Tuesday, August 01, 2006

IBM and Novell folks talk about AMD64

See AMD enterprise event #6 (video). Three IBM execs talk about the new Opteron servers. Novell's CEO and CTO are there pumping SuSe Linux 10. IBM+Novell+AMD join forces. I bet they have seen K8L in action.

This report claims that socket F Opterons have been shipping for weeks. Socket F Opteron has built-in PCI-E support.

The specs of IBM x3655 is here, and the specs of IBM x3455 is here. From the x system spec, you can see Rev F opteron's clockspeed is 2.8GHZ. The specs for IBM blades is here. Since a 3GHZ Woodcrest is only 2% faster than 2.6GHZ Opteron in Apache benchmark, a 2.8GHZ Rev F should beat 3GHZ Woodcrest in that benchmark. In FP performance, expect Rev F Opteron to smash Woodcrest.

50 Comments:

Blogger TheKhalif said...

People just don't want to accept that K8L will give a similar increase as 4x4. 70-80% over Opteron.

2 FP
2 SSE
32B prefetch
2 loads per cycle
2MB L3
3 AGU
3 ALU
perhaps cluster-based Multithreading.

It will kill. PERIOD.

1:15 PM, August 01, 2006  
Anonymous Anonymous said...

"It will kill. PERIOD."

unfortunately old (ie K7 times) arch limitations are also still there. hardcoded ALUs to input ports. still only 3 input ports (conro is 4 + macro op fusion).

3 AGU and 3 ALU is in fact 3 AGU/ALU.

u also forget better branch prediction in k8l, which is supposed to be secret conroe weapon.

2:39 PM, August 01, 2006  
Anonymous Edward said...

"unfortunately old (ie K7 times) arch limitations are also still there. hardcoded ALUs to input ports. still only 3 input ports (conro is 4 + macro op fusion)."

Don't be fooled by Intel propaganda. Micro op fusion-similar stuff exist in K8 and Core Duo already (maybe even Pentium-M). Core 2 Duo just added a few (maybe one or two) such cases - probably due to its better memory disambiguation capability.

3:04 PM, August 01, 2006  
Anonymous Anonymous said...

Any idea when we are going to see 64 bit programs like Nero and other encoding type programs, or am I just unaware that they are already out? Are many of the large software vendors waiting for Vista to put these out?

3:16 PM, August 01, 2006  
Anonymous Anonymous said...

AMD64 for Digital Content Creation :

- Cinema 4D R9.6 64-bit

- Softimage|XSI 5.1 Essentials/Adv. (Native 64-bit)

- NewTek LightWave 3D v9.0 64-bit

- SideEffects Houdini v8.1 64-bit

- NextLimit Realflow v4.1 64-bit



- Autodesk 3ds Max 9 64-bit -> October 2006


- Autodesk Maya 8.0 64-bit -> August 2006


New ZBrush 2.5 is optimized for 32 bit systems as well as 64 bit systems, taking full advantage of the processing power, multiple CPUs and available RAM.
ZBrush



- [Game] Crysis will also take advantage of 64-bit CPUs :
According to Yerli [Cevat Yerli, Crytek president and CEO], the 64-bit version "will bring a performance difference of up to 10 to 15 percent on each thread" compared to the 32-bit version.
Crysis Exclusive Preview

5:23 PM, August 01, 2006  
Anonymous enumae said...

TheKhalif said...

"People just don't want to accept that K8L will give a similar increase as 4x4. 70-80% over Opteron."

When K8L arrives it will be directly aimed at enthusiast (or am I wrong), and thereby carrying a premium price.

It would also (most likely) be one of the next in the FX series.

So here is my questions...

Why bother with 4x4 if a native quadcore has the same performance?

This seems a little strange that AMD would do this, but only AMD knows for sure what kind of performance K8L is getting, maybe 70-80% is a little high.

Or is 4x4 a stop gap against C2D?

Before you slam me for that, think about it, whats the point of 4x4 if K8L is going to make it obsolete.

Now 8x4 makes sense, but it is a different story all together, and only in a wetdream could I afford it, but damn thats a good dream.

9:51 PM, August 01, 2006  
Anonymous Anonymous said...

This Paul O. quote is priceless. When asked why he had a 4.3 Billion stock on hand.

"If I have made one mistake in the last two years, it was that we did not build enough chips at the right time…that got us in trouble in terms of market share," Paul Otellini, chief executive, told analysts

Do ya think he has enough yet !

10:59 PM, August 01, 2006  
Anonymous Anonymous said...

to Enumae

The 4X4 and K8L are not exclusive. Think of it, 4X4 gives performance advantage of Opteron over C2C. When Intel has its 4-core CPU, 4X4 gives K8L the same kind of performance advantage. In AMD roadmap, 4X4 is going to stay no early that 2008.

BTW, I guess 4X4 is aims at PC station.

1:38 AM, August 02, 2006  
Anonymous Anonymous said...

About 4x4

2006: 4x4 (2 x dual-core) (vs Kentsfield)

"Buyers should be able to get a bundle with two FX-branded processor to use with the 4x4 platform for less than $1000."

2007: 4x4+ (2 x quad-core, DDR2)

2008: 4x4++ (2 x quad-core, DDR3)

Technologies Roadmap: Desktop (PDF, p.32)

3:50 AM, August 02, 2006  
Anonymous Anonymous said...

"Don't be fooled by Intel propaganda. Micro op fusion-similar stuff exist in K8 and Core Duo already (maybe even Pentium-M). Core 2 Duo just added a few (maybe one or two) such cases - probably due to its better memory disambiguation capability."

Edward, u mix up mIcro-op fusion and mAcro-op fusion. micro-op fusion exists for years, macro-op fusion is feature new to conro. with it conro can process 5 commands in one pass. unlike K8 which is still hard limited to 3.

it has nothing to do with propaganda, k8 has arch drawbacks as well.

4:05 AM, August 02, 2006  
Blogger Mad Mod Mike said...

"Or is 4x4 a stop gap against C2D?"

4x4 is a method to increase performance on multi-tasking, not a stop gap solution to stop C2D. 4x4 was not thought up overnight, I guarantee it has been on the minds of AMD for quite some time after seeing the success of 2P Opteron64 servers and workstations.

The point of 4x4 is that it gives the user the ability to buy 1 CPU now, and another later on when he/she feels it is necessary; something you cannot do with C2D. 4x4 is also upgradable, likely to support DDR3 since it supports K8L as well; meaning you buy a Dual-Core CPU now, and a few years later, Dual Quad-Core's are in your sockets.

5:20 AM, August 02, 2006  
Blogger pointer said...

People just don't want to accept that K8L will give a similar increase as 4x4. 70-80% over Opteron.

Oh, you mean that it will beat C2D for the same freq CPU comparison (since C2D beat Opteron by tens % but less than 70% and lower power cosumption).

ok, let's see. when is the K8L supposed to be release? let do a bet here. I bet it will still lose (that's the reason AMD has to come out with 4x4 propaganda). If you accept the bet, just reply to this comment. if I lose, I'll come here and say that I lose and Intel lose. if you lose, you come here and say that you lose and AMD lose.

6:01 AM, August 02, 2006  
Anonymous Wirmish said...

"Why bother with 4x4 if a native quadcore has the same performance?"

"Before you slam me for that, think about it, whats the point of 4x4 if K8L is going to make it obsolete."

Because you can put 2 K8L in 4x4. :Þ

6:36 AM, August 02, 2006  
Blogger Archit said...

Nice Info!

For some Funny 'n' Weird Jokes Visit

topjokes.blogspot.com

6:40 AM, August 02, 2006  
Anonymous Edward said...

"This Paul O. quote is priceless. When asked why he had a 4.3 Billion stock on hand."

His answer just showed that he's a bonehead who refuse to admit (at least publicly) the reality.

This kind of people can shamelessly tell you the color is blue when it's really green.

8:00 AM, August 02, 2006  
Anonymous Anonymous said...

Should smash. SHOULD frag. Should this, should that.

This is non-news. We'll be excited when it actually happens. Until then it's kind of silly to speculate, don't you think?

8:40 AM, August 02, 2006  
Anonymous enumae said...

Thanks to everyone who answered my question.

9:51 AM, August 02, 2006  
Anonymous the ka leaf said...

"The point of 4x4 is that it gives the user the ability to buy 1 CPU now, and another later on when he/she feels it is necessary; something you cannot do with C2D. 4x4 is also upgradable, likely to support DDR3 since it supports K8L as well; meaning you buy a Dual-Core CPU now, and a few years later, Dual Quad-Core's are in your sockets."

AMD has specifically said that in the initial release of "4x4", the user will have to purchase TWO chips together -- STARTING at $1000! AMD management does not want to see 1 chip and then later another 1 chip and deal with compatibility issues, support issues, etc.

To anyone except an Another Microbrained Dumbshit (AMD), 4x4 is scam for AMD to sell more chips. "Megatasking" is something very few people do. It is such a small market segment, one wonders why AMD didn't spend the money making 2P workstations better. Just saving the cost of a heatsink by moving two sockets close together is not much of a win.

AMD has made *nothing* that is upgradable across memory types. So expecting semi-crippled Opteron chips to suddenly support DDR3 is just really really dumb. Stop the lies.

The user will be forced to throw away their old system (like they did moving to AM2, like they did when going from AM2 to AM4x4). AMD obsoletes platforms way faster than Intel and marginalizes customer investment.

What really needs to happen is AMD has to make a good chip. No amount of smokescreen is going to fool the market. AMD makes dud chips today.

It is sad to see that AMD fired all the Opteron team so all they have is ATI and globbed-together sockets. That is pathetic.

10:24 AM, August 02, 2006  
Anonymous Anonymous said...

Report: Intel Chip Glut Swells Inventory

Semiconductor stockpiles rose to $2 billion the second quarter, up 78% from the first quarter, research firm iSuppli says.

http://www.informationweek.com/security/showArticle.jhtml?articleID=191601092

11:18 AM, August 02, 2006  
Anonymous Anonymous said...

I can see a future ATI coprocessor in a 4X4 setup.

K8L also get a hypertransport upgrade, right? You think IBM offers new lines of Opteron servers for short-term gains? Nope.
IBM probably likes what it sees down the road.

11:28 AM, August 02, 2006  
Blogger whorush said...

it has pci-e in it?

how do you know?

thats gotta give it a boost for a few things and further simplify mobo/chipset design.

can u send me a link showing it has that?

12:33 PM, August 02, 2006  
Anonymous Anonymous said...

"AMD obsoletes platforms way faster than Intel and marginalizes customer investment."

Thanks for the laugh.

12:57 PM, August 02, 2006  
Anonymous Anonymous said...

"Megatasking" is something very few people do.

Thats because with single core CPUs the average user wasn't able to get good performance running several applications at once. Now with increasingly parallel chips, users are able to have a ton of crap open and not get bogged down (if you have the RAM for it). That appeals to me, as I'm sure it does to others.

AMD has made *nothing* that is upgradable across memory types.

Apparently AM3 Opterons will be able to use either DDR2 or DDR3. Just because it hasn't heppened yet doesnt mean it wont. Considering the cost of upgrading both motherboard and RAM, I don't really see the problem in getting a new CPU as well, unless you are trying to stay at the bleeding edge (in which case you probably are getting a new CPU anyways)

1:07 PM, August 02, 2006  
Anonymous Anonymous said...

"His answer just showed that he's a bonehead who refuse to admit (at least publicly) the reality.

This kind of people can shamelessly tell you the color is blue when it's really green."


He was referring to the chipset shortage awhile back where Intel could not sell processors because chipset supply was so constrained. Were you interperting his comments to mean that he thought he should have been making more CPU's?

1:14 PM, August 02, 2006  
Anonymous Anonymous said...

"4x4 is also upgradable, likely to support DDR3 since it supports K8L as well; meaning you buy a Dual-Core CPU now, and a few years later, Dual Quad-Core's are in your sockets."

So by upgradeable from K8/DDR2 to K8L/DDR3, you mean you only have to change the chips and memory; basically you are saying you can reuse the motherboard?

Are DDR2 and DDR3 physically interchangeable on a board, I don't have much info on DDR3. I assume DDR3 works at a slightly lower voltage?

1:17 PM, August 02, 2006  
Anonymous Anonymous said...

"The user will be forced to throw away their old system (like they did moving to AM2, like they did when going from AM2 to AM4x4). AMD obsoletes platforms way faster than Intel and marginalizes customer investment."

I don't think it is fair to say this only applies to AMD, both Intel and AMD do this, on a positive note it is a good thing for motherboard makers as they get to sell new boards with most new chip architecture launches. The difference of late is Intel has been doing this via chipset compatibility, however that generally means a new board.

1:22 PM, August 02, 2006  
Anonymous Anonymous said...

"AMD has specifically said that in the initial release of "4x4", the user will have to purchase TWO chips together -- STARTING at $1000! AMD management does not want to see 1 chip and then later another 1 chip and deal with compatibility issues, support issues, etc."

--> And in recent statements, they have said that you can buy 2 FX chips for <$1k. By your logic, you probably said F*%^ Nvidia/ATI for making me buy 2 GPU's!!

"To anyone except an Another Microbrained Dumbshit (AMD), 4x4 is scam for AMD to sell more chips. "Megatasking" is something very few people do. It is such a small market segment, one wonders why AMD didn't spend the money making 2P workstations better. Just saving the cost of a heatsink by moving two sockets close together is not much of a win."

==> You think they've been sitting on their a$$es all this time? K8L... You are really naive and shows how much of a moron you are. Do you have any idea of design cycles, technology qual, to ramp, and how much time it really takes?

"AMD has made *nothing* that is upgradable across memory types. So expecting semi-crippled Opteron chips to suddenly support DDR3 is just really really dumb. Stop the lies."

==> Your Intel Fanboism has resorted you to the IQ of a broomstick. Has Intel offered ANYTHING upgradeable across memory types? Much less CPU? Intel is just abusing its monopoly on the CPU market to force the market in a certain direction because they CAN. AM2 supports both DDR2/3.

"The user will be forced to throw away their old system (like they did moving to AM2, like they did when going from AM2 to AM4x4). AMD obsoletes platforms way faster than Intel and marginalizes customer investment."

==> You are continuing in proving yourself as an even bigger moron and blinded by all else. 4x4 is a path to migratability, for YOU, the consumer, to CHOOSE to upgrade at your own pace and for what you need.

"What really needs to happen is AMD has to make a good chip. No amount of smokescreen is going to fool the market. AMD makes dud chips today."

==> They have had the best chip for the past 3-4 years. And Core2 will only hold a slight lead for so long, whenever it is actually available. Heard they are having issues with the C1 stepping...

"It is sad to see that AMD fired all the Opteron team so all they have is ATI and globbed-together sockets. That is pathetic."

==> Fired? Ok, seriously now... whatever you're smoking, it's really really good. Mind sharing it with the rest of us? :)

***The previous posted by this moron exemplifies what an extreme Intel Fanboy is***
--> There are intelligent ones out there able to hold a conversation.

1:54 PM, August 02, 2006  
Blogger TheKhalif said...

When K8L arrives it will be directly aimed at enthusiast (or am I wrong), and thereby carrying a premium price.



K8L is for servers. I would say that it will be for 16Way and up. there ill probably be a versiogn for dual core along with BullDozer, but they may also go with non-native quad by slapping two X2s together -Intel got away with it.

1:59 PM, August 02, 2006  
Blogger TheKhalif said...

Oh, you mean that it will beat C2D for the same freq CPU comparison (since C2D beat Opteron by tens % but less than 70% and lower power cosumption).

ok, let's see. when is the K8L supposed to be release? let do a bet here. I bet it will still lose (that's the reason AMD has to come out with 4x4 propaganda). If you accept the bet, just reply to this comment. if I lose, I'll come here and say that I lose and Intel lose. if you lose, you come here and say that you lose and AMD lose.



Awww poor baby. Intel won't hold the lead long. Wait no one can even get Core 2. Woodcrest is not the same as Core 2 because Opteron is designed for server use and IF 51xx wins anything it will be wksta apps like encoding.

Opteron still owns DBs. There are some new scores on TPC but the chips se so much power that they aren't worth it.

2:03 PM, August 02, 2006  
Anonymous Anonymous said...

Intel only 1 Billion dollar cash left. Look at this link

http://knobias.10kwizard.com/filing.php?repo=tenk&ipage=4303150&doc=1&total=&back=2&g=&attach=on\

2:06 PM, August 02, 2006  
Blogger TheKhalif said...

First, you wouldn't be makign fun of my name would you? That is bad for your health and well-being.

AMD has specifically said that in the initial release of "4x4", the user will have to purchase TWO chips together -- STARTING at $1000! AMD management does not want to see 1 chip and then later another 1 chip and deal with compatibility issues, support issues, etc.

To anyone except an Another Microbrained Dumbshit (AMD), 4x4 is scam for AMD to sell more chips. "Megatasking" is something very few people do. It is such a small market segment, one wonders why AMD didn't spend the money making 2P workstations better. Just saving the cost of a heatsink by moving two sockets close together is not much of a win.



It desn't matter if you can get one chip or two. People who need that kind of power &70% fatser than FX62) will pay for it. My next upgrade will be 4x4. I need it for my dev box.

I could run all my VMs with Exchange and SQL.

face it. There are two major chips manufs around and unless Intel goes out of business there will be.

2:08 PM, August 02, 2006  
Anonymous Edward said...

"AMD has specifically said that in the initial release of "4x4", the user will have to purchase TWO chips together -- STARTING at $1000!"

Based on this link (search for x2) 4x4 will initially also work with X2, not just FX-only.

Frankly, it would be moot if 4x4 was more expensive than dual-socket Opteron. So the claim of $2k just for two CPUs got to be false.

I know Intel fanboys have hard time to believe it, because their great Intel would never give users anything that expansive without it being super expensive. ;-)

3:52 PM, August 02, 2006  
Anonymous Ho Ho said...

I know Intel fanboys have hard time to believe it, because their great Intel would never give users anything that expansive without it being super expensive. ;-)


If Woodchrest would be availiable it would cost around $1200 for good motherboard and two CPUs. Of cource with top-of-the-line CPUs that would increase to around $2400.

Btw, you can build a "4x4" around socket 940 for around $900 for 2x 265+mobo even today. Of cource if you would want to get 2GHz+ CPU's it would cost you much more. Of cource that price will most likely drop considerably with the new Opterons just around the corner.

4:18 PM, August 02, 2006  
Anonymous Anonymous said...

When K8L arrives it will be directly aimed at enthusiast (or am I wrong), and thereby carrying a premium price.


Did you not hear Dirk, Hector, and the other AMD executives don't give a hoot about benchmarks and you enthusiast. You guys don't make the maket, nor effect it.

4:29 PM, August 02, 2006  
Anonymous Anonymous said...

When K8L arrives it will be directly aimed at enthusiast (or am I wrong), and thereby carrying a premium price.


Did you not hear Dirk, Hector, and the other AMD executives don't give a hoot about benchmarks and you enthusiast. You guys don't make the maket, nor effect it.

4:30 PM, August 02, 2006  
Anonymous enumae said...

Edward said...

Here is a very interesting part of the article you link to...

"The 4x4 initiative will be directly targeted at enthusiasts, according to Moorhead. Each socket will include an AMD Athlon64 X2 chip in a new AM2 socket, although what limitations will be placed on the clock speed will not be released until a future date, he said."

The only thing to do is wait and see.

There are articles saying FX only and some saying X2, but the part that should really jump out is that its a new socket, so anyone who bought an X2, thinking they could use it, is not going to be able to use it in 4x4.

To the anonymous poster who said...

"Did you not hear Dirk, Hector, and the other AMD executives don't give a hoot about benchmarks and you enthusiast."

Then why is "Patrick Moorhead, AMD's vice president of marketing." saying that this (4x4) is targeted at enthusiaist?

It would seem that they can make alot of money on the enthusiast crowd, not near as much as mainstream or servers but it has to be a good return.

TheKhalif said...

"K8L is for servers."

I am not trying to prove you wrong but here is a slide saying otherwise.

K8L-preview

6:07 PM, August 02, 2006  
Anonymous Anonymous said...

Intel only 1 Billion dollar cash left. Look at this link

http://knobias.10kwizard.com/filing.php?repo=tenk&ipage=4303150&doc=1&total=&back=2&g=&attach=on\


From the link:

Cash and cash equivalents
$ 3,436


that is not 1 billion, that is 3.4 billion.

7:56 PM, August 02, 2006  
Anonymous Edward said...

"There are articles saying FX only and some saying X2, but the part that should really jump out is that its a new socket, so anyone who bought an X2, thinking they could use it, is not going to be able to use it in 4x4."

yeah, we all know that 4x4 is targeted at the enthusiasts, so what? does that translate to it being FX-only? does that limit its price to over $1k per CPU? my comments was to rebuff these silly claims, and what's your problem with that?

also, "a new AM2 socket" IS an AM2 socket; otherwise it'll just be called "a new socket". Besides, IF 4x4 would require a different type of CPU, it WILL use a new socket, otherwise AMD will likely get hundreds if not thousands dissenting customers and RMAs every week.

you're just obviously in denial here. ok, ok, just cover your eyes and pretend that nobody can afford 4x4, alright? you're not forced to believe in anyway.

10:42 PM, August 02, 2006  
Anonymous Edward said...

"Edward, u mix up mIcro-op fusion and mAcro-op fusion. micro-op fusion exists for years, macro-op fusion is feature new to conro."

As a matter of fact, NO. I was talking about Core's micro-op fusion all the way. That's why I said it's probably improved due to Conroe's better memory disambiguation.

Macro-op fusion, OTOH, is conceptually simple: it groups a CMP or TEST with a conditional branch into one macro-op. AnandTech claimed that could improve performance by 11%, which is totally BS from/for amateurs. For this to work effectively, the branch prediction needs to work with a good value prediction of the CMP or TEST preceding the branch; this dramatically reduced the predict-right probability of the conditional branch, and could even hurt performance due to more pipeline flush. I'd be rather surprised if they could get 3% improvement out of a lot of efforts & complexity (which would contribute to eventual lower yield of the chip).

It's PURELY Intel propaganda as it is described by all those websites. Well, please show me hard data to prove its effectiveness if you could, and I'll appreciate it, and be humbly corrected. But until then, enough with those blue crystals that only make you feel good about a product.

11:07 PM, August 02, 2006  
Anonymous Ho Ho said...

AnandTech claimed that could improve performance by 11%, which is totally BS from/for amateurs. For this to work effectively, the branch prediction needs to work with a good value prediction of the CMP or TEST preceding the branch; this dramatically reduced the predict-right probability of the conditional branch, and could even hurt performance due to more pipeline flush

Wasn't it so that conditional jump can't create a branch misprediction? At least that is what I've heard from fellow programmers who came up with branchless KD-Tree traversal code. They simply replaced a bunch of "if"'s with conditional branches and got instant speedboost on that ugly 31-stage pipline CPU.

12:46 AM, August 03, 2006  
Anonymous Anonymous said...

"As a matter of fact, NO. I was talking about Core's micro-op fusion all the way. That's why I said it's probably improved due to Conroe's better memory disambiguation."

please have a read initial posts again. I started this with macro-op fusion, u continued that micro-op fusion. as a matter of fact u missed the point.


"Macro-op fusion, OTOH, is conceptually simple: it groups a CMP or TEST with a conditional branch into one macro-op."

yes, its extremely simple. the fact is that almost each CMP or TEST is followed with conditional branch. so this macroop-fusion works almost always on this very common code combination.

What Im trying to say is that K8L is very good enhancement to current K8 core, but it still holds some very old arch limitations. one of this limitation is decoder/execution limit on 3 op per tick. second is shared L2 cache. etc.

1:51 AM, August 03, 2006  
Anonymous enumae said...

Edward said...

"also, "a new AM2 socket" IS an AM2 socket; otherwise it'll just be called "a new socket". Besides, IF 4x4 would require a different type of CPU, it WILL use a new socket, otherwise AMD will likely get hundreds if not thousands dissenting customers and RMAs every week."

I was not trying to say that it would be $1000 per CPU, if you had read the whole post you would have seen I was pointing something out to someone else in regards to enthusiast.

Do you try to take my post out of context or as a personal attack?

In regards to a new socket, well thats what he said a "NEW SOCKET", not me. Wasn't AM2 for DDR2, so maybe they need to have a new socket for the other HyperTransport Links?

I don't know, but some here have stated they are needed for 4x4.

You also said...

"you're just obviously in denial here. ok, ok, just cover your eyes and pretend that nobody can afford 4x4, alright? "

I am not saying or pretending that no one can afford it.

I am not bashing AMD.

From what is out there the price seems to be at least $1000 for 2 processors and about $250 for the MB. $1250 +/- to an enthusiaist is not out of the question.

To respond to your sarcasm...

ok, ok, just cover your eyes and pretend that AMD doesn't want to make money.

8:08 AM, August 03, 2006  
Anonymous Edward said...

enumae said: "From what is out there the price seems to be at least $1000 for 2 processors and about $250 for the MB. $1250 +/- to an enthusiaist is not out of the question."

Since you didn't seem to get my points well, let me to re-state it: if 4x4 takes X2, it would be $300 for 2 processors, not $1000. If the CPU for 4x4 was more expensive than 2xxx Opterons, it will be moot. If the motherboard for 4x4 is more expensive than dual-socket Opteron boards, it will also be moot. Got it?

While there's nothing wrong to have $1250+/- for the enthusiasts market, there's also nothing wrong to have $500 for that (for JUST the processors and motherboard).


"Do you try to take my post out of context or as a personal attack?"

What are you talking about? I was originally commenting ka leaf's claim that 4x4 processor starting at $1000. I said, "my comments [were] to rebuff these silly claims, and what's your problem with that?" How could I have taken YOUR post out of context, when I was rebuffing OTHER's claims?

And where's the 'personal attack' you got? Point that out, please, or I'll regard this as some kind of dirty blackmail from you.

"In regards to a new socket, well thats what he said a "NEW SOCKET", not me. Wasn't AM2 for DDR2, so maybe they need to have a new socket for the other HyperTransport Links?"

NO! In the page I linked, Moorhead said "a new AM2 socket"! Why did you conveniently omit the AM2?

Well, from this link it seems 4x4 processors will be new and will bear the "FX" brand. But it also says the price tags will be well under $1000. So the "FX" there is really different from the "FX" we currently see. Don't take other's words out of their context!

Also, the idea that 4x4 will require "new socket" probably comes from here, where it says [AMD] unveils its new "Socket 4x4", which was from the "Socket 4x4 chipset" context. Again, don't take other's words out of their context!

10:51 AM, August 03, 2006  
Anonymous enumae said...

You know what Edward, we have been down this road before and always without any sort of achievement, or definitive conclusions in regards to the debates.

Edward said...

"Point that out, please, or I'll regard this as some kind of dirty blackmail from you."

Well if I recieve any comments from you I am automatically put on the defensive, just look at the comment above.

Or...

"you're just obviously in denial here. ok, ok, just cover your eyes and pretend that nobody can afford 4x4, alright? you're not forced to believe in anyway."

How would you take those comments?

After our last discussion, I am trying to be very civil with you, and will continue to try and do.

"Since you didn't seem to get my points well, let me to re-state it: if 4x4 takes X2, it would be $300 for 2 processors, not $1000."

Up until today the latest news about 4x4 was it would be around $1000 for the two processors.

Here is your original comment, while your saying it is directed to thekhalif, you make no mention of that...

"yeah, we all know that 4x4 is targeted at the enthusiasts, so what? does that translate to it being FX-only? does that limit its price to over $1k per CPU? my comments was to rebuff these silly claims, and what's your problem with that?"

Therefore I though it was directed at me.

"NO! In the page I linked, Moorhead said "a new AM2 socket"! Why did you conveniently omit the AM2?"

I was not leaving anything out, I had said that in my previous post an pretty much quoted the whole statement, the later comment was to you "NEW SOCKET".

A new AM2 socket, well I interpret that as refering to a new socket based on AM2, if it was referring to AM2, he could have said a new socket called AM2.

"Well, from this link it seems 4x4 processors will be new and will bear the "FX" brand. But it also says the price tags will be well under $1000. So the "FX" there is really different from the "FX" we currently see."

I am done being polite with you, you ...

I swear you are back peddiling very hard right now, trying to make it seem as if you are right.

People have already stated that there would be a new FX line, I guess you didn't want to hear it, just keep slaming people for not agreeing with you about AM2.

So all that time you were slamming people, you were wrong and the people who said it was FX were right, sucks to be you, punk a$$.

"Also, the idea that 4x4 will require "new socket" probably comes from here, where it says [AMD] unveils its new "Socket 4x4""

Edward go jump off a fu@K'n cliff, I understand the differences, you just got your hopes and dreams about AM2 on 4x4 dismantled and, I am so glad you did.

Try not to make a mistake this as rude, I had tried to be patient but it doesn't work with you.

Don't bother responding, although youll probably try and spin it so you were right.

In regards to anyone else that reads this all comments were directed at Edward and only Edward, those of you who thought you would be able to use AM2 on 4x4, agin none of what was said is directed at you.

12:40 PM, August 03, 2006  
Anonymous Edward said...

"Well if I recieve any comments from you I am automatically put on the defensive, just look at the comment above."

It's you who decided to put yourself on the defensive, not me. I did not personally attack you, but you claimed I did. Is that how you put yourself on the defensive? I'd say that's a wrong thing to do, really, because that's more like a false accusation to ME.

"How would you take those comments?"

Didn't you just return the same style comments back to me? Did I make a sweat on that? I didn't accuse you for personal attacks when you think I was covering my eyes... because it's not.

"Up until today the latest news about 4x4 was it would be around $1000 for the two processors."

Where do you see that "around $1000"? What about the "well under $1000" link of 7/26 Tech Report?

"Here is your original comment, while your saying it is directed to thekhalif, you make no mention of that...

Therefore I though it was directed at me.


Yes that was directed at you. Is that a personal attack? That was directed at you because I don't understand why you insist 4x4 processors will cost $1000+ while AMD SAID it costs "well under $1000".

"People have already stated that there would be a new FX line, I guess you didn't want to hear it, just keep slaming people for not agreeing with you about AM2."

Why do you turn blind to my comments that the "FX" you/people mentioned was OUT OF THE CONTEXT? As yourself said, it's a NEW FX, and as AMD claimed, the two processor bundle will sell for "well under $1000". Then THAT FX is just the FX brand.

So since Woodcrest is also branded as Xeon, do you suggest it draw as much power and runs as slow? If in February I said, the new Woodcrest will be different from Xeon, are you going to bash me?

"So all that time you were slamming people, you were wrong and the people who said it was FX were right, sucks to be you, punk a$$."

AMD decided to brand the 4x4 processors as FX, and I agree with that, just like I agree with Intel branding Woodcrest as Xeon.

That, as I said, doesn't give anyone the excuse to confuse the "4x4 FX" with current FX, nor does it give you the right to be uncivil again

"Edward go jump off a fu@K'n cliff, I understand the differences, you just got your hopes and dreams about AM2 on 4x4 dismantled and, I am so glad you did.

Try not to make a mistake this as rude, I had tried to be patient but it doesn't work with you.
"

You are just plain rude when you lost your patience, no matter how many "thanks", "please", "appreciated" you said elsewhere. I have and will NEVER use dirty words in whatever heated discussion.

Anyway... so much for the claim that 4x4 will be useless because it'll be too expensive. Two processors and 50% performance over Conroe for "well under $1000", if an enthusiast is not attracted by that, he might as well not buy anything from AMD (read: Intel fanboy). So what if these CPUs are branded FX in the future? They are not the FX we're seeing today (where two of them surpass $1000 by much)! That's the point.

2:28 PM, August 03, 2006  
Anonymous enumae said...

Edward... I do not know why I get defensive when debating with you, maybe my interpretation from past discussions is why, but I am sorry for having lost patience.

"Where do you see that "around $1000"? What about the "well under $1000" link of 7/26 Tech Report?"

The link of 7/26 Tech Report, I had not seen it, but here is the link for the around $1000

"Anyway... so much for the claim that 4x4 will be useless because it'll be too expensive."

I am not sure if this is towards me, as I have no recolection of saying this, if it was over $1000 for the two processors (but its not) it would have been to expensive for me.

"Two processors and 50% performance over Conroe for "well under $1000"

This, I am all for, I would like to own an AMD 4x4 system, though it would not be the high end version.

"if an enthusiast is not attracted by that, he might as well not buy anything from AMD (read: Intel fanboy)."

I completely agree with you, but I still do not understand the fanboy label.

"AMD decided to brand the 4x4 processors as FX, and I agree with that, just like I agree with Intel branding Woodcrest as Xeon."

That said, and this is a legitimate question, will all AM2 FX series CPU's work in 4x4?

If so my understanding about the new socket (AM2 socket) will be incorrect.

Again, I am sorry about that...

3:28 PM, August 03, 2006  
Anonymous Edward said...

"That said, and this is a legitimate question, will all AM2 FX series CPU's work in 4x4?"

There is no single answer, but -

* If current AM2 FX will work with 4x4, nothing would prevent current X2 to also work with 4x4.

* OTOH, if current AM2 X2 will NOT work with 4x4, there's no reason that current FX could work with 4x4.

In short, no evidence (at least not revealed to me) shows FX and X2 having different architectures. They have the same core with the same number of HT links, except FX's multiplier is unlocked..

"If so my understanding about the new socket (AM2 socket) will be incorrect."

Well, I think what is fair to say is that AMD has been pretty vague about it. They probably are debating it themselves. On one hand, it'll be a good source of cost and confusion if yet another socket/processor type is released; on the other hand, they don't want 4x4 to eat the 2p Opteron market.

But IMO, 4x4 wouldn't eat any high-margin 2p Opteron market simply due to its lack of RAS capability. The ONLY reason for AMD to release a new (FX) brand of processor should be to beef up the HT links.

BTW, ALL Athlon64 processors have 3 HT links: HT0, HT1, and HT2. It's just two of them are much slower - at 200Mhz. If you own an Athlon64 system, download & run CPUInfo from AMD website and you'll see.

"Again, I am sorry about that..."

Not a problem. It happens when those silly discussions that got too heated up. ;-)

4:10 PM, August 03, 2006  
Anonymous enumae said...

Thanks Edward.

So when you say...

"The ONLY reason for AMD to release a new (FX) brand of processor should be to beef up the HT links."

Is that to make them all the same speed, and replace the 2x200Mhz links with a higher clocked link?

The inner workings are where I get confused.

8:48 PM, August 03, 2006  
Anonymous Edward said...

""The ONLY reason for AMD to release a new (FX) brand of processor should be to beef up the HT links."

Is that to make them all the same speed, and replace the 2x200Mhz links with a higher clocked link?
"

I honestly don't know.

It should be possible to have two 1Ghz HT link, like the 940-socket 2p Opteron. But I'm not a motherboard design guy, and I really don't know.

9:24 PM, August 03, 2006  
Anonymous Anonymous said...

I thought you'd want to know that Socket F has been delayed again.

http://www.theregister.com/2006/08/05/survey_server/

As it stands, however, the Rev F chips will be more delayed than previously expected. AMD already pushed back its official "launch" announcement from July to August. Now we hear that HP doesn't expect to slot Rev F chips into boxes until "late September or the first week of October."

So it looks like Socket F will be a paper launch in August with availability in September/October. Now I know you made fun of the lack of Woodcrest availability, but it seems that AMD is no better. It seems now that Socket F has slipped a full quarter. Supposedly the reason is that AMD is going to leave Woodcrest alone by having a quiet Socket F launch just to get the platform on the market and then stage a come-back next year when quad cores are ready.

10:20 AM, August 05, 2006  

Post a Comment

Links to this post:

Create a Link

<< Home