IBM to launch five Opteron servers tomorrow
It's Opteron time again. SUN introduced the 8P 16 core x4600, the 24TB x4500, the massive 4P Opteron blade 8400. The SUN x4600 smashes 16P HP Itanium Superdome and is very competive against 16P Power 5+, but at a fraction of the cost. Not to be left behind, IBM is going to unleash five new Opteron product lines, including two blade designs, which can be upgraded from 2P to 4P.
As I told you, 4P will be the predominant configuration in enterprise computing. AMD Direct Connect Architecture with glueless 8P ccNUMA made this possible. Just as Intel trying to catch up, the rule changes, and 4P is the name of the game. I expect IBM to EOL its x3 chipset -- too costly to follow Intel's changing FSBs. Woodcrest will be pushed to the ultra low end. The same is happening in performance desktop, moving from 1P to 2P.
In business desktop market, AMD64 is now inside ThinkCentre.
Digitimes reported that DELL's all out AMD64 flood will start in September 2006. That's only 30 some days away. Exciting times ahead, folks.
HP better hurry up.
posted by Sharikou, Ph. D. @ 7/31/2006 09:49:00 PM
41 Comments:
And Digitimes said Dell AMD PCs in September and Dell AMD notebooks in October or November.
Amazing, AMD must have enough volume to lower the price points to seal these deals.
Amazing, AMD must have enough volume to lower the price points to seal these deals.
AMD's capacity right now can supply 33% of the market.
Sun claims “Performance - Up to twice the performance of dual core Xeon servers” This is just the beginning kids. There’re so many brilliant minds in partner with AMD that I here by proclaim; GAME OVER.
"IBM is going to unleash five new Opteron product lines, including two blade designs"
How about the 550 Core 2 Duo design wins (already!!)...?
Oh hey BTW Shakira, I read in one of your other posts that you have finally accepted that Core 2 Duo is faster than Opteron, even though you say its only for a few benchmarks. Wow, must have been hard for you ain't it?
Oh and looks like you had to open up a new topic to get away from all that PhD discussion...
The Chinese link basically says that Taiwanese NB vendors confirm that Dell orders a full line-up of "AMD inside" notebooks, which will be on sale in October or Novenber this year.
Note that this means the Dell's Intel-only-ness will be completely gone, from server, desktop, to notebook.
It also says that Dell had prepared for AMD-based notebooks for a number of years before; every time it was given up due to Intel's favoritism. Now it seems Intel no long offers the incentive good enough for Dell to remain its exclusive loyal.
However, even after the launch of these AMD-based machines, Intel is still Dell's largest provider.
Here's one more example of Intel's business ethics.
http://www.khaleejtimes.com/DisplayArticleNew.asp?xfile=data/business/2006/August/business_August28.xml§ion=business&col=
How about the 550 Core 2 Duo design wins (already!!)...?
Oh hey BTW Shakira, I read in one of your other posts that you have finally accepted that Core 2 Duo is faster than Opteron
1) Intel is going to supply 1 million Conroes in the next 7 weeks. 1 million/7/7/550= 37
2) Woodcrest is 16% slower, just scroll down and look at the benches. That's why IBM is releasing five Opteron servers, 0 woodcrest.
With applied science from the Sun and IBM brain trust coupled with AMD’s flexible architecture a lot of these exciting technologies in vanilla favor will drip down to future desktop computing.
I’m curious. What rebuttal can an Intel aficionado make? Are they still debating 32 bit, FSB and over-clocking on the eve of a revolution in computing? A flood gate of innovation has just been opened. New debates on coprocessors married to GPU and Torrenza hasn’t even begun yet. Untold hundreds of scientist and programmers have recently redirected there thoughts to these new technologies.
Intel’s proprietary model is not part of this knowledge base; they are still making P4’s.
Woodcrest and Conroe are fast but obsolete, they are cousin to P4. Intel is at the end of the road with its contemporary offerings.
With AMD now living on Main Street and Intel’s Xeon / Woodcrest banished to life in the back alleys serving the low end server market. The Intel brand has taken an enormous hit. I’m sure this massive physiological contusion has damaged the once proud Intel fan club to a point of utter embarrassment.
Dell/AMD server, desktop & laptop news must be painful also. Intel’s market share just slipped another notch. Imagine that, Dr. S right again.
I don’t receive any personally satisfaction from your agony over Intel’s demise, but I do wonder what it will take for you to relinquish yourself from Intel’s grip. I’m fascinated by the brand loyalty at all cost mentality. I need help to understand the affection.
I don't quite understand it. Pretty much every technical site on the planet has conceded that the Core 2 Duo line-up simply dominates its AMD counterparts. In every single segment except the MP server segment, Intel's new wonder has a substantial lead. And in the MP segment, the only reason AMD is still competetive is because of its interconnect (which is admittedly quite good), not because of its microprocessor architecture. Keep in mind that as soon as Intel gets around to releasing it's serialized interconnect (slated to replace FSB by 2008), AMD will be competetive in NO segment.
Another thing I don't understand: where exactly is this "revolution" in computing coming from? AMD's 4x4? Even the most pro-AMD tech sites have been skeptical that people will go for two underperforming processors over a single higher performing core. Maybe this revolution comes with AMD's aquisition of ATI and the announcement to develop platforms? Intel has been a self-proclaimed platform company for years. So where is this innovation? Where is this "revolution"? I keep looking for it, but all I can see is an AMD struggling to survive.
As the competetiveness of AMD has dropped like a rock as of late, so too has the credebility of this site.
See, IBM is a partner of AMD so there is no surprise if they announce Opteron servers with fanfare. They also have Xeon servers, which probably are selling much more without the need for such publicity.
Sun is an ex-Intel competitor which lost out to Intel, so it suits them to be in Bed with AMD.
What counts is not what's on the market. It is what the customers are buying. Even with Prescott Xeons, Intel had a dominant position in x86 Servers though they were losing marketshare to Opterons. Now with Woodcrest the trend will reverse. Volume of Opterons will not go down to 0 overnight. So there will be Opteron based products on the market. That doesn't mean those products are leading the market. Far from it.
"Oh hey BTW Shakira, I read in one of your other posts that you have finally accepted that Core 2 Duo is faster than Opteron, even though you say its only for a few benchmarks. Wow, must have been hard for you ain't it?
Oh and looks like you had to open up a new topic to get away from all that PhD discussion... "
Do you have anything intelligent to say? Or are you Intel fanboy's only tactic to continue personal attacks?
YOU might not like it... but there are actually others who come here for a different view on the way of things (the NON-Intel way).
Now, can you just go out, buy a Conroe and be happy with it?
Pretty much every technical site on the planet has conceded that the Core 2 Duo line-up simply dominates its AMD counterparts.
You were only looking at 32 bit desktop performance, where Con XE 6800+ is about 10% faster than FX62. But the game has changed even on desktop, 4x4 will boost performance by 80%. Now, catch that.
On servers, Intel is at the ultra low end, while everyone has gone at least 4P. That's why IBM and DELL must go AMD full force. As you can see from the benchmarks, the SUN x4600+ frags 16P Itanium 2 Superdome and is very competitive against the Power 5, at a fraction of the cost.
4x4 will boost performance by 80%. Now, catch that.
Where will that other 20% go?
As you should know, there will be dualchip/quadcore Core2 availiable roughly by the launch time of 4x4 to combat against it.
One thing I would like to know is that how will NUMA act on desktop PC? Has anyone made any tests on, say, 2GHz dualcore vs 2x 2GHz singlecore? Let's say that both systems have their best RAM setups (2x1G for x2 and 4x512 for 2P). I'm not interested in server tests but in stuff what people run on their PC's (games, video editing, rendering etc).
shakirou said...
It's Opteron time again. SUN introduced the 8P 16 core x4600, the 24TB x4500, the massive 4P Opteron blade 8400. The SUN x4600 smashes 16P HP Itanium Superdome and is very competive against 16P Power 5+, but at a fraction of the cost. Not to be left behind, IBM is going to unleash five new Opteron product lines, including two blade designs, which can be upgraded from 2P to 4P.
does that mean opteron killed the sparc architecture?
Woodcrest is 16% slower, just scroll down and look at the benches. That's why IBM is releasing five Opteron servers, 0 woodcrest.
More cherry picking by Shakira (I personally like Sharifraud)... Chooses the one or two benches that Opeteron outperforms. Then he concludes that is why IBM is releasing 5 new Opteron servers and no Woody's. Hmmm, maybe there is another explanation... maybe it's because Woodcrest if FREAKIN' new and it takes some time to get a product out the door. I guarantee you IBM will be announcing Woodcrest based server systems in this quarter. Will you conclude that the new Xeons are better based on that? I didn't think so because of your obvious biases.
According to Shakira it's because Opteron and AMD are better, better, always better. Man, if you're not getting paid to pump by AMD you are the dumbest PhD I've ever known. At least you should be paid for your sycophantism rather than run a pathetic little blog for free.
It is the Xeon which killed Sparc.
As the competetiveness of AMD has dropped like a rock as of late
Maybe in your own little world this is true, but in the rest of ours the opposite has occurred. With more competitively priced chips, and steadily increasing production volume, AMD has been getting MORE competitive. Maybe in a month when the average early-adopter can start buying conroe chips this will change, but as of late, it has been AMD's ballgame.
does that mean opteron killed the sparc architecture?
Sparc is OK for now. In 256P configuration, Sparc has linear scaling. Also, the coming Rock processor will be 16 core and SMP, that goes up to 256 core very easy. Also, rock will use AMD's Socket F, so you will see SUN servers running both SPARC and AMD64 in the same machine...
Why is this big news?
It's interesting to see how much hooplah is generated by something that should mean so little to so few.
There is no computing revolution these days. Things are just becoming faster. Nothing new here.
What AMD and Intel need to do is to come up with new technology that doesn't rely on transistors and silicon and die sizes.
Sometime soon a new company will come along and blow away the companies that streamline and reproduce this old technology.
Until then, this is all very meaningless in the grand scheme of things.
Where’s the revolution you ask? I’m sure you are kidding. P4’s are for under privileged third world countries, Intel has seceded the server market on there own conference call, Conroe is a temporary product fix, Woodcrest already obsolete to 4X4, Xeon is now very undesirable. Please keep up.
Any computing in x86 64 that is compiled with Intel technology will automatically be handicapped, will require more clock cycles, jump routines just to make the same calc. It is unlikely if not impossible for Intel to compete without AMD64 original code.
Intel has already reached its zenith, they desperately needed 65nm just to maintain not to surpass. Think ahead friend, what’s going to happen when AMD is at 65nm? Power to watt ratios attained will be out of Intel’s league. MS Vista has compounded Intel’s problem.
If you have to ask, “What revolution”? You are already too late. My next statement may shock you; the only hope for Intel is to learn 64 then beat AMD to 128 bit computing. The next window of opportunity for Intel is a decade away. Proof, last month Intel and its fanboys were making the case for 32 bit, need I say more.
Please read past articles on this and other blogs to come up to speed.
Sharikou, Ph. D ?
What does the Ph. D. stand for.
A doctorate in manpulating data to favor an AMD position. No amount of spining even around 64bit can change the fact that Conroe, Merom and Woodcrest are the best chips and platforms overall for dual-core applications.
I've heard of cold fusion sillyness
and cloning manpulation
But this Doctorate deserves his Ph.D revoked.
Why do you even have it in your title?
"The Intel brand has taken an enormous hit. I’m sure this massive physiological contusion has damaged the once proud Intel fan club to a point of utter embarrassment."
Except the Intel brand, and I may be wrong, is better known for Pentium than Xeon.
"I do wonder what it will take for you to relinquish yourself from Intel’s grip."
For what I do, my Intel 940 is fine, I can play games and do some photo manipulation. It runs pretty cool and have had no problems with it.
"I’m fascinated by the brand loyalty at all cost mentality"
As am I looking at your statements. Would you claim to not be brand loyal to AMD?
I just have to ask, do you accept that a E6700 beats an FX62?
Sorry if that comes across a little harsh, but there is no point in being a fanboy, quality/performance/price is all that should influence your purchasing.
I know C2D's are not available in quantity, probably be a few weeks, but I would not mind having one.
4x4 looks even better, though I would like to see that price come down quickly after launch, but well just have to wait and see.
“No amount of spining even around 64bit can change the fact that Conroe, Merom and Woodcrest are the best chips and platforms overall for dual-core applications.”
Intel’s inefficient 64 bit code is like dragging an anchor. Can you comprehend the nature of this problem?
Again, I ask "What revolution"?
Computing will still be the same, only faster. You will still have a little beige box sitting next to you, generating heat and consuming electricity, pumping out pixels to a screen and manipulating electrons across simple wires and pathways.
There really is nothing new. The computer essentially hasn't changed since the transistor was invented, it's only gotten faster.
I don't have the answer, but I know that sticking more CPU's in a package will not change the world. You speak of a "revolution", but there is no drastic change in ways of thinking at present in the computing world.
AMD and Intel think 'inside the box', no matter how much either party believes. Constantly improving on old methods. Making things smaller and faster.
32, 64, 128 bit processing is nothing but expanding on old ways of thinking. God, the way things are portrayed in the little land of cpu's, you'd think there was something important and life changing going on. But there's not. Same crap, year after year.
I don't have any sort of ideas, but I do know that people are excited over nothing really.
I'll be excited when something truly new comes around.
“I just have to ask, do you accept that a E6700 beats an FX62?”
You make a good point and I commend you on the rest of your well thought out post
"I just have to ask, do you accept that a E6700 beats an FX62?"
If you want the fastest (not readily available chip for now), then go with Conroe.
But for some of us, we like choice, and the "future" upgrade ability of the AM2 socket. I think of it as an investment, and if someone can offer something to me now, and say... "hey, in the future when our prices drop, or Quad core comes out, you are able to upgrade easily". Well, that's my choice.
Talking about revolution. It's narrow minded to only think of more processors. Torrenza offers up multi-socket to other co-processor, etc in the future... migration, upgradeability!!
So what if I get 68FPS with Conroe on Oblivion and 62FPS on an Athlon running at very high resolution? My eye can't tell the difference since the human eye can't detect such.
MESSAGE TO INTEL FANBOYS: If you have nothing but personal attacks, and no intelligent argument, go back to Anandtech or Tom's and praise Intel over there. :)
- The Rest of the Bloggers sick of childish behavior.
the only hope for Intel is to learn 64 then beat AMD to 128 bit computing
Considering one of the primary motivations for the jump to 64 bits is/was memory address space limitations, the jump to 128 bits is millenia away. We wont be using silicon by that time, so effectively it wont matter.
There really is nothing new. The computer essentially hasn't changed since the transistor was invented, it's only gotten faster...I'll be excited when something truly new comes around.
AMD and Intel are essentially engineering firms at their core. What you are looking for is research breakthroughs, which aren't going to come from them, but more likely from Academia. You will be waiting a long time for something that big to happen. I suggest you grab a snack.
Anonymous said...
"But for some of us, we like choice, and the "future" upgrade ability of the AM2 socket. I think of it as an investment, and if someone can offer something to me now, and say... "hey, in the future when our prices drop, or Quad core comes out, you are able to upgrade easily". Well, that's my choice."
It is definetly your choice, but are you trying to take my comment out of context, or are you just trying to say that if you bought an E6700, you could not upgrade that as well?
I am not going to get into quadcore, but I will say one thing, there is a guy running Kentsfield on an Intel 975 chipset, which also supports Conroe. Here is a link.
I will not bash AMD, they have great products, and I would like to own a 4x4 system, I think it would be great.
My whole point was a lower clocked/price/power conuming chip is beating AMD's best chip.
This is temporary, and I understand that, but from a consumers point of view it is very impressive.
As to your other question...
"So what if I get 68FPS with Conroe on Oblivion and 62FPS on an Athlon running at very high resolution?"
It's all about the price you paid to get 62fps on that Athlon FX62, it would cost you about $200 more to be 9% slower, when comparing to the E6700.
If you look at it like an investment, you will see this.
There really is nothing new. The computer essentially hasn't changed since the transistor was invented, it's only gotten faster...I'll be excited when something truly new comes around.
AMD and Intel are essentially engineering firms at their core. What you are looking for is research breakthroughs, which aren't going to come from them, but more likely from Academia. You will be waiting a long time for something that big to happen. I suggest you grab a snack.
Got my sandwich here.
This is what people need to invest their brain cells into. http://www.research.ibm.com/autonomic/manifesto/
Maybe not so far in the future? I'll enjoy seeing both AMD and Intel grovelling and withering up because they are not putting enough into R&D and instead choose to outdo each other in the same game time after time.
But again, no revolution. Nothing new here. AMD INTEL BLAH BLAH.
I think 4X4 will come after AMD lounchs 65 nm procesors. It's more reasonable, cheep, energy efficient and so on. much bettter opportunity than conroe, with puture upgrade fot K8L, so intel has no chances.
"Considering one of the primary motivations for the jump to 64 bits is/was memory address space limitations, the jump to 128 bits is millenia away. We wont be using silicon by that time, so effectively it wont matter."
I would not agree on first phrase. its intel historical view on amd64, that its about 64 bit memory adressing space only. so it took some time for them to get to whole amd64 specification (and its still not fully functional even in conroe - macro op fusion does not work, etc).
however 64 memory adressing is only part of it. amd64 is arch enhancement (like sseX) as well. more registers, additional commands. Im working in company providing solutions for financial companies. just simple amd64 recompilations shows +20% performance enhancement for our software. thats because of additional registers (which can be seen as ultra fast L0 cache) and new commands.
"With AMD now living on Main Street and Intel’s Xeon / Woodcrest banished to life in the back alleys serving the low end server market. The Intel brand has taken an enormous hit."
While I myself prefer AMD's technologies, the comment above got to be false. After false, Intel still commands 70%+ server market. Xeon is still the hottest (pun intended) x86 server on the market yet.
Xeon is still the hottest (pun intended) x86 server on the market yet.
I doubt it. Once you take out those Celeron and Pentium 4 servers, the picture may change. AMD, on the other hand, is mainly Opteron.
I'm quite sure I already replied to the following quote but for some reason I can't find it here.
the only hope for Intel is to learn 64 then beat AMD to 128 bit computing
Care to explain why is 128bit computing better than 64bit computing? Better yet, start with the definition of 128bit CPU.
If that definition says "external memory bandwidth" then you are couple of years late. If it sais "some registers are 128bit" then yet again you are years late. If it sais "GP registers are 128bit" then you are right on definition but wrong overall.
There is no need for so big GP registers, especially when address space and pointer sizes increase too. Even now you can see that in some apps moving from 32bit to 64bit makes things worse. Mostly that is because of bigger pointers, increased memory bandwidth and increased inefficiency of cache line usage.
just simple amd64 recompilations shows +20% performance enhancement for our software
I can confirm that on EM64T too, with certain software. At least ray tracing can benefit 50-100% increase in speed (depending on scene) thanks to twice the number of registers and the ability to trace sixteen rays simultaneously without registry spills.
Of cource in software that uses pointers a lot, increased memory bandwidth might not let the benefits to show that well. IIRC I havent seen any gaming benchmarks where 64bit versions would beat 32bit ones by big margain, usually speed difference is less than 5%
[rant]
Too bad that x86 has so little registers, even under AMD64/EM64T. Cell's SPU has 8 times as much registers and even though it has only a fraction of the transistors of modern x86 CPU's it performs almost as good. That is dual Cell machine with 16 SPE's can be up to 12x faster than single-core K8 on same clocks in ray tracing. In terms of transistor count and performance Cells beat K8 five to one. Not that bad, I would say.
http://www.sci.utah.edu/~wald/Publications/2006///Cell/download//cell.pdf
Too bad world hasn't yet moved to massively multithreaded apps. I personally would trade around 50% of single-core performance for 8 cores any day. Seems I'll just have to either wait for couple of more x86 NGMA's or just go with Cell or it's future revisions :)
[/rant]
I doubt it.
As have been said several times, better CPU does not translate to market domination, at least not very quickly. I can say that in our server room we have around 80% of Intel-based servers. Of course if I had saying in what should be bought that number would most likely be much less by now.
however 64 memory adressing is only part of it. amd64 is arch enhancement (like sseX) as well. more registers, additional commands. Im working in company providing solutions for financial companies. just simple amd64 recompilations shows +20% performance enhancement for our software. thats because of additional registers (which can be seen as ultra fast L0 cache) and new commands.
Yes but alot of that is largely independant of whether it uses 64 bit addressing or not. Opcodes can be added to existing instruction sets, often without changing addressing size. As such, instruction set improvements will continue through extensions, even if it will never be practical to use 128 bit addressing.
“Care to explain why is 128bit computing better than 64bit computing? Better yet, start with the definition of 128bit CPU.”
Making reference to some existing technology where 128 bit is used for very long instruction or 4 instructions per clock cycle using 128 bits.
http://www.transmeta.com/crusoe/vliw.html
YOU might not like it... but there are actually others who come here for a different view on the way of things (the NON-Intel way).
You mean the close-minded Sharikou way... sure I see that.. This place is junk and as such there is no harm done when you spew junk here...
Of course, some mis-guided souls think they are discussing *important and meaningful* things on Sharikou's blog.. good for you, whatever helps you to sleep...
Wow IBM has 5 more lines of servers.
Its like Ranger Rover annoucing they have another country distributing their SUVs. Sure sales will go up a bit... but in the end won't amount to a hill gass out of Sharkou's ass.
Sure Opertons are the best.. or people believe its the best. What the hells does best mean anyway. INTEL will get closer in two quarters. It'll be close enough and their leadership in the mainstream which represent like 80% of all cpu's shipped will win the marketshare wars, profits, and MONEY.
We all know Sharikou Ph.D ( of ? education and degree ) has stated again and again INTEL will soon lose money and be bankrupt in 5 quarters.
Of course he has the caveate that if they cut 30% of the workforce they might survive. In 5 quarters the Doctor of lies will have an excuse why INTEL didn't go bankrupt and how it is still the most profitable semiconductor company and the largerst by 2x will be turned into more nonsense.
PhD... is that a joke or something. He has a PhD in spewing BS.
Wow IBM has 5 more lines of servers.
Its like Ranger Rover annoucing they have another country distributing their SUVs. Sure sales will go up a bit... but in the end won't amount to a hill gass out of Sharkou's ass.
Sure Opertons are the best.. or people believe its the best. What the hells does best mean anyway. INTEL will get closer in two quarters. It'll be close enough and their leadership in the mainstream which represent like 80% of all cpu's shipped will win the marketshare wars, profits, and MONEY.
We all know Sharikou Ph.D ( of ? education and degree ) has stated again and again INTEL will soon lose money and be bankrupt in 5 quarters.
Of course he has the caveate that if they cut 30% of the workforce they might survive. In 5 quarters the Doctor of lies will have an excuse why INTEL didn't go bankrupt and how it is still the most profitable semiconductor company and the largerst by 2x will be turned into more nonsense.
PhD... is that a joke or something. He has a PhD in spewing BS.
wow, it's awesome, very interesting
Post a Comment
<< Home