Intel Itanium outsells AMD Opteron

[Katanai]

A $200K HP Superdome

Hahahaha!

Try multiplying that by 10. See, that's why these systems are unknown to the general public. They are in a totally different league. That's also why I think this comparison to the Opteron is a bit unfair. The only thing they got in common is that they are both CPU's. Any other similarity ends there. Still there is no one stopping AMD to come out with a high end architecture of their own. Yet they don't, the only thing they got is the Opteron, that to be frank, was totally killed, not by Itanium but by Nehalem. We'll see though, they have a chance to make a comeback with these 12 core chips that are soon to be released...

[saaya]

Hahahaha!

Try multiplying that by 10. See, that's why these systems are unknown to the general public. They are in a totally different league. That's also why I think this comparison to the Opteron is a bit unfair. The only thing they got in common is that they are both CPU's. Any other similarity ends there. Still there is no one stopping AMD to come out with a high end architecture of their own. Yet they don't, the only thing they got is the Opteron, that to be frank, was totally killed, not by Itanium but by Nehalem. We'll see though, they have a chance to make a comeback with these 12 core chips that are soon to be released...

considering that amd still manages to sell opteron chips for a thousand bugs a piece, i wouldnt really say nehalem has killed the opteron
especially since your saying a certain architecture killed a marketing name and series of another company... which doesnt make sense :P

and whether amd will have a come back or not is not a question... their cpu designs are way focussed on the server space, and we all know that bulldozer is a sever chip first and a desktop and mobile chip second... just like nehalem...im sure bulldozer will be a great chip... the thing is just that im sure sandybridge will at least be as much of a great chip as bulldozer...

id really like to know why bulldozer didnt come out in 2009 as planned... that would have been quite something...
amd could have caught up to intel and even surpassed them...

[therightway]

The only graph dating back about 10 years that I could find with a minimal search shows both AMD and Intel's overall market share. Not server CPU specific.

This graph must be of retail CPU marketshare. I know in server and mobile CPUs, Intel has a higher share.

[JF-AMD]

Bulldozer in 2009 would have been a DDR-2 based processor. There were changes made to the processor that also allowed it to get more performance and scalability. The Bulldozer that was defined for 2009 is not what you will be seeing in 2011. We made the changes and realigned the schedule to make it more compeititive.

[Katanai]

considering that amd still manages to sell opteron chips for a thousand bugs a piece, i wouldnt really say nehalem has killed the opteron
especially since your saying a certain architecture killed a marketing name and series of another company... which doesnt make sense :P

Yeah but that's exactly what happened and it does make sense. The Nehalems might still be called Xeon but we're talking about a totally different chip here. A chip that was specially designed for the server market with only one goal in mind: to destroy the Opteron. And quite frankly it has succeeded in doing that. There's no need to be indirect here, we all know how Intel thinks as a company. The Opteron had some key advantages over the old Xeon chips, which made it successful, especially in multiprocessor configurations. Namely the HyperTransport link, a faster integrated memory controller and you could argue that it also could have a higher CPU count per socket as the 6 core versions were more accessible and a bit more widespread than the Intel equivalent. The Nehalem architecture negated all these advantages, not only that but it surpassed it in all these aspects with the QuickPath Interconnect, integrated triple channel DDR3 controller and Hyper-Threading Technology. That's the truth really and that's why Nehalem is spreading like crazy on the server market while the Opteron, at best, stagnates.

[god_43]

that graph is utter bull, do not post it again (especially without a source).

amd and intel were never at parity. i don't see any mention of amd being over 30% market share in any segment, in this link?


http://www.marketwatch.com/story/int...n-server-chips


lets not revise history shall we?



edit: wtf is "http://www.kalionzes.net/" (where the graph is from)? its not even a technology site?

[Dresdenboy]

that graph is utter bull, do not post it again (especially without a source).

amd and intel were never at parity. i don't see any mention of amd being over 30% market share in any segment, in this link?


http://www.marketwatch.com/story/int...n-server-chips


lets not revise history shall we?



edit: wtf is "http://www.kalionzes.net/" (where the graph is from)? its not even a technology site?

I know that graph in earlier variants (less quarters displayed). IIRC it shows retail desktop market share. Numbers should be from one of the known ones (Gartner, Mercury or IDC) - but these are also estimations if I'm correct.

[wuttz]

my day is never complete w/o reading an awesome spin ad from spintel. keep on spinning! xD

[god_43]

AMD owned 25.3% of the total market, up from 23.3% a year ago, based on calculations from AG Edwards analyst David Wong, who cited preliminary data from Mercury Research.

theswe are estimates from mercury at that time....but everyone else was pretty much saying the samething?


http://www.tgdaily.com/business-and-...2-report#close

http://arstechnica.com/old/content/2006/01/6053.ars


http://www.digitimes.com/systems/a20070215PD218.html


i also find it interesting to read the old comments from these articles...its like going back in time!





edit: here is another good one: http://www.digitimes.com/bits_chips/a20070316PR203.html

[anubis]

IIRC AMDs push of x64 was the thing Intel was afraid the most. It was the stinger in their ass that showd you dont need 4k cpu to make a supercomputer

And intel was forced to introduce x64 in "ordinary" cpus, although they really-really didnt want to.

edit: http://en.wikipedia.org/wiki/X86-64

History of AMD64
AMD64 was created as an alternative to Intel and Hewlett Packard's radically different IA-64 architecture. Originally announced in 1999 with a full specification in August 2000,[1] the architecture was positioned by AMD from the beginning as an evolutionary way to add 64-bit computing capabilities to the existing x86 architecture, as opposed to Intel's approach of creating an entirely new 64-bit architecture with IA-64.
The first AMD64-based processor, the Opteron, was released in April 2003.

[xlink]

+1
Each Itanium CPU costs at least 10x more I bet... Not even talking about all the other hardware that's required to run it.

the OTHER hardware is comparable to everything else, I do seem to recall hearing about socket compatibility with nehalem systems...

[xlink]

I think his main point is that Itanium doesn't even execute the same code.

yes it can. do some research. IA64 can run x86 code, just not well.

[JumpingJack]

IIRC AMDs push of x64 was the thing Intel was afraid the most. It was the stinger in their ass that showd you dont need 4k cpu to make a supercomputer

And intel was forced to introduce x64 in "ordinary" cpus, although they really-really didnt want to.

edit: http://en.wikipedia.org/wiki/X86-64

http://www.theregister.co.uk/2008/02...d_weber_scc13/

Here is an interview of former AMD CTO Fred Weber (one of the great technologist in the industry, and one of the fathers of the K6/K7/K8 lineage)... it gives a great insight into what AMD was thinking when they put in AMD64 into the mix.

[duploxxx]

Yeah but that's exactly what happened and it does make sense. The Nehalems might still be called Xeon but we're talking about a totally different chip here. A chip that was specially designed for the server market with only one goal in mind: to destroy the Opteron. And quite frankly it has succeeded in doing that. There's no need to be indirect here, we all know how Intel thinks as a company. The Opteron had some key advantages over the old Xeon chips, which made it successful, especially in multiprocessor configurations. Namely the HyperTransport link, a faster integrated memory controller and you could argue that it also could have a higher CPU count per socket as the 6 core versions were more accessible and a bit more widespread than the Intel equivalent. The Nehalem architecture negated all these advantages, not only that but it surpassed it in all these aspects with the QuickPath Interconnect, integrated triple channel DDR3 controller and Hyper-Threading Technology. That's the truth really and that's why Nehalem is spreading like crazy on the server market while the Opteron, at best, stagnates.

are you actually active in the server world? Nehalem didn't destroy the opteron at all, yes Intel gained some market share against opteron due to the performance lack in some SW parts previous xeon versions had and nehalem no longer, but the response with istanbul stagered that again in many areas. Lots of sales on AMD is now waiting for the new platform launch, which Intel is quite affraid off, since they are changing roadmaps all the time with gainestown to counter that launch. Don't forget AMD doubled the amount cores and 4 channel ddr3 MC in a 2s box that is for example a lot of virtualization performance.

Not everyone buys the top bin you know, and AMD systems are way cheaper with OEM discount then Intel.

[Katanai]

are you actually active in the server world? Nehalem didn't destroy the opteron at all, yes Intel gained some market share against opteron due to the performance lack in some SW parts previous xeon versions had and nehalem no longer, but the response with istanbul stagered that again in many areas. Lots of sales on AMD is now waiting for the new platform launch, which Intel is quite affraid off, since they are changing roadmaps all the time with gainestown to counter that launch. Don't forget AMD doubled the amount cores and 4 channel ddr3 MC in a 2s box that is for example a lot of virtualization performance.

Not everyone buys the top bin you know, and AMD systems are way cheaper with OEM discount then Intel.

Yes actually I am and like I've said Nehalem is selling like crazy. It's a very successful architecture. And the part with AMD systems being way cheaper is more of a myth than anything. The platform as a whole is a bit cheaper in most cases but the difference is not as big as you make it out to be. I don't know where you have the info from that AMD is making lots of sales right now. As for virtualization performance, until AMD implements an analogue to Hyper-threading they don't got anything going for them. Doubling the amount of cores is not really the best solution as it also almost doubles the costs. We'll have to wait and see, 12 core CPU's sound good. If they manage to take the next step and manage to implement a form of Hyper-threading technology to the next generation, like they are talking now, those CPUs should be monsters. Don't get me wrong though, I got absolutely nothing against AMD. The Opteron was a revolutionary CPU and it actually sold less than it should have. I'm just saying it how it is at the moment...

Anyway enough thread derailing, this is about Itanium not Nehalem...

[duploxxx]

Yes actually I am and like I've said Nehalem is selling like crazy. It's a very successful architecture. And the part with AMD systems being way cheaper is more of a myth than anything. The platform as a whole is a bit cheaper in most cases but the difference is not as big as you make it out to be. I don't know where you have the info from that AMD is making lots of sales right now. As for virtualization performance, until AMD implements an analogue to Hyper-threading they don't got anything going for them. Doubling the amount of cores is not really the best solution as it also almost doubles the costs. We'll have to wait and see, 12 core CPU's sound good. If they manage to take the next step and manage to implement a form of Hyper-threading technology to the next generation, like they are talking now, those CPUs should be monsters. Don't get me wrong though, I got absolutely nothing against AMD. The Opteron was a revolutionary CPU and it actually sold less than it should have. I'm just saying it how it is at the moment...

Anyway enough thread derailing, this is about Itanium not Nehalem...

hyperthreading in virtualization is only usefull when you have huge amount of very small vm's, once you have many vCPU VM with avarage/high load hyperthreading is useless, real cores aren't. perhaps check this and look at my remark below about price and question again why so many people buy intel for virtualization: (not to mention before nehalem was there or the crap 7xxx series in virtualization) http://it.anandtech.com/IT/showdoc.aspx?i=3571&p=10
It's only the top bin that is a bit faster but also double price.

I assume that you are talking about retail prices, I talk about major volume where an AMD 2435 is almost equal to E5530 in price. Now check again where the price/performance ratio will be..... not to mention soon to be released 4000-6000 series adding cores isn't the biggest cost, that is small die size.

[kl0012]

I assume that you are talking about retail prices, I talk about major volume where an AMD 2435 is almost equal to E5530 in price. Now check again where the price/performance ratio will be..... not to mention soon to be released 4000-6000 series adding cores isn't the biggest cost, that is small die size.

I would say that even if we accept all your assumptions about the volume prices, even then E5530 will be probably a bit better option then 2435.

Spec.org dual-cpu system.
SpecInt_rate/SpecFP_rate:

Opteron 2435 205/143
Xeon E5530 209/165

Sorry for offtop.

[JF-AMD]

The comparison of the X5570 to the Opteron 2435 is always interesting. They have a raw performance edge, but the processor is priced at $1386 vs our $1019. That makes it 36% more expensive. If they have less than a 36% performance advantage (which is almost always the case) then the AMD is going to deliver better price performance.

I've done the math in other places because there are some that will arge those are the list prices and system price is different. When you configure like for like system on, for instance, Dell's site, you see the price delta go UP and at the platform level the price differential was higher.

[qurious63ss]

The comparison of the X5570 to the Opteron 2435 is always interesting. They have a raw performance edge, but the processor is priced at $1386 vs our $1019. That makes it 36% more expensive. If they have less than a 36% performance advantage (which is almost always the case) then the AMD is going to deliver better price performance.

I've done the math in other places because there are some that will arge those are the list prices and system price is different. When you configure like for like system on, for instance, Dell's site, you see the price delta go UP and at the platform level the price differential was higher.

What about overall system prices? or do most people buy only upgrade CPUs?

[kl0012]

The comparison of the X5570 to the Opteron 2435 is always interesting. They have a raw performance edge, but the processor is priced at $1386 vs our $1019. That makes it 36% more expensive. If they have less than a 36% performance advantage (which is almost always the case) then the AMD is going to deliver better price performance.

I've done the math in other places because there are some that will arge those are the list prices and system price is different. When you configure like for like system on, for instance, Dell's site, you see the price delta go UP and at the platform level the price differential was higher.

I am little surprised to hear such things from a marketing guy.
First, a cpu price difference is never proportional to actual performance difference even in the same line of cpus. But you probably know it even better then me.
Second, your performance/price equation is partially useful for home PC user but in busines/server space it is completely BS. Even if X5570 only 20% faster then 2435 do you really think that %20 more work which potentially can be done in, lets say, two years, costs only $250 diff between cpus?

[therightway]

I am little surprised to hear such things from a marketing guy.
First, a cpu price difference is never proportional to actual performance difference even in the same line of cpus. But you probably know it even better then me.
Second, your performance/price equation is partially useful for home PC user but in busines/server space it is completely BS. Even if X5570 only 20% faster then 2435 do you really think that %20 more work which potentially can be done in, lets say, two years, costs only $250 diff between cpus?

No need to get angry or combative. I think it's cool that JF posts in this forum. It's always nice to get someone from inside AMD to post. I wish more executives would do that. As far as the price/performance ratio, obviously you have to take into account price vs performance. If performance is more important, then obviously Nehalem will be a better option. But if it isn't, Opteron presents a very compelling argument. Not every server runs at 100% capacity all the time. Most servers are at like 30% most of the time. If that were the case, Nehalem doesn't present much value beyond Opteron.

[Macadamia]

do you really think that %20 more work which potentially can be done in, lets say, two years, costs only $250 diff between cpus?

Do you really think that servers are even loaded 100% all the time?

Funny, mainstream users don't need good graphics and now servers process data all the time without ever being in a state of idle. Any more shareholder quips?


BTW my take on why Nehalem is selling well: It's a proper open-ended platform with at least a guaranteed upgrade path. G34 and C32 are going to provide that too but as of now AMD's Socket F is a dead end platform and many businesses might not have proper motherboards (split power plane supports) etc to justify an Istanbul upgrade.

[JF-AMD]

I am little surprised to hear such things from a marketing guy.
First, a cpu price difference is never proportional to actual performance difference even in the same line of cpus. But you probably know it even better then me.
Second, your performance/price equation is partially useful for home PC user but in busines/server space it is completely BS. Even if X5570 only 20% faster then 2435 do you really think that %20 more work which potentially can be done in, lets say, two years, costs only $250 diff between cpus?

Here's the reality of the situation: arguing about the performance of top bin processors is really meaningless. 95% of the server processors sold are not top bin.

As a marketing person, this tells me that raw performance (i.e. the highest test score) is not that relevant. What is, is either price/performance or performance/watt.

If you look at Mercury Reaseach (the leading company that does processor-level market research), you see that most server processors are sold in the N-3 or N-4 categories.

We do exceptionally well there because as you move down our stack, you still get top speed memory and top speed I/O. Intel artificially limits those products (I won't even try to guess as to why.)

If more people saw a value in getting proportionally more work done over 2-3 years, then you are absolutely right. However, all of the trends in the server market tell me that only ~5% of the workloads merit the top performance and the other 95% is looking for more balance. And that is what I believe we bring to the market.

[JF-AMD]

What about overall system prices? or do most people buy only upgrade CPUs?

Actually at the system level, Intel's price premium goes up. This is from the work I did back in december (posted on anandtech forum), prices may have changed slightly, but should be close these levels.

I actually went back to do a little math on this.

I went to Dell's site and configured 3 servers (PowerEdge 710 and PowerEdge 805). I held all components, warranty and other options equal.

The 3 servers were:
One with 2 Xeon 5570's and 36GB of RAM (3 channels) - $7408
One with 2 Xeon 5570's and 32GB of RAM (only 2 channels, but who would do that?) - $6978
One with 2 Opteron 2435's - $4565

The more expensive one is 62% more expensive, the lower performing dual channel box 2as 52% more expensive. So both are more expensive than the 40% delta on processor pricing.

For fun I configured an E5540 and that was still 20% more expensive. So I started thinking, where do you get to price parity? Basically that is an E5520. So, from a price perspective, you can get 8 2.26GHz Intel cores for roughly the same cost as AMD's most expensive 12 cores @ 2.6GHz.

Knowing that my most popular SKU is the 2431 (2.4GHz), to get to the same price for an Intel system, you're at the bottom, a 1.86GHz DUAL CORE, so 4 total cores of Nehalem vs. 12 2.4GHz Opteron cores. Oh, and you are still paying a price premium for those 4 cores (~4%).

So, yes, you can point to all types of benchmarks to show you how the top speed xeon parts are fast, but when it comes to pure economics, your choice is a 50-60% premium on the top bin vs our 2435 (to match those benchark charts), or you can try to match the price our our most popular dual socket six core platform, but in trying to match budget dollars, you are going to end up with a pair of 1.86 dual core processors. No turbo, no HT, 4.8GT QPI and only 800MHz memory.

If you check Gartner and IDC you'll see that the average ASP for 2 socket servers is ~$3600. That tells me that there are far more servers being bought in that price band than any others. And at $3600, look at what you get with Intel and what you get with AMD. It's all about value in the data center. Benchmarks are nice, but that is not how the overwhelming majority of servers are bought. I've been in this business for close to 20 years and I have never seen more of a focus on value. We're in exactly the right spot.

[kl0012]

No need to get angry or combative. I think it's cool that JF posts in this forum. It's always nice to get someone from inside AMD to post. I wish more executives would do that. As far as the price/performance ratio, obviously you have to take into account price vs performance. If performance is more important, then obviously Nehalem will be a better option. But if it isn't, Opteron presents a very compelling argument. Not every server runs at 100% capacity all the time. Most servers are at like 30% most of the time. If that were the case, Nehalem doesn't present much value beyond Opteron.

I'm sorry if it looks like I am angry. No, I'm not angry at all.

Do you really think that servers are even loaded 100% all the time?

Did I say this? potentialy - please pay attention.

Funny, mainstream users don't need good graphics and now servers process data all the time without ever being in a state of idle. Any more shareholder quips?

The first part of your sentence please address someone else, I never said this. For the second part I can say that if you really don't need performance, then again Nehalem servers provides much higher performance/watt value especially when the server spends more time in idle. Until AMD will bring power-gated technology into its processors there is no competition here.