AMD 65nm in production
See FAB Tech report here, 65nm AMD procs are coming sooner.
This site reported a dual core duel between Athlon 64 X2 4200 (2.2GHZ, 2x512KB) and Conroe 6300 (1.86GHZ, 2MB), X2 4200 wins except the SuperPi bench. The X2 4200 had 18% clockspeed advantage, that's enough to defeat Conroe. In particular, in the HDBENCH floating point test, the X2 4200 won by a 35% margin. From SpecFP_rate2000 scores published by SUN, a Socket F Opteron 2.6GHZ beats a Woodcrest 3GHZ by 38%. Intel's Core 2 architecture is not well suited for HPC applications.
With 65nm parts, I expect AMD to achieve a clockspeed increase of 25%, sufficient to decisively defeat Conroe XE 6800 in all benchmarks by brute force. Coupled with architectural enhancements with K8L, AMD may achieve a whopping 60% performance/core advantage over Core2 on integer performance. On floating point, the K8L core will be at least 2x faster than Core2.
Tyan launched six socket F 1207 boards.
This CNET page has some nice DELL quotes over the years. From saying AMD was fragile to embracing AMD64 across the board, Michael Dell finally learns a tough lesson as his company's future looks increasingly bleak.
posted by Sharikou, Ph. D. @ 8/18/2006 01:12:00 PM
102 Comments:
How cool, now we know a little about the deal was with Dell. Dell steps up to the plate and gets a slow pitch from AMD, then Dell knocks the ball out of the park with new stuff that know one else has yet.
that's the beauty of this business. it's like you're a game publisher with a sleeper hit title that will sell like hot cakes. you don't want to hype it too much, then bam, you release it. the rest is history. this is what i've been waiting to hear and i kinda got the hints from the doc himself. i hope i'll have some extra money this holidays to upgrade my rig. xmas will be fun this year. once again, amd will save dell as they did for ati. i wish amd's stock would split, hopefully my 85 shares will multiply :).
To the first poster:
I wouldn't get my hopes up about the 65nm K8's (atleast the ones introduced this year) performance wise, chances are these will not have any performance improvements over their 90nm counterparts aside from lower power consumption, which is of course something important as well.
The reason I think this is because at the time AMD can put out these new 65nm K8 parts, K8L will be knocking at the door, and if I were AMD, I'd try to bet my money on that horse.
Moving to 65nm right now seems to be to increase capacity more than anything. I think for real performance improvements we'll have to wait for K8L. I leave out 4x4 because the performance improvements I'm referring to are ones strictly inherent to superior CPU architecture design.
Anyway, since this is an uneducated guess, I'd like Sharikou or anyone else knowledgeable enough to share some thoughts about performance increases for 65nm K8 aside from power consumption, given the light of upcoming K8L.
Note that this post is not intended as a flame or anything, just bringing conversation. With all the name calling and bashing of each other's opinions around here, a disclaimer seems appropriate.
I'm really curious as to exactly what amounts of power and clock improvements AMD will get with 65 nm K8. That, along with the new competitive pricing... I'm betting that Dell's seemingly late announcement of AMD offerings isn't as of a fluke as some people believe!
*Yawn* Oh..uh..wha....AMD FINALLY making it to the 65nm party? Sheesh, bout time...AND finally getting into bed with Dell...Greeeat timing there..Dell is sinking and wanted a flotation device named AMD. This would have been a great idea, TWO FREAKING YEARS AGO!
hmm seems the Tyan board could be the ultimate gamer Platform. if you can afford it that is:P
http://www.theinquirer.net/default.aspx?article=33780
Consider 2 x 2,6 ghz opteron and 4x GeForce 7950GX2:D.
mork, I don't have all the answers but for the last one
3 - Any clues on what chipsets will support Quad Core?
It's obvious you haven't been following AMD's technology. :) The HT architecture means that it is pretty much the same chipset. AMD doesn't force you to change motherboards and chipsets every 6 months or so (*sob* *sob* tell that to my 955X uber motherboard that was top of the line not too long ago)
In fact, many AM2 boards use the same exact chipsets as older Socket 939 motherboards.
And AM2 boards out there today are already quad core ready. At most, you'll need a BIOS reflash to support the new CPUs.
ps this was the same case moving from single core to Athlon64 X2s on Socket 939, so you got backwards compatibility for over 2 years, rather than just 6 months
Best feature of the new Tyan boards:
"Four onboard 5-phase digital VRMs"
Tyan's previous boards had stability issues and other problems because of inferior power systems on the board.
It looks like Tyan has fixed things.
65nm production was scheduled to begin mid-2006 so AMD is right on schedule. However, with Fab 36 not reaching 50% production until Q1 2007 according to AMD themself, their levels are very low right now and it's doubtful anybody much less Dell will be shipping any 65nm chips until the official December launch.
"hmm seems the Tyan board could be the ultimate gamer Platform. if you can afford it that is:P
http://www.theinquirer.net/default.aspx?article=33780
Consider 2 x 2,6 ghz opteron and 4x GeForce 7950GX2:D."
You have to remember that Charlie over at the Inquirer is something of a retard.
The Tyan boards mentioned are all non-SLI.
So if you really want a non-SLI motherboard for gaming, go ahead and plunk your dollars down for one of those Tyans.
A better bet is to get a real gaming motherboard with SLI and put TWO 7950GX2 cards in it (and this gives you the most Nvidia's drivers can deliver on, Quad-SLI).
Of course if you have a CPU limited game, then go for a quad-Opteron board and ONE 7950GX2. If your game actually supports 8 threads. Which is unlikely.
Or just get a PS3 :-)
Every chipset can support AM2 quad core. This isn't intel silly. The memory controller is in the CPU. Hello, any AM2 socket will support anything up until AM3 is dead. Kind of a silly question. All it is is from a bios update. Has nothing to do with chipsets.
Luckally AMD doesn't have the same problems with chipsets like intel does because of chipset and memory controler problems. Since the memory controler and north bridge controler are in the cpu, the sockets will support any cpu from AM2 and AM3 cpus into AM2 sockets regaurdless of the chipset.
And 65nm AMD's have been under production in fab 30 since may 06 because of a transission they did. And fab 36 as of about a month ago give or take. We will see 65nm's in the 4th Q of 06. And the advantages of 65nm will be the Overclockability boosts. Just like what intel got, expect atleast a 30% difference in better OC's. And best of all way less volts heat and watts.
Theres always a performance difference in some way or another. Besides they added a 4th FPU in the 65nm construction so there will be atleast some improvments to performance. We all know the die shot of REV G so don't start with its not going to have a performance difference. It has 4 darn FPU's pointed out all over the net. Thank you.
Moving to 65nm right now seems to be to increase capacity more than anything.
Capacity is key. Even in 1Q07, Intel will be producing 60% legacy chips. So as long as AMD can beat Netburst, it can laugh all the way to the bank with massive flood of X2 chips. Remember, AMD's CPU only costs $40 each. By mid 2007, FAB36 will be fully converted into 65nm...At that point, 60% of AMD's chips will be produced at 65nm.
Also, AMD should be able to bump clockspeed by at least 30% at 65nm... We can expect 3.4 GHZ X2 chips.
let us not forget,a so called slow Athlon64 x2 running at 3.4ghz would be on a performance scale of intel's latest and greatest. scary,just to think they could possibly match the performance of the x6800 without the k8L tweaks.
Yo Pretender...
Did you read the link. The bulk of Dells AMD is coming from 90nm as 65nm is only starting to ramp. What is amazing is that it takes very little volume of 65nm to make a million unit run rate. Things must be going badly.
A window tour of a fab showing tools and robots are moving is no big deal. Without more detailed understand AMD could claim it was 45nm and the Dell monkeys like you would lap that cumm and not know the difference.
Some PhD prenteder wrote
"Capacity is key. Even in 1Q07, Intel will be producing 60% legacy chips. So as long as AMD can beat Netburst, it can laugh all the way to the bank with massive flood of X2 chips. Remember, AMD's CPU only costs $40 each. By mid 2007, FAB36 will be fully converted into 65nm...At that point, 60% of AMD's chips will be produced at 65nm.
Also, AMD should be able to bump clockspeed by at least 30% at 65nm... We can expect 3.4 GHZ X2 chips. "
Sorry the pretender he is wrong wrong wrong.
INTEL has validated and released Conroe, Woodcrest, and Merom. The majority of their 65nm capacity is running Core2. Their 90nm capacity is running some old PentiumD and Chipsets... They are pumping out Core2 that whips AMD chips on every benchmark there is.
Secondly a dumb shrink in principle speeds up the transistors by ~20%. Since the 65nm parts aren't a new architecture nor deeper pipeline they won't get 30% frequency pop it simply isnt the physics of devices and logic design. Now if you had a real PhD and any knowledge of VLSI design you wouldn't have typed something so stupid. but you also talk about how INTEL will go bankrupt so its no surprise you again and again post stupid conclusions on wrong analysis.
Why don't you go to school and get a real Phd. Maybe you'll learn something new and post some real conclusion...
Even in 1Q07, Intel will be producing 60% legacy chips.
While they may still have stock of legacy chips and may still be selling them I don't think they would be producing anymore of them. Intel has just EOL'd almost all the 90nm chips and also the 930D and 940D. The main legacy chips that could possibly be in production in Q1 are the 925D, 945D, and some Celeron Ds.
That's nice, AMD is playing catch up ever since they've reversed engineeered an 8080A, http://www.amd.com/us-en/Weblets/0,,7832_10554_10531,00.html.
Ohh, I forgot this was in the past and we now looking at the future.
I still find the timing very wierd. Between Merom, Conroe, and $93 805Ds there isn't a reason to go AMD in desktops. 2S servers Woodcrest is more than competitive with Socket F and even in 4S Tulsa seems to do okay given it's huge 16MB shared L3 cache. Now K8L will probably be a different story but that's next year. Dell avoids AMD for years when they have the performance lead and when Intel finally regains it in multiple segments Dell goes to AMD. Their timing is awful.
Hopefully they can get it out soon. Tigerdirect's top 10 is showing a negative trend for AMD.
1. Intel Core 2 Duo E6600 2.40G
2. Intel Pentium D 840 3.2GHz /
3. AMD Athlon 64 X2 4800+ / 2MB
4. Intel Core 2 Duo E6700 2.67G
5. AMD Athlon 64 X2 4600+ 2.40G
6. Intel Core 2 Extreme X6800 2
7. Intel Core 2 Duo E6700 2.67G
8. AMD Athlon 64 FX-62 2.80GHz
9. Intel Core 2 Extreme X6800 2
10. AMD Athlon 64 X2 4600+ / 1MB
While they may still have stock of legacy chips and may still be selling them I don't think they would be producing anymore of them.
No. Intel is stuck with Netburst until 2008 due to its outdated FAB automation process -- Copy exact. But, by then, Intel will be BKed already.
The PhD pretender is again wrong..
he wrote "No. Intel is stuck with Netburst until 2008 due to its outdated FAB automation process -- Copy exact. But, by then, Intel will be BKed already."
He put two wrongs and came to anothe wrong conclusion. For those that are educated it is so clear how wrong he is. It is entertaingin to see someone claim to be a PhD but continuing to post laughable wrong conclusion.
Cope Exactly that Intel has published at various conferences is the method of using the same proccss, tools, operation, monitor methods on a process no matter where the fab is ( Isreal, US, Ireland ).
Automation is part of it but actual not what Copy Exactly is about. What is more funny it has nothing to do with what you make with the process. Copy Exactly in Ireland, Arizona and Oregon on 65nm means the tools, process, monitors, limits, targets are matched. What you run on the proces could be a Cedar Mill chip, a Yonah Chip or the new Merom, Conroe, Woodcrest. It has NOTHING to do with copy exactly. You don't even know what AMD's APM is nor INTELs copy exatly is. Copy exactly insures you get the same high yield, and speedy parts from all 3 factories.
Keep spewing the shit.. its worth a good laugh on a Friday...
Are you really that stupid?
Hopefully they can get it out soon. Tigerdirect's top 10 is showing a negative trend for AMD.
It's interesting that AMD's top selling model is the discontinued X2 4800+. Looks like last minute buyer demand.
Dell avoids AMD for years when they have the performance lead and when Intel finally regains it in multiple segments Dell goes to AMD. Their timing is awful.
They held out as long as they could. They were getting a sweet volume discount deal on Intel, and if they went AMD they would have lost that. They would have had to charge a higher price on both AMD and Intel based PCs.
Now that Dell isn't getting as sweet a deal (if there is one at all) there isn't any reason not to use AMD. Their timing is awful - because they should have foreseen what the price war would have done. If they had been smarter they would have "gone AMD" 4 months ago so that by the time the price war got in effect they would actually have AMD PCs out and making corresponding high sales numbers. By now they just get AMDs leftover chips after everyone else who AMD has a better relationship with gobbles them up.
Hopefully they can get it out soon. Tigerdirect's top 10 is showing a negative trend for AMD.
1. Intel Core 2 Duo E6600 2.40G
2. Intel Pentium D 840 3.2GHz /
3. AMD Athlon 64 X2 4800+ / 2MB
4. Intel Core 2 Duo E6700 2.67G
5. AMD Athlon 64 X2 4600+ 2.40G
6. Intel Core 2 Extreme X6800 2
7. Intel Core 2 Duo E6700 2.67G
8. AMD Athlon 64 FX-62 2.80GHz
9. Intel Core 2 Extreme X6800 2
10. AMD Athlon 64 X2 4600+
Half of the CPUs in that top 10 are $600 or more. Either Tigerdirect caters to a very strange crowd, or thats just a completely unrealistic profile. If they actually sell this many high end CPUs compared to other CPUs they aren't making very much money off of CPUs.
By AMD's own numbers X2-FX CPUs account for only 0.2% of their Athlon volume. No chip that accounts for 0.2% of the total volume is going to be selling that high.
On the other hand, Intel itself hadn't made the E6600 and E6700 available until fairly recently, their first shipping chips were E6300's and E6400's.
And lastly, Tigerdirect is a rip-off, if you are buying a high end Core 2 Duo chip you can get it $50-$100 cheaper at Newegg. Newegg by the way has its top5 selling chips as all Athlon's under $200, 3 of which are X2s. That is much more realistic.
http://129.15.202.185/athlon_rev_g/wtf_mates.html
http://www.theinquirer.net/default.aspx?article=32322
Rev G will not have 4 FPU's, they will have 4 decoders and more buffers for the ability to reach higher clock speeds. Rev H is the huge difference because its K8L.
So we should see 4ghz more common in AMD's 65nm tech according to this data. This will help, there should be some extra performance differences in over all performance just by going to 65nm.
You said...
Did you read the link. The bulk of Dells AMD is coming from 90nm as 65nm is only starting to ramp. What is amazing is that it takes very little volume of 65nm to make a million unit run rate. Things must be going badly.
I said...
As badly as it is for conroes sad yeild rate? Don't make me laugh. This just means AMD is ahead of the release date. We should see them turning up in a few months. Before dec 2006 even. Atleast it won't be as hard to get as it is for sad old conroe. Once you post a conroe link it will be sold out so don't even bouther.
The prob with intel is as soon as conroe comes out there is not enough to supply the people with back orders so there is a long line. Example I had the release date on conroe change on me 5 times. AMD's sell so good they make enough but are sold out as soon as they reach the door and shipped to all the people that bought them.
The difference is AMD can keep up but its so much so where they almost can't because the sales already happen as soon as it hits the stores. Wile on the other end you have conroe in so much demand with a lack of yeild it hurts intel and the retailers who sell them because people don't like to wait. You end up with a serious problem with only back orders for weeks if not months on end. AMD doesn't have fab problems like intel does. Too much P4 crap as simple as that.
"Besides they added a 4th FPU in the 65nm construction so there will be atleast some improvments to performance. We all know the die shot of REV G so don't start with its not going to have a performance difference. It has 4 darn FPU's pointed out all over the net"
Funny, I must be using another Internet since I couldn't find a link that would say that. Are you sure you are not mixing up K8 and K8L?
"No. Intel is stuck with Netburst until 2008 due to its outdated FAB automation process"
By/at 2008 Intel should be moving to 45nm and producing only Core2's. Later that year there might be a new core availiable.
"by then, Intel will be BKed already."
Are you willing to bet on that?
3.4GHz X2s on 65nm? After heavy overclocking perhaps, I certainly don't expect any retail 65nm X2 chips to run at 3.4GHz.
From the latest AMD roadmap the fastest 65nm X2 I see is the 5200+, which will be well short of 3.4GHz.
AMD has nothing to counter C2D until K8L is out mid 2007.
Ignorant consumers will continue to buy Celerons and P4s and PDs simply because they know no better, and Intel is the 'safe' brand.
You can blaim AMD's marketing department, or lack of one, for that.
http://www.theinquirer.net/default.aspx?article=33793
Well sharikou you are mentioned on The Inquirer again. The "ultimate AMD Fan Boi". I guess some fame even bad fame is better than no fame. ;)
Rev G will not have 4 FPU's, they will have 4 decoders and more buffers for the ability to reach higher clock speeds.
That is not true. Rev G is a simple optical shrink. The blue 65nm chip that everyone keeps talking about is a prototype for some K8L features and not an actual product. AMD has confirmed this themself. What Rev G brings is lower power consumption not even higher clock speeds since the fastest launch chips are going to be a X2 4600+. The FX doesn't even use 65nm all through 2007.
One of the articles wrote:
"Intel takes a very definitive lead in performance and power management at 65 nanometers...If we thought AMD was going to be supercompetitive in the spring and fall of next year, we'd be introducing AMD products right now," Michael Dell, October 2005.
And one year lateer... shows you how wrong DELL was about predicting and evaluating its own industry.
Shareholders should make Michael and Rollins resign from DELL.
The only reason I posted the top 10 at tigerdirect is because Sharikou was posting it in his previous blog entrys to show that AMD was kicking some serious Intel butt.
"By/at 2008 Intel should be moving to 45nm and producing only Core2's. Later that year there might be a new core availiable."
You Intel people make this sound so simple and elegant? I'm not about to educate you here, but I will at least send you to various references so that you can learn a few things:
Depreciation
Amortization
Write off
Not maximizing your investments is bluntly STUPID! For Intel to be moving so fast is a move of desperation. What effects will that have on the company? Well you know when Sharikou was talking about DELL's Shareholder Equity? Well for one, that will decrease Intel's shareholder equity in the long run. Not very good if your an Intel investor.
What other effects will this have? In the long run, it will increase Intel's cost of manufacturing processors?
So regardless of what the current cost may be for Intel/AMD to produce processors, let me assure you of this fact:
1) Evidence shows that Intels processor manufacturing costs are (and keep) rising, while
2) Evidence shows that AMD's processor manufacturing costs are constantly being lowered.
In regards to Sharikou' comment:
"by then, Intel will be BKed already."
Ho ho said...
"Are you willing to bet on that?"
Although most within the AMD camp would be honored to take you up on such challenge let me inform you (and any others who have asked Sharikou to bet) that it makes absolutly no sense for us or Sharikou to bet with you Intel people on such subject.
Here is why:
1. Accepting the bet:
If Sharikou were to accept your bet, say of $1000, then there are only two possible scenarios...
a) You are right... and Sharikou is $1000 poorer by the end of the year.
b) He is right and Sharikou is richer by $1000 at the end of the year.
OR
2. Putting the bet money at work:
He takes the $1000 and buys a straddle on Intel, for which there are then two scenarios...
a) You are right... Sharikou makes a ton of money.
b) You are wrong... Sharikou makes a ton of money.
My point is two fold:
1) Learn to invest and you will NEVER wast your money on bets again!
2) You can make a fortune from this AMD/Intel uncertainty without really caring who comes out on top (although I already know that it will be AMD).
" Anonymous said...
*Yawn* Oh..uh..wha....AMD FINALLY making it to the 65nm party? Sheesh, bout time...AND finally getting into bed with Dell...Greeeat timing there..Dell is sinking and wanted a flotation device named AMD. This would have been a great idea, TWO FREAKING YEARS AGO!
3:06 PM, August 18, 2006 "
hello? intel moved to 65mm this year and got ZERO improvement in everything ( like less than 5-10% less heat. )
nice to see intel fanboys still spitting crap as usual.
" Anonymous said...
Yo Pretender...
Did you read the link. The bulk of Dells AMD is coming from 90nm as 65nm is only starting to ramp. What is amazing is that it takes very little volume of 65nm to make a million unit run rate. Things must be going badly.
A window tour of a fab showing tools and robots are moving is no big deal. Without more detailed understand AMD could claim it was 45nm and the Dell monkeys like you would lap that cumm and not know the difference.
7:33 PM, August 18, 2006 "
so now you bitching about AMD's ramping process (wich is actually faster... but yet a bit late ) and yet you lick intel's shoes and say awes of their "impresive" conroe ramping?
pathetic..
"Keep spewing the shit.. its worth a good laugh on a Friday...
Are you really that stupid?"
you look so serius.. smart... and intelligent with these insults..
feeling like dick cheney lately?
go see snakes on a plane plz..
From one of the inq arcticles:
"This process encompasses embedded SiGe with dual stress liner and stress memorization technology on silicon on insulator- or e-SiGe with DSL and SMT on SOI for those so inclined. AMD and IBM have stated publicly that this technology allows for a 40% faster switching transistor than from a standard 65 nm design without all the three letter acronyms (TLA’s)."
So thats where you get the 40% higher clock nubmers!
continued...
"In a complex design like a CPU this could mean a theoretical 50% overall clockspeed increase going from AMD’s 90 nm process to AMD/IBM’s 65 nm process all the while staying within the same power envelope."
huh? I hope they know that thansistor switching speed has little to do with CPU speed. Transistors that make up CPU's switch way >10GHz. The biggest obstacle in increasing the clock speed is the wires between the transistors.
Of cource they can increase the speed a bit since the transistors switch a bit faster => less leak => less heat.
"You forgot 5400+ and 5600+"
You know those are 90nm models and not 65nm, do you? Highest 65nm models seems to be 5400+ that is out Q3 07
Assuming AMD keeps the same speed increase steps as it has now I expect those FX64 and 66 to reach 3 and 3.2G in Q4 06 and Q3 07. x2 line should max out around 3GHz in Q1 07
Jeach! said...
"Not maximizing your investments is bluntly STUPID! For Intel to be moving so fast is a move of desperation."
How do you see desperation, they have always had the manufacturing lead over AMD, and has usually been a 2 year cycle (Intels manufactoring process).
Do you think that the factories using 65nm are going to be converted to 45nm?
If so you would be incorrect.
Here is a quote...
"Intel's initial 45nm development efforts are under way at the D1D facility in Oregon. In the future, we also plan to manufacture 45nm products at Fab 32 in Arizona and Fab 28 in Israel, two high-volume fabs currently under construction."
January 2006
Also couldn't they use the 65nm factories for chipsets and what ever other 65nm chips, they do have 3 of them...
Anonymous said...
"Wile on the other end you have conroe in so much demand with a lack of yeild it hurts intel and the retailers who sell them because people don't like to wait."
I think its lack of ramp, not yield.
If you could show a link about this I would believe yield, but we all know that finding out Intel or AMD yields is very hard to do.
As far as waiting, your right, but the E6700 beats an FX62, so I will gladly wait.
"hello? intel moved to 65mm this year and got ZERO improvement in everything ( like less than 5-10% less heat. ) "
I've been cheated! They must have sold me a X2 isntead of 920. It is impossible for that CPU to run at 25C in idle and 40C in full load wit zalmantec 9500 cooler set to minimum. As it runs @4GHz with slightly increased voltage it might even be a 65nm part!
[/sarcasm]
enumae said ... Also couldn't they use the 65nm factories for chipsets and what ever other 65nm chips, they do have 3 of them...
Yes, you are right. Intel has been (started with broadwater, there was a small gap before where chipset was on N-2 for a brief period) using the N-1 strategy on the chipset (read as CPUs use Nth gen process, chipsets use the N-1th)
by having a broad portfolio of chips allow intel to maximize its factory usage. As you can read from the Q2 briefing. Intel has kept the factory fully loaded in Q2.
AMD is likely to do the similar thing when it has enough capacity, it will produce ATI chips at N-1, and CPU at Nth.
http://www.fabtech.org/content/view/1390/74/
"The reports authors said that yields were at 50 percent at Chartered's Fab7 but sometimes even better than those achieved within AMD's own fab!"
Over 50% sounds promising -_-
http://www.tgdaily.com/2006/08/18/amd_4x4_on_course_for_second_half/
""good bet" that items in that product line - whose final name is to be announced - will be available at Christmastime."
Yay 4x4 for Christmas:)
Michael has not learned sh#@. Consumers will not buy something that looks like crap because it is cheap!
I had a dell laptop and in 2 years, I replaced the AC adapter 3 times, my CDROM keeps disconnecting itself, and my one USB port is broken. It is about time for Dell to wake up and realize his computers look and act ugly. He needs to know that PCs are becoming consumer devices and most people choose their appliances based on how cool they look and fit with other furnature.
AMD will not help Dell. It is not about what is in, it is about what is out.
THe AMD Dell deal could prove to pull them both down. Time will tell.
lets make the bet more interesting: If Intel goes bankrupt, Intel fannies will buy AMD chips. If Intel does not go bankrupt, Sharikou would have to show the world his PhD.
" Anonymous said...
" Anonymous said...
*Yawn* Oh..uh..wha....AMD FINALLY making it to the 65nm party? Sheesh, bout time...AND finally getting into bed with Dell...Greeeat timing there..Dell is sinking and wanted a flotation device named AMD. This would have been a great idea, TWO FREAKING YEARS AGO!
3:06 PM, August 18, 2006 "
hello? intel moved to 65mm this year and got ZERO improvement in everything ( like less than 5-10% less heat. )
nice to see intel fanboys still spitting crap as usual.
1:59 PM, August 19, 2006 "
Move out of the north pole, not all the world lives in your frozen throne.
and netburst on 64nm = almost identic to 90nm, only better overclock ( in a few cases )
you're of the very very few who I've seen claim these are colder..
"Ain't exactly seen a flood of upper level cornrows...."
http://www.pricegrabber.com/search_attrib.php/page_id=42
Upper Conroes control 3/7 of pricegrabber's popularity list.
6300 is #1 on Zipzoomfly
"Ain't exactly seen a flood of upper level cornrows...."
http://www.pricegrabber.com/search_attrib.php/page_id=42
Upper Conroes control 3/7 of pricegrabber's popularity list.
Intelers failed to realize the problem I have long discussed:
The better and more popular the Conroe is, the faster Intel will BK.
When I said Conroe is the last straw that pushed DELL to AMD, it was not because Conroe is not good, but because it's good: 40% faster than Netburst -- 95% of Intel's production and 99% of inventory.
Get it??
It's called Osborne -- a company went BK because of a next gen. Back in 2005, I told AMD folks, as long as AMD does not hype its next gen, Intel will be the only one suffers from Osborne.
They're running the company. You're not. I'm sure they could've stopped production earlier if they needed too:)
Dell and others still have to plug in their remaining motherboards with something and there isn't anything Core below $183. And China and India are stilll exploding, someone has to buy them.
Anonymous said...
"Move out of the north pole, not all the world lives in your frozen throne."
Here is a link of an old computer I had, it is a single core but is both lower temps and good overclocker.
As I do not live in the North pole this would seem to be a very good sign of 65nm, for AMD or Intel.
Intel 651 @ 4.5GHz
AMD 65nm Generation v1 (2006)-> v3 (2007) :
- 3rd génération of strained-silicon (DSL + Stress Memorization Technology + e-SiGe + SOI)
- Addition of Nickel Silicide
- 4th génération of low-k dielectric stack
"No. Intel is stuck with Netburst until 2008 due to its outdated FAB automation process -- Copy exact."
Harikou - This is the 2nd or 3rd time you've made this ignorant statement. Both Core2 and Netburst are produced on 65nm technologies! Intel doesn't use 1 set of tools for Core2 and 1 set for Netburst. The only major difference in the fabs are the reticles/masks used in the litho tools (which can get changed out in less than 1 hour). Copy exact ensures the tooling is matched within a fab and from fab to fab within a given process technology, not for specific product!
You're not really this ignorant about manufacturing are you? Or are you just making stuff up trying to make Intel look bad again?
Intel doesn't use 1 set of tools for Core2 and 1 set for Netburst. The only major difference in the fabs are the reticles/masks used in the litho tools (which can get changed out in less than 1 hour).
Intel use the same kind of tools for the Core2 and Netburst--that is for sure. Intel doesn't make the tools, it buys them. The problem is Intel's copy exact means one set of tools can only be used for one kind of chips. That's why Intel can't do Itanic at 90nm or 65nm. Because Itanic is so low volume, Intel can't dedicate one set of tools to make a few dozen chips per quarter. Both xscale and Itanic are stuck at 130nm for this reason.
AMD is different, with APM3.0, all AMD chips, Opteron, Athlon, Sempron, Turion, rev E, rev F, rev G, rev H, are made on the same set of tools.
"intel moved to 65mm this year and got ZERO improvement in everything ( like less than 5-10% less heat. )"
If you had any clue about manufacturing there is more to manufacturing transitions than your performance comparisons about Netburst improvements. Cost/die? Ever think perhaps the Netburst design could not take advantage of the improved technology performance? (as measured by publicly disclosed transistor performance data).
Correct me if I'm wrong here, but do AMD's 665nm desktop roadmap show a 40% clock improvement? a 40% heat improvement? The numbers I've seen show similar clockspeed and TDP's! Like Intel, AMD will be able to capitilize on the benefits on future product designs (Intel - Core 2, Nehalem, AMD - K8l, K10?); but both get a pop in terms of die size scaling and manufacturing costs.
intel moved to 65mm this year and got ZERO improvement in everything ( like less than 5-10% less heat. )
Actually, Intel's 65nm Netburst has substantial power reduction compared to 90nm. The pentium D 840 was a heat monster, at least 173 watts, but Pentium D 945 is better, approximately 130 watts. So there was improvement from shrinking. This is expected, because smaller size means smaller capacitance.
However, AMD's 65nm will be much much better than AMD's 90nm. It has some nice tricks in transistors that is specific to SOI.
"The problem is Intel's copy exact means one set of tools can only be used for one kind of chips."
This is simply wrong! Intel has how many product types? You're saying that have a different set of manufacturing tools for each product type - I can't believe you are this stupid! The different product types use different RECIPES on the fab tools not different tools.
I had no idea you were serious about this, I though you were just spinning the anti-Intel rhetoric again. Now I realize you just have no idea how a fab works...
So what you are saying is that over the next 2 quarters Intel will be completely retooling all of their 65nm fabs as they migrate from Netburst to Core2? Ha, too funny...
This is simply wrong! Intel has how many product types? You're saying that have a different set of manufacturing tools for each product type - I can't believe you are this stupid! The different product types use different RECIPES on the fab tools not different tools.
It is Intel who is stupid. Their copy exact approach is retarded. Ideally, you have recipes with transistors at specific node, this is FAB technology, you make transisitors on wafers. This should be orthogonal to your chips. Once you nailed the process, you should be able to making any transistor combinations with the process -- Turion or K8L. That's what AMD is doing. Once it nailed the 65nm, it can make all of its chips under that node.
Intel is stupid. They are like writing bad code, copy&paste, mixing different indepdent technologies. The result is they are screwed. That's why they can't swap out the masks in their Pentium D FABs to make Conroes. Instead, they need a new FAB to make Conroes.
May I just ask, how long does it take to search for some random Chinese site to 'prove' the X2 4200+ is 'superior' to the E6300?
You must have a lot of spare time Sharikou, but I guess your blog proves that already. Do you have a day job by any chance? Or are you still living off your Intel severence payout?
With similar price points between a 4400+ and E6300, it seems AMD knew exactly what to expect from Intel. The only reason to go for CONroe would be very high-end. But even then with AMD's platform and drop in quad-cores coming. It looks like the fan-boi's have jerked themselves off over catching up with K8. K8L is going to do to CONroe what K8 did to Netbust. Guess history repeats itself.
Sharikou said ...
That's why Intel can't do Itanic at 90nm or 65nm. Because Itanic is so low volume, Intel can't dedicate one set of tools to make a few dozen chips per quarter. Both xscale and Itanic are stuck at 130nm for this reason.
...
It is Intel who is stupid. Their copy exact approach is retarded. Ideally, you have recipes with transistors at specific node, this is FAB technology, you make transisitors on wafers. This should be orthogonal to your chips. Once you nailed the process, you should be able to making any transistor combinations with the process -- Turion or K8L. That's what AMD is doing. Once it nailed the 65nm, it can make all of its chips under that node.
Well, another industry (technogically and economically) outsider point of view.
chips are designed with Fab process in mind. you cannot produce the chip using the new process without re-design (mainly the circuit part) and re-validate. Different process has different electrical characteritic.
And to the contrary of what Sharikou said, Copy Exactly gives Intel a signicant advantage over AMD in terms of High Volumn Manufacturing. I'm not bashing APM here. APM is good for AMD for its size. Copy exactly fit Intel for its size.
As what i said before, Intel has been using the N, N-1 or even N-2 strategy in manufacturing its chip to fully utilize its factoriessss. The N-1 and N-2, as proven technology, also give lower development cost to the chip
Randy said...
"With similar price points between a 4400+ and E6300, it seems AMD knew exactly what to expect from Intel."
That is not a tough one to figure out, expected prices were laid out a few monthe before AMD's price cuts.
"But even then with AMD's platform and drop in quad-cores coming. It looks like the fan-boi's have jerked themselves off over catching up with K8."
I am sorry to burst your bubble but Kentsfield is also drop in quad core.
Also it will be out about 6 months before K8L (server chips), and its only speculation on how long after that we see desktop versions.
"K8L is going to do to CONroe what K8 did to Netbust. Guess history repeats itself."
We are all hoping that it does, as it is the people who benefit from AMD and Intels competition.
But to keep things in perspective, we have seen no samples, real world and projected performance are two different things.
chips are designed with Fab process in mind.
Of course. But AMD designs its chips for its process. Intel designs its chips for its process. Intel won't design a chip for AMD FAB36 or TSMC. So, this is not even a question.
I am truely amazed at how retarded the PhD pretender is and the flies that hover around his turds are. They again prove it to us...
THank god they neither work for INTEL nor AMD.. Take pity on their employer at their stupidity and incompetence.
They wrote:
"The problem is Intel's copy exact means one set of tools can only be used for one kind of chips."
"The problem is Intel's copy exact means one set of tools can only be used for one kind of chips. That's why Intel can't do Itanic at 90nm or 65nm. Because Itanic is so low volume, Intel can't dedicate one set of tools to make a few dozen chips per quarter. Both xscale and Itanic are stuck at 130nm for this reason."
Having worked at all three companies you flies are wrong.. dead wrong was it clear enough.
You are wrong..
LOL too bad none of you have a real job... like me..
bye now
Sharikou, Ph. D said...
"With 65nm parts, I expect AMD to achieve a clockspeed increase of 25%, sufficient to decisively defeat Conroe XE 6800 in all benchmarks by brute force."
I ran the numbers and for an average of 5% improvement over an X6800, they need the full 25% clock increase, it would take them to 3.5GHz.
At the same time I ran the numbers of the 3.5GHz FX62 vs a 3.2GHz X6800. The numbers are reversed, 4% advantage for Conroe.
Looking at there road map, it would seem the FX series all the way through Q3 2007 will be 90nm, and given the increases they usually use (200MHz) that would make the FX66 a 3.2Ghz processor, and unable to claim the crown from Intel.
If I am missing something let me know.
Thanks
If anyone would like to see the numbers look here
Note: the basis for the benchmarks is Anandtechs benchmarks.
Sharikou, im just wondering at roughly 3.4ghz,these kind of speed increases should ulitilized even more of that untapped bandwidth on that great HT architecture,hell we can be looking at an additional 5 to 10 percent performance increases.
"If Intel goes bankrupt, Intel fannies will buy AMD chips."
That's certainly interesting. I myself cannot imagine what I'd have done had AMD bankrupted before, say in 2002 (I've never feared it though). How boring the PC market would be if there was only Intel making the processors! I don't think anyone would like to be domincated by Netburst and Itanium!
On the other hand, what would it be if Intel goes bankrupt? Will the PC industry be as/more interesting? Will AMD keep innovate on CPUs (probably not)? Will those Intel admirers stop buying new PCs and switch to.... uh.... Mac using Power? It would be a really interesting scenario. ;-)
http://www.theinquirer.net/default.aspx?article=33795
Is it me or does this read a move by a sad, desperate company that is totally afraid?
On the other hand, what would it be if Intel goes bankrupt? Will the PC industry be as/more interesting? Will AMD keep innovate on CPUs (probably not)? Will those Intel admirers stop buying new PCs and switch to.... uh.... Mac using Power? It would be a really interesting scenario.
It wouldn't be "interesting," it'd be downright frightening. There would be no-one to stop AMD from pumping prices to absurd levels and only releasing new speed bumps once every year. Even if the DoJ launched an investigation, it'd probably take 4 or 5 years to get anything meaningful done.
And I know some people like to claim that AMD are perpetual do-gooders and would continue to work as normal in a monopoly, but consider this quote from George Lucas:
"I like to think that the best type of ruler would be a kind of benevolent dictator, as he would actually be able to get things done. Sadly though, there's no such thing as a benevolent dictator, and I don't think there ever will be."
I was reading some posts a few months ago regarding K6's and the upcoming athlon classic architechture or something like that
7 years later, all Intel got is a faster Pentium 3 on a faster bus. Poor Intel.
I <3 Sharikou's made up performance numbers =P
Azary Omega said...
"Son, remember that Anand got his buck for what he did and you just a... well i dont think intel will cut you anything no mater how hard you promote em."
Just a ...?
Say what your thinking, don't punk out, you arn't gonna hurt my feelings, just so you know, anytime someone here points out something theoretically negative about AMD... FAN BOI!!!
It would seem people are very closed minded, don't you think?
I am not promoting Intel, just running numbers to debate the Dr. statements.
Would you rather no body do this, and just go on believing what ever he says?
Pick a site and I will run them again, I don't care, I am not trying to show bias either way.
The numbers clearly show AMD's future speeds and how they supposedly would perform, I say supposedly because I am just scaling.
Stoolman at Xtremesystems, the place to be seen for any famous overclocker, got a sample of the Kentsfield E6700, which is basically a quad-core version of the E6700, clocked at 2.67GHz. Well, that could be it, but wait. He took the chip and overclocked it to 4.75GHz. Obviously, he used some hardcore solutions like liquid nitrogen, but he proved that teh chippery can reach over 4GHz. His benchmarks did not show a huge difference when compared to a dual dual-core 3GHz Woodcrest, however.
"The scores are relatively poor, the platform considered. I also did optimizations towards 3d mark 06, ram tweaks, real time and relatively nice clocks on the 7950, air considered, but the cpu acts more like a slow x6800 than a potent quad core..." - Stoolman
Looks like newegg is doing the dirty work for Intel now ..
http://www.theinquirer.net/default.aspx?article=33825
"Would you rather no body do this, and just go on believing what ever he says?"
Yes, Enumae YES!
Obviously, he used some hardcore solutions like liquid nitrogen, but he proved that teh chippery can reach over 4GHz.
Intel demoed 5GHZ Pentium 4 in 2001. But it had to cancel 4GHZ P4 in 2005.
Going forward, you will see that Intel will repeat history, the only way to boost Conroe performance will be hiking up clockspeed. AMD 4x4, on the other hand, is always 80% faster.
netrama said...
Looks like newegg is doing the dirty work for Intel now ..
http://www.theinquirer.net/default.aspx?article=33825
That article cracks me up. "It would have been better had the ad read 'The worlds best Desktop, Mobile, 1P server, and DP server processor has just arrived'"...
LMAO.
7 years later, all Intel got is a faster Pentium 3 on a faster bus. Poor Intel.
and 7 years later with the said design beating AMD across the board, run ay lower power and sell at cheaper price :)
7 years later with the said design beating AMD across the board, run ay lower power and sell at cheaper price :)
Dude, face the reality. Today, AMD firmly controls the high end. Intel is ultra low end. The role has reversed. Intel's only strength today is desktop, and that strength will evaporate soon.
Shakirou said ...
Dude, face the reality. Today, AMD firmly controls the high end. Intel is ultra low end. The role has reversed. Intel's only strength today is desktop, and that strength will evaporate soon.
Dude, face the reality. Right now, Intel firmly controls the low end and ibm with power controls the highend. amd is somewhere in between.
Sharikou, Ph. D said...
"Intel is ultra low end. The role has reversed. Intel's only strength today is desktop, and that strength will evaporate soon."
Low end, meanin anything below 4P?
What percentage of processor sales are desktop processors?
Only strength...
What do you call AMD in the mobile segment if Intels only strength is in desktop?
How will it evaporate?
Aside from K8L, what is going to out perform Conroe price/performance?
WTF?
Some pretender said
"Dude, face the reality. Today, AMD firmly controls the high end. Intel is ultra low end. The role has reversed. Intel's only strength today is desktop, and that strength will evaporate soon"
Today wins every dame benchmark for desktop/server ( dual-core ).
Unless you go look for some narrow high memory specific bandwidth benchmark do you see AMD lead.
Gone are the days of running around talking about CPU one running circles around CPU two. At this point if you look at the overall suite of benchmarks COre2 is the winner. Nothing to argue about.
65nm will incrementally help AMD a bit with higher clockes. Adding more cache will be of little improvement for AMD with their superior buss architecture. So in the end until K8L comes out in 2nd half 2007 INTEL is king. Hard to win the Market share wars with an inferior product and being capacity constrained. AMD ain't going away but it is not the winner in the larger scheme of things.
enumae said:
"That is not a tough one to figure out, expected prices were laid out a few monthe before AMD's price cuts.
I wasn't simply refering to price, but to real performance, not the fudged up numbers we saw prior to release. AMD knew exactly how CONroe would perform. A $200 cpu won't out-perform an $800 one as Intel would have you believe. SuperPi not withstanding.
"I am sorry to burst your bubble but Kentsfield is also drop in quad core."
Seems I'm the one netbursting your bubble, as Kentsfraud even OC'd doesn't give much more performance than Woodcrest. But why would it, being on the same FSB.
"I am not promoting Intel, just running numbers to debate the Dr. statements."
Any open-minded person likes a spirited debate, but when those numbers are based on Anand's, you might as well just be promoting Intel.
Dude, face the reality. Today, AMD firmly controls the high end. Intel is ultra low end. The role has reversed. Intel's only strength today is desktop, and that strength will evaporate soon.
Correction: Intel beats AMD at desktops (including gaming), laptops, AND 1P and 2P servers (The only thing poor opty is good at is FP, which no one other than educations institutions uses). With Tulsa, AMD's leadership in 4P is questionable at best. So the only place where AMD has confirmed leadership is 8P. What's the volume there? 0.1%??
And K8L is vaporware at best at this point (I know it has taped out, but so has 45 nm Core 2). There is no reason to believe that it will perform better than Core 2 uArch. Especially considering Core 2 will be on 45 nm by then. Again 8P and 4P (maybe, considering Tigerton is Core 2 based) is where AMD will have leadership, but that is miniscule of the market.
The bottom line is, Intel has best uArch, AMD has best interconnect. Sadly, for AMD, interconnect technology does not matter as far as 99% of the market is concerned.
And don't give me that crap about upgradability. Average Joe doesn't care about upgradability. He buys a computer from Dell or HP, throws it away after 3 years, and buys a new one.
Seems I'm the one netbursting your bubble, as Kentsfraud even OC'd doesn't give much more performance than Woodcrest. But why would it, being on the same FSB.
Correction: Not much performance improvement compared to 2 Woodcrests. Well, if the chip can give as much performance as two woodcrests, it will kill any 4x4 AMD has to offer. Also, those guys were running at FSB 1200, while even Inquirer agrees that Intel will be releasing Kentsfield at 1333 FSB. So I would wait for somewhat more official numbers on Kentsfield numbers to come out before passing final judgement.
Correction: Intel beats AMD at desktops (including gaming), laptops, AND 1P and 2P servers (The only thing poor opty is good at is FP, which no one other than educations institutions uses). With Tulsa, AMD's leadership in 4P is questionable at best. So the only place where AMD has confirmed leadership is 8P.
Most of you said it untrue. Woodcrest is a solution that doesn't solve any problem
1) For clustered HPC env, Opteron rules with 38% FP performance lead.
2) For transctional SMP env, Woodcrest is unsuitable, because it's 2P only.
3) For Web, Opteron rules on apache performance.
The only thing Woodcrest is probably good is workstation, but that market is tiny -- and ruled by Opteron any way.
On desktop, Conroe unit is so tiny, it is negligible. Conroe was a baby delivered prematurely.
Randy said...
"A $200 cpu won't out-perform an $800 one as Intel would have you believe."
I never said it would, but clock for clock and price/performance the new Core is beating K8, all the way up the line.
"Seems I'm the one netbursting your bubble, as Kentsfraud even OC'd doesn't give much more performance than Woodcrest."
Randy, your original statement was about upgrading, lets stay in that context, not performance.
"Any open-minded person likes a spirited debate, but when those numbers are based on Anand's, you might as well just be promoting Intel."
Looking at your comments, are you open minded?
THIS IS TO EVERYONE...I could have just left out the source... but why try and hide the numbers, hell if it didn't have a source someone would have said I made them up, so sorry everyone if you don't like them.
Back to you randy...
Point me to a site that has benchmarks you like, and I will run them again, like I said before I don't care.
The whole point was to show proposed numbers, thats all, they are not real.
yes premature enough to wipe the floor with AMD's best by the biggest margin we have seen a new product do.
Another brilliant deduction Doctor. At this rate they will probably admit you in a mental patient ward.
"That's what AMD is doing. Once it nailed the 65nm, it can make all of its chips under that node."
So why so long for AMD to transition the FX series to 65nm? Hmmm... According to you they should be able to do this with their fantastic APM3.0 instantaneously... perhaps there actually is some design work when transitioning from one node to another even if it is a simple shrink. Of course you have no clue about this as you don't have any clue about fab technology except what you can read on the web.
Oh wait never mind I'm sure AMD prefers to run them on a process with a bigger die size (as it costs more to produce) and to not take advantage of the 25% speed improvement YOU claim by simple node transistion.
Yeah, I'm sure AMD prefers to keep their top of the line desktop chip on a lower performing, higher cost tehcnology node, given that according to you they can just drop it in on 65nm with this fabulous APM3.0 invention!
...or maybe as it accounts for <0.25% it is not worth the design effort? (i.e. technology is not "orthoganol" to product design!)
" Anonymous said...
http://www.fabtech.org/content/view/1390/74/
"The reports authors said that yields were at 50 percent at Chartered's Fab7 but sometimes even better than those achieved within AMD's own fab!"
Over 50% sounds promising -_-
http://www.tgdaily.com/2006/08/18/amd_4x4_on_course_for_second_half/
""good bet" that items in that product line - whose final name is to be announced - will be available at Christmastime."
Yay 4x4 for Christmas:) "
considering they for sure mention 65nm, id say its a good news..
btw, that news is around.. 3 months old
so let's presume they've improved it now . besies 50% yield in 65nm is quite remarcable, considering AMD just started to switch while pumping a lot of X2's, 64's..etc..
So why so long for AMD to transition the FX series to 65nm? Hmmm... According to you they should be able to do this with their fantastic APM3.0 instantaneously... perhaps there actually is some design work when transitioning from one node to another even if it is a simple shrink. Of course you have no clue about this as you don't have any clue about fab technology except what you can read on the web.
Sharikou is an idiot, and everyone knows it. The guy is like a two-year old--keeps on repeating Opty beats Woody Opty beats Woody Opty beats Woody .... forever! Doesn't matter what benchmarks you show him, or what real data you show him. He just goes "Opty beats Woody Opty beats Woody Opty beats Woody .... forever!"
The simple question is: if AMD's APM is sooooo great, why are they still stuck with 90 nm? Didn't Intel move to 65 nm a year ago? Heck, 6 months after AMD moves to 65 nm, Intel moves to 45 nm. Some APM 3.0!!!
Here's the deal: APM or no APM, you need to validate whether a tool is good enough for performing the job. You know there's a lot of crappy process hardware out there, and you cannot just start processing your stuff using anything lying around. Now, to perform that kind of correlation on 1 tester, it takes about 4 hours (2 wafers detailed analysis). Considering your entire fab is idle, and you can cycle the correlation wafers through your fab one after another, it would take 800 hours to qualify a floor with 200 testers. That is more than a month. Now you start to get a feeling for why it takes so much time to ramp up a product.
APM is just some BS AMD pulled out of its @$$! No one in the industry could care less!
BTW, Sharikou is such an idiot that he think Copy Exact means running the same product throughout all the fabs! What a moron.
Sharikou: here is a small tutorial for you.
Copy Exact has nothing to do with what or how many products you can process simultaneously. Heck, you don't even have to have same product on the same wafer (that's right, you can have 4 fully different chips on the same wafer alternating each other). What Copy Exact means is, as long as you have a particular process/product, you are going to process it exactly the same way, no matter where in the world it is being processed. And by process means everything: including recipes, masks, test programs, database settings, heck, even file names and file paths. That reduces confusion and chances of errors. This ensures that products coming out of Ireland have the same quality as those coming out of oregon and those coming out of New Mexico. Thus, the customers don't have to worry about which fab is produing their products, since they are all "Copy Exact".
to enema or enumae whatever:
Heres a benchmark I like. Intel shoves a 20 stage(later 31 stage) dildo named Nutbust in the consumer's ass and it takes the faithful 1.2 sec. to reassume the position for the CONroe shafting. I'm laughing all the way to my nearest AMD dealer. Why are so many willing to throw money at a Co. that has ripped you off before without seeing actual proof. Its not like they've changed, after all but admitting Netbust is junk, they're still pushing it on consumers. Intel could write that junk off and ramp they're new wonder chip. Makes you wonder what corners were cut with CONroe. Remember how long it took to find out about Replay in P4. But as long as its all about a few benchies, the P4 won benchies in its day too, but you couldn't give me one for the cost of having to buy a mobo.
"Any open-minded person likes a spirited debate, but when those numbers are based on Anand's, you might as well just be promoting Intel."
And should I assume that you believe Sharikou's claims in the next blog that have been quoted off AMDZone with no mention of HW setup?
Everyone seems to be using benchmarks selectively when they suit their purpose...although at least Anandtech tells the reader what the actual HW they are testing is!
Perhaps Shari-kook might want to think about doing his own benchmark - oh wait he said he was going to do this with Woodcrest and then chickened out!
"Correction: Intel beats AMD at desktops (including gaming), laptops, AND 1P and 2P servers (The only thing poor opty is good at is FP, which no one other than educations institutions uses)"
Corrections: (too many - allow me to break them into numbers)
1) Intel's Core 2 Duo beats AMD's K8 at the same clock rate for desktop gaming, 1P server, and 2P with less than 4GB RAM.
2) Intel's Core Duo notebook performs on par with AMD's Turion X2 for the same price, with slightly lower power consumption, but much worse wireless connectivity.
3) The performance per dollar (street price) of Intel solutions are beaten soundly by all AMD ones except the FX.
Your statement sounds like anything from Intel is beating that from AMD; it's completely false. The truth is you can't do meaningful comparison without setting the boundary conditions.
For general performance conception, the fact is, I use both Core 2 Duo and A64 X2; for all that matter, including java/C++ programming, photoshop editing, and transcoding, none is considerably faster than another. That is within 15%. (I also noticed how much faster Core 2 Duo is at SuperPi and Ogg/DivX, though).
Randy said...
"Heres a benchmark I like. Intel shoves a 20 stage(later 31 stage) dildo named Nutbust in the consumer's ass and it takes the faithful 1.2 sec. to reassume the position for the CONroe shafting."
It would seem you have lost the argument, and now we are going to resort to name calling... congratulations on a spirited argument... lol
"I'm laughing all the way to my nearest AMD dealer. Why are so many willing to throw money at a Co. that has ripped you off before without seeing actual proof."
What kind of proof do you want to see?
"Its not like they've changed, after all but admitting Netbust is junk, they're still pushing it on consumers."
Mainstream consumers will never see the difference, maybe you are missing that.
"Intel could write that junk off and ramp they're new wonder chip."
Sure, makes for smart business practice.
"Makes you wonder what corners were cut with CONroe."
Why don't you explain these cut corners for me?
"Remember how long it took to find out about Replay in P4."
Sorry but I am not familiar with "Replay".
"But as long as its all about a few benchies, the P4 won benchies in its day too, but you couldn't give me one for the cost of having to buy a mobo."
A few benches, well it would seem that you have missed alot of the reviews out there.
Here is an actual question. When K8L comes out, will you believe the benchmarks?
to enumae:
If your not familiar with Replay, what it is and more importantly why its there, then all I can say is Intel's marketing has worked. These are the kind of things you don't find in benchmarks.
As far K8L, I don't need benchies, I believe in the µ-arch and the K8 before it.
Randy said...
"If your not familiar with Replay, what it is and more importantly why its there, then all I can say is Intel's marketing has worked."
I went and looked it up and it seems to have a negative effect, but at the same...
"if each thread is processing different types of operations, the replay system will not interfere, and a performance increase can appear."
I am not in the processor field, so my understanding is limited, but I do understand that no processor is perfect, hell, P4 would usually beat K8 in video encoding.
"These are the kind of things you don't find in benchmarks."
OK, but how does that effect the consumer, the average consumer who will be buying Netburst?
"As far K8L, I don't need benchies, I believe in the µ-arch and the K8 before it."
If you do not need them, then you are AMD's perfect consumer.
Without comparison, how do you know what performance advantage they have or do not have over the competition?
You want the best performance for your dollar, right?
PS: I am still waiting for your explanation on how Intel cut corners on Conroe...
enumae - since randy isn't going to post anything useful, here you go. From wikipedia: "The Replay system is a little known subsystem within the Intel Pentium 4 processor. Its primary function is to catch operations that have been mistakenly sent for execution by the processor's scheduler. Operations caught by the replay system are then re-executed in a loop until the conditions necessary for their proper execution have been fulfilled."
Intel basically made its scheduler super-aggressive (perhaps a necessary evil given their retartedly long pipeline) and it just throws instructions down the pipe whether they are legal to be executed or not. Replay catches the failed instructions and re-executes them in sequence.
Replay is basically the reason why Hyper-Threading on Netburst is trash. Instead of idling while waiting for a dependency to be worked out, it tries to execute anyways. This wastes the second threads resources, and increases power consumption.
see more here
Thanks SaintGreg, much appreciated.
enumae said:
"If you do not need them, then you are AMD's perfect consumer."
No, I could just as easily be a great customer of Intel's, one thing stops me, I just don't believe in the µ-arch.
Replay doesn't just trash HT, although some of that was fixed with Prescott. No, its what it represents.
Its a watt-eating, cycle-wasting, latency-increasing fix for speculation so aggressive you can't come to any other conclusion than Intel designed Netburst for clockspeed above all else, even performance.
Corners probably weren't cut on Conroe, shouldn't be, it took them enough time to get it out. Its a matter of design philosophy, do you want a Co. that seems to have a market expert as their head engineer. Netbust was made for AMD, not for you, the people who bought it.
Randy said...
"Its a matter of design philosophy, do you want a Co. that seems to have a market expert as their head engineer."
I am not trying to sound sarcstic here, but do you have a name for the head engineer?
Thanks.
"P4 would usually beat K8 in video encoding."
My personal experience is that Intel has much better SSE2 implementation than AMD.
For most softwarecryptographic algorithms, K8 usually outperform P4 more than 60% at the same clockrate. However, if the algorithm is SSE optimized, K8's advantage becomes something like 25%.
That superiority in P4 seems also transferred to Conroe, BTW.
Tell us all please..
How is Core2 Conroe a corner cutter.. Please share your claim. If none then like the PhD pretender you are just a plain pretender lier... LOL at your stupid claims. Backup them up if you got a dick.
Intel only has 2 fabs in 65nm that work. AMD has 2 in 65nm fab 30 aready in transission of may 06 and fab 36 of now. Silicon Germanium (e-SiGe) with Dual Stress Liner (DSL) and Stress Memorization technology (SMT) on Silicon-On-Insulator (SOI) wafers, will be in K8L. The new process technologies reduce interconnect delay through the use of lower dielectric constant (lower-K) insulators, which can improve overall product performance and lower power consumption. In addition, the new technologies have shown ability to be manufactured at the 65nm generation and scaleable for use in future generations.
AMD's fab processes are more advanced then intels and take 2x the time to make them because of there complexity. Intel uses a standard process since the only one, they use only 1 process type. COM. Now enhanced on conroe cores but not much different then it was on netburst. Well AMD uses 5 process types at the least on its 65nm generation. 3 on its 90nm generation.
Just because intel is in a smaller process doesn't make it more advanced infact AMD just matches it with 90nm tech if not giving better heat and watt reduction. AMD made perfect 90nm processes is why. Intels 65nm isn't perfected like AMD's is, further more AMD uses processes that are too advanced for intel to put into there cpus. Intel can also make more because COM is cheaper and takes half the time to make But not as reliable as AMD's process witch makes a stronger chip with greater heat resistance and almost 0 transistor leakage unlike intels high transistor leakage.
Next time before you say something about fab tech do more research into it. You really don't know much about intel processes let alone AMD's.
Besides AMD is also useing partner fabs to do more 65nm processes. They will be out even sooner.
AMD's 90nm process alone is enough and the EE vers prove that agenst 65nm Intel's.
"May I just ask, how long does it take to search for some random Chinese site to 'prove' the X2 4200+ is 'superior' to the E6300?"
As a matter of fact, that is not a random Chinese site, but one of the largest tech site. The article that Sharikou linked is among the highest-viewed article of the week.
Sharikou is good at digging out news, I have to say. But this link isn't that hard to find if you know Chinese (or if you can employ someone who does to help you).
BTW, the test in this article is actually performed in Japan by some AMD-favoring people (maybe AMD reseller). Look at the pictures more carefully and you would see that. I remember some ignorant fellow here on the comments demanded me to show him a single AMD-favoring site or page on the Internet. Well here it is. Plus the AMDZone article that Sharikou linked elsewhere. Apparently some people are too ignorant to understand how free (of big company benchmarketing) the Internet can be.
Post a Comment
<< Home