Sign in with
Sign up | Sign in
Your question

AMD K10 Cinebench benchmark - Computex

Last response: in CPUs
Share
June 6, 2007 10:25:16 AM

This is not much ... but better than nothing I guess ...

Link

Anyway it's the first time a K10 was benchmarked by an independent reviewer.
June 6, 2007 10:50:10 AM

One benchmark is not enough to show a true reflection, but your right it's better than nothing. The AMD was slower, but before people start bashing it remember the AMD was at '1.6Ghz' and the Intel at '2.4Ghz'. Clock the AMD higher and it will leave Intel for dust in this benchmark I believe 8) I have always liked AMD processors and I can't wait to buy one :)  .
June 6, 2007 10:56:24 AM

Except the Intel processor was 57% faster in performance while only being clocked 50% higher. In fact, if you do a rough estimate, the dual-core version of the K10 would be slower than either the E4300 or the 3600+ in this particular test.
Related resources
June 6, 2007 10:59:08 AM

Quote:
Except the Intel processor was 57% faster in performance while only being clocked 50% higher. In fact, if you do a rough estimate, the dual-core version of the K10 would be slower than either the E4300 or the 3600+ in this particular test.


8O 8O 8O
June 6, 2007 11:01:24 AM

Quote:
Except the Intel processor was 57% faster in performance while only being clocked 50% higher. In fact, if you do a rough estimate, the dual-core version of the K10 would be slower than either the E4300 or the 3600+ in this particular test.


8O 8O 8O
Mind you, the dual-core K10 would be at 1.6GHz. Still it doesn't appear that the K10 has improved on the K8 at all in regards to Cinebench 9.5, which is a big suprise given the improved SSE.
June 6, 2007 11:02:39 AM

3rd party benchmarks needed. AMD is hiding something. They pick the benchmark and use a low clocked version? Boo and hiss! Give the procs to all the review sites so we can get some real insight.
June 6, 2007 11:09:22 AM

Quote:
3rd party benchmarks needed. AMD is hiding something. They pick the benchmark and use a low clocked version? Boo and hiss! Give the procs to all the review sites so we can get some real insight.


This is as close to a third party benchmark as you are going to get before the official launch. Dailytech is a highly credible site. They scooped the R600 fiasco early with benchmarks that all the ATI fanbois screeched about like stuck pigs. Only to eat crow a week later when it was shown that the DT benchmarks were 100% accurate.

I for one dont doubt the accuracy of these results one bit even if it is one early silicon.

Anyway you are missing the main point of these results: This tells me K10 will, at best, barely compete with the Penryn class cpus. While you can squeeze better clock speeds out by tweaking the silicon. You CANNOT significantly change the performance per clock without significantly altering the actual chip design. AMD barely has time to tweak the silicon for better clock speeds they cannot change the design.

AMDs hopes of having a Conroe/Penryn killer in K10 are dashed.

Not good for CPU consumers!
June 6, 2007 11:14:26 AM

Ouch!

Even assuming perfect scaling, it falls behind the 2.4 GHz Clovertown.

No wonder AMD is staying quiet. They can't ramp the clock and IPC doesn't beat the available products from the competition.
June 6, 2007 11:26:33 AM

Quote:
Except the Intel processor was 57% faster in performance while only being clocked 50% higher. In fact, if you do a rough estimate, the dual-core version of the K10 would be slower than either the E4300 or the 3600+ in this particular test.

I wouldn't say that too soon.

The following is the Cinebench from Q6600, QX6700, and QX6700 OC.
humm...

If my rough calculation is correct, and assuming performance progression is linear, a 2.6Ghz Barcelona will supposedly get 11 seconds. That's about 1~2 seconds faster than QX6700 @ 3.2 Ghz.

So yeh, I would agree that Barcelona will be competitive with Penryn on this particular test. However, more data is needed to determine the real performance of K10.
June 6, 2007 11:29:50 AM

What's even more shocking is that rendering is a highly FP intensive task. This is exactly the type of calculation AMD said Barcelona would "wipe the floor" with Intel 8O
June 6, 2007 11:30:14 AM

Quote:

If my rough calculation is correct, and assuming performance progression is linear, a 2.6Ghz Barcelona will supposedly get 11 seconds. That's about 1~2 seconds faster than QX6700 @ 3.2 Ghz.

The 1.6GHz K10 requires 27s to complete the test, it was the 2.4GHz Kentsfied that took 18s.
June 6, 2007 11:33:22 AM

Quote:

If my rough calculation is correct, and assuming performance progression is linear, a 2.6Ghz Barcelona will supposedly get 11 seconds. That's about 1~2 seconds faster than QX6700 @ 3.2 Ghz.

The 1.6GHz K10 requires 27s to complete the test, it was the 2.4GHz Kentsfied that took 18s.
Yep.
Since 2.6Ghz is 62.5% faster than 1.6Ghz, so if we assume performance progression is linear, we should also see a 62.5% increase in benchmark score.

So:
27 secs * (1-62.5%) = 27 secs * (37.5%) = 10.125 secs.

However, its four in the morning here, and I haven't slept yet. I could've got this third grade math wrong though... :oops: 
June 6, 2007 11:34:52 AM

Quote:

AMDs hopes of having a Conroe/Penryn killer in K10 are dashed.
Not good for CPU consumers!


Two points:

First, these results are far from spectacular because K8 was already on par with C2D performance in this particular benchmark Link, so the architectural improvements of K10, are not translating into real performance gains, at least not in this benchmark

Second, we have to recon that the tested chip was still early silicon and performance can still improve but not by much ... at best 10% or so.
I would not expect miracles though.

It’s still unclear how K10 will perform at the tasks where AMD currently lags behind C2D(video encoding and SSE2 optimized programs in general and programs sensitive to INT performance).

What we’ve seen by now is two benchmarks were K10 is ~ on par with K8 performance.

We can draw 2 conclusions from that:

1)Either AMD is sandbagging and deliberately hiding their real perfomance (dubious, because at this point in the game there’s really nothing to be gained from that)

2)Either K10 is really not that much faster than K8 in these particular tasks and we need other programs to fully unleash the power of K10 architecture(remember that K10 was designed primarily as a server architecture and will mostly benefit in high memory intensive applications and multi-threaded environments, it will not necesarily kick ass in absolute desktop application performance)

So before starting to predict the demise of K10 architecture, we should at least expect for a full set of benchmarks on a final revision of the core …
June 6, 2007 11:36:57 AM

The correct method would be to divide 27s by (1+62.5%), which would give you about 16.6s. Using your method, plugging in a 3.2GHz value would result in a time of 0s.
June 6, 2007 11:39:31 AM

Quote:
The correct method would be to divide 27s by (1+62.5%), which would give you about 16.6s. Using your method, plugging in a 3.2GHz value would result in a time of 0s.

You're right about my calculation error. Thank you for pointing that out. Maybe I'll take a nap, and come back and solve this :lol: 

EDIT: I just rethought about the mathematical model you proposed, and they're right.

Using solely this piece of information, the same conclusion can be drawn with POV-Ray performance : AMD's floating points might only get a minor boost, that is not enough to topple Intel's position. This might also be the reason that AMD is being hush hush about this. However, more data is needed, as this piece of information only represents a small piece in the whole puzzle.
June 6, 2007 11:54:01 AM

Quote:

Two points:

First, these results are far from spectacular because K8 was already on par with C2D performance in this particular benchmark Link, so the architectural improvements of K10, are not translating into read performance gains, at least not in this benchmark


And this explains the POVray beating the K10 recently took how? POVray does show significantly better clock for clock performance for Core over K8: http://www.hwupgrade.com/articles/cpu/27/the-last-of-th...
Further, we can now much more accurately estimate the clockspeed that the recent K10 POVray benches were run at.

Lets not forget AMD is claiming +40% on the Xeon E3255 (presumably another simulation).

Quote:

Second, we have to recon that the tested chip was still early silicon and performance can still improve but not by much ... at best 10% or so.
I would not expect miracles though.


You arnt going to get another 10% performance per clock 1-2 months from launch, ESPECIALLY when you are still trying to fix the serious clockspeed disadvantage. Hell, Penryn is a 12 month refresh with slightly modified silicon and it only gets +10% per clock over Core.
June 6, 2007 11:58:14 AM

Quote:
The correct method would be to divide 27s by (1+62.5%), which would give you about 16.6s. Using your method, plugging in a 3.2GHz value would result in a time of 0s.


To be more exact, it would result in exactly 16,875s for a 2,6 GHz K10
June 6, 2007 12:12:56 PM

Quote:

And this explains the POVray beating the K10 recently took how? POVray does show significantly better clock for clock performance for Core over K8: http://www.hwupgrade.com/articles/cpu/27/the-last-of-th...
Further, we can now much more accurately estimate the clockspeed that the recent K10 POVray benches were run at.

Lets not forget AMD is claiming +40% on the Xeon E3255 (presumably another simulation).


No offense, but the link you have provided its a little bizarre, hwupgrade??? What about in the link I’ve provided, witch is from TechReport by the way , on the bottom of the page it clearly shows a list of AMD CPU's vs. Intel CPU's, benchmarked din POV-ray and the results speak for themselves…

Anyway I really don’t want to get in a polemic on this topic because we can’t verify the results ourselves.


Quote:

You arnt going to get another 10% performance per clock 1-2 months from launch, ESPECIALLY when you are still trying to fix the serious clockspeed disadvantage. Hell, Penryn is a 12 month refresh with slightly modified silicon and it only gets +10% per clock over Core.


I tend to agree with you on this one … that’s exactly why I wrote: "at best 10% or so.
I would not expect miracles though
"
a b à CPUs
June 6, 2007 12:39:37 PM

Dailytech (Anandtech) is as credible (pffft!!) as this site (under Toms Hardware) - both are Intel sponsored sites ... both sidestep the NDA's with manufacturers accordingly.

Both have been actively creating FUD for AMD.

The single socket comparisons are useless from a server perspective anyway. http://www.overclockers.com/tips01162/

Opteron still rules the roost in that high end area.

Has Cray approached Intel ?? Yeah ... right ... I forgot ... you ppl here are noobs and teenyboppers who believe what they read from forums ...

Mebbe some of you should have stayed at school a bit longer.

The maths debate was interesting ... I can see none of you work in accounting either ... thankfully.
June 6, 2007 12:55:56 PM

Quote:

No offense, but the link you have provided its a little bizarre, hwupgrade???


http://www.anandtech.com/cpuchipsets/intel/showdoc.aspx...
http://www.extremetech.com/article2/0,1697,2112086,00.a...

In the Anandtech test they are very close ... so I guess we can say that K8 is at least competitive with C2D

X6800(2,93 Ghz) - 1134,9
E6700(2,66 Ghz) - 1041,5
X2 6000+(3,00 Ghz) - 1022,3

In the second test there is a big difference, I think we both agree that both Anandtech and TechReport are credible sources an if we make a arithmetic average between the two of them, we’ll get to the conclusion that K8 is at least competitive with C2D.

Anyway, it’s a known fact that K8 is competitive with C2D on FPU intensive application so I don't thinks we need to continue this anymore...
June 6, 2007 1:01:58 PM

Quote:
Dailytech (Anandtech) is as credible (pffft!!) as this site (under Toms Hardware) - both are Intel sponsored sites ... both sidestep the NDA's with manufacturers accordingly.

Both have been actively creating FUD for AMD.

The single socket comparisons are useless from a server perspective anyway. http://www.overclockers.com/tips01162/

Opteron still rules the roost in that high end area.

Has Cray approached Intel ?? Yeah ... right ... I forgot ... you ppl here are noobs and teenyboppers who believe what they read from forums ...

Mebbe some of you should have stayed at school a bit longer.

The maths debate was interesting ... I can see none of you work in accounting either ... thankfully.


What's the matter, did you buy AMD shares at $40 or something? Why all the anger? :lol:  :wink:
June 6, 2007 1:03:18 PM

Quote:
Dailytech (Anandtech) is as credible (pffft!!) as this site (under Toms Hardware) - both are Intel sponsored sites ... both sidestep the NDA's with manufacturers accordingly.

Both have been actively creating FUD for AMD.

The single socket comparisons are useless from a server perspective anyway. http://www.overclockers.com/tips01162/

Opteron still rules the roost in that high end area.

Has Cray approached Intel ?? Yeah ... right ... I forgot ... you ppl here are noobs and teenyboppers who believe what they read from forums ...

Mebbe some of you should have stayed at school a bit longer.

The maths debate was interesting ... I can see none of you work in accounting either ... thankfully.


Hey dude just chill, what's with all these insults???
You think you're better or what???
You’re speaking about people you don’t even know!!!
I would suggest you re-think you're attitude towards the people and this forum ... no need to be rude when trying to say something!!!

You can state you're opinion in a mannerly way or you can just take you're insults elsewhere...

Get It????
June 6, 2007 1:11:37 PM

Quote:

AMDs hopes of having a Conroe/Penryn killer in K10 are dashed.
Not good for CPU consumers!


Two points:

First, these results are far from spectacular because K8 was already on par with C2D performance in this particular benchmark Link, so the architectural improvements of K10, are not translating into read performance gains, at least not in this benchmark

Second, we have to recon that the tested chip was still early silicon and performance can still improve but not by much ... at best 10% or so.
I would not expect miracles though.

It’s still unclear how K10 will perform at the tasks where AMD currently lags behind C2D(video encoding and SSE2 optimized programs in general and programs sensitive to INT performance).

What we’ve seen by now is two benchmarks were K10 is ~ on par with K8 performance.

We can draw 2 conclusions from that:

1)Either AMD is sandbagging and deliberately hiding their real perfomance (dubious, because at this point in the game there’s really nothing to be gained from that)

2)Either K10 is really not that much faster than K8 in these particular tasks and we need other programs to fully unleash the power of K10 architecture(remember that K10 was designed primarily as a server architecture and will mostly benefit in high memory intensive applications and multi-threaded environments, it will not necesarily kick ass in absolute desktop application performance)

So before starting to predict the demise of K10 architecture, we should at least expect for a full set of benchmarks on a final revision of the core …In AMD's defense, they were using DDR2-667. I think we're all well aware of the importance of running DDR2-800 on AM2 systems. What we don't know is if it will be as important, with the L3 cache, and other changes on K10. Remember the last couple months when AMD was finalizing AM2, and Anand did several reports on it's terrible memory performance with DDR2.... after AMD tweaking the IMC, and Anand retesting.... (3 tries i think) :wink: they improved a fair amount. So, any thing's possible, i guess. I think it's entirely possible that they might find another 10% improvement in some apps, but not quaranteed. :) 
June 6, 2007 1:13:04 PM

Quote:

Anyway, it’s a known fact that K8 is competitive with C2D on FPU intensive application so I don't thinks we need to continue this anymore...


+10% per clock FP performance is not an insignificant advantage given that overall Core is only about +20% on K8 clock for clock.
June 6, 2007 1:14:51 PM

Quote:
You CANNOT significantly change the performance per clock without significantly altering the actual chip design.


YES YOU CAN - just disable the L3 CACHE!!!
June 6, 2007 1:16:57 PM

Quote:

In AMD's defense, they were using DDR2-667. I think we're all well aware of the importance of running DDR2-800 on AM2 systems.


Rendering isnt a bw intensive app.
June 6, 2007 1:19:48 PM

Quote:


YES YOU CAN - just disable the L3 CACHE!!!


June 6, 2007 1:21:22 PM

Quote:

In AMD's defense, they were using DDR2-667. I think we're all well aware of the importance of running DDR2-800 on AM2 systems.


Rendering isnt a bw intensive app.

Someone on Dailytech mentioned that Barcelona only accepts registered DIMMs up to 667MHz? Any truth to this statement?
a b à CPUs
June 6, 2007 1:39:29 PM

Gee ... there is no way I could be as rude as Baron Matrix anyway ... he takes the cake, eats it and leaves the crumbs all over the opposition.

I see no-one has replied regarding the server arguement ... good.

In terms of a single socket I doubt the K10 will actually beat the high end Core2 gaming boxes as even with the tuned prefetchers and other enhancements as the Core2 has an advantage from an arch perspective running single intensive threads that are not FPU reliant.

AMD's gamble is on intra and inter die communications and split power planes ... all which is great for servers ... perhaps not so great until the gaming coders catch up with making the most out of multi-threaded games.

Here is a very good article that I am sure most would see relevant.

http://www.realworldtech.com/page.cfm?ArticleID=RWT0516...
June 6, 2007 1:45:02 PM

Quote:

Someone on Dailytech mentioned that Barcelona only accepts registered DIMMs up to 667MHz? Any truth to this statement?


I remember hearing something like 1066MHz will be supported at launch.
June 6, 2007 1:47:41 PM

Do you guys fail to realize something... even IF Barcy is -10% to +10% better than Penryn (take your pick) - the chip is going to cost $$$$$ to produce. 8O

Wherease the Penryn will actually lower Intel's costs.

Price war... Intel wins... not even close.
June 6, 2007 1:53:27 PM

With this new data point, I can't see AMD escaping chapter 11. They simply won't be able to charge a premium over the intel offerings like they could during the K8 vs P4 era. The writing is on the wall.
June 6, 2007 1:55:45 PM

Quote:
Do you guys fail to realize something... even IF Barcy is -10% to +10% better than Penryn (take your pick) - the chip is going to cost $$$$$ to produce. 8O

Wherease the Penryn will actually lower Intel's costs.

Price war... Intel wins... not even close.


That's hardly relevant to this topic though. We are purely discussing the performance of Barcelona at this point. Obviously Intel will have an advantage in manufacturing cost as it is 45nm vs 65nm, hardly a fair fight.

Of course if Barcelona/K10 doesn't end up competitive with Penryn that will only make matters worse... if that is the case it may very well spell the end of AMD I'm afraid, and I've been refraining from making such statements for a VERY long time... but these results, plus AMD's recent track record (R600 anyone?) don't exactly fill me with a lot of confidence...

Of course I expect Baron to act all cheery and claim everything is fine with AMD, what do I know... :? :lol: 
June 6, 2007 1:58:28 PM

Quote:
With this new data point, I can't see AMD escaping chapter 11. They simply won't be able to charge a premium over the intel offerings like they could during the K8 vs P4 era. The writing is on the wall.


Yup. Glad I dont own AMD stock. It is quite clear that AMD wont be able to repeat its K8 performance leap.
June 6, 2007 1:58:52 PM

Quote:
Gee ... there is no way I could be as rude as Baron Matrix anyway ... he takes the cake, eats it and leaves the crumbs all over the opposition.


Don't be rude, period. You’re excuses or explications are worthless, apologies would have been much better.

Quote:

In terms of a single socket I doubt the K10 will actually beat the high end Core2 gaming boxes as even with the tuned prefetchers and other enhancements as the Core2 has an advantage from an arch perspective running single intensive threads that are not FPU reliant.

AMD's gamble is on intra and inter die communications and split power planes ... all which is great for servers ... perhaps not so great until the gaming coders catch up with making the most out of multi-threaded games.


I agree with you on this one ... as I already pointed out in one of my previous post here:"remember that K10 was designed primarily as a server architecture and will mostly benefit in high memory intensive applications and multi-threaded environments, it will not necesarily kick ass in absolute desktop application performance"
June 6, 2007 2:01:40 PM

Do you guys fail to realize something... even IF Barcy is -10% to +10% better than Penryn (take your pick) - the chip is going to cost $$$$$ to produce. 8O

Wherease the Penryn will actually lower Intel's costs.

Price war... Intel wins... not even close.
June 6, 2007 2:09:47 PM

Jesus, what’s with all this fatalism????

One benchmark appears and you’re all making a tragedy out of it …
I’ve already said, we need to wait for a proper testing with a full benchmark suite and an up to date revision of the core…

To all of you who think that AMD will not survive if K10 will not be super-competitive… YOU'RE WRONG!!! … AMD survived on K5, K6 days when they were not even close on the current level of performance and market share so just relax and wait for the final benchmarks.

:lol:  :lol:  :lol: 
June 6, 2007 2:19:11 PM

Quote:
This is not much ... but better than nothing I guess ...

Link

Anyway it's the first time a K10 was benchmarked by an independent reviewer.


Its not doubt AMD is playing the R600 game with the Barcelona, sad really. They are hiding non stellar performance.

Anandtech has already commented on how the Tiawanese community of vendors at Computex are unimpressed with the Barcelona. They should know since they are the ones developing products for the chip.
June 6, 2007 2:34:31 PM

Quote:
This is not much ... but better than nothing I guess ...

Link

Anyway it's the first time a K10 was benchmarked by an independent reviewer.


Its not doubt AMD is playing the R600 game with the Barcelona, sad really. They are hiding non stellar performance.

Anandtech has already commented on how the Tiawanese community of vendors at Computex are unimpressed with the Barcelona. They should know since they are the ones developing products for the chip.

I don’t want to appear as I’m defending AMD or something … I usually chose the better value for my money and I’m don’t have a preference for INTEL or AMD (used them both in the past), but I really think that we can’t draw this kind of conclusions yet… and by the way I happen to think that R600 is a good product but maybe not polished enough. :p  :p  :p 
June 6, 2007 2:34:48 PM

Quote:


YES YOU CAN - just disable the L3 CACHE!!!




Question... is the bike "owning" the guy, or the guy "owning" the bike????

:?
June 6, 2007 3:50:49 PM

Quote:
To all of you who think that AMD will not survive if K10 will not be super-competitive… YOU'RE WRONG!!! … AMD survived on K5, K6 days when they were not even close on the current level of performance and market share so just relax and wait for the final benchmarks.

:lol:  :lol:  :lol: 


AMD never had the huge debt load they have today, debt they have no ability to service with no profits in sight. The ATI acquisition gutted their balance sheet. They always had lots of room to manuver in the Jerry Sanders days, now they don't, and failure is no longer an option.

If K10 is not a big success, and that is looking increasingly doubtful, then AMD will not be able to survive in its current form. Either they will go under with a C11 and lose their X86 licence, or they will have to sell off their fabs, which will destroy their ability to be a serious competitor to Intel.
June 6, 2007 4:08:51 PM

Barcy is a server CPU.
The a dual core 4 socket Opteron, beats the 4 core, dual socket Xeon in most business based benchmarks (SAP being one of them) see

http://www50.sap.com/benchmarkdata/sd2tier.asp.

Intel does not have a 4 socket solution yet for core 2 based Xeons.
Sun already has an 8 socket solution, Currently opteron does not scale as well up to 8 cores but with the extra HT link this should not be a problem.
So with a 4 core, 4 to 8 socket market, AMD has the only competive solution. Not even itanium comes close. The only other contender is IBMs P6, but that won't run Windows.
This is a huge market segement, with virtualisation and server consolidation being the key words in most medium to large organizations. That is thousands if not millions of servers worldwide.
Most companies tend to have a 3 to 5 year write off period for most kit, the early adopters of Opteron will probably be looking at refreshing kit and consolidating.
Most companies do not upgrade CPUs, rather they will put in a new server and redeploy older kit to less applications.
June 6, 2007 4:12:42 PM

Quote:

This is a huge market segement, with virtualisation and server consolidation being the key words in most medium to large organizations.


Nonsense. More than two socket servers is a small segment in the overall market.
June 6, 2007 4:19:04 PM

Quote:
You CANNOT significantly change the performance per clock without significantly altering the actual chip design.


YES YOU CAN - just disable the L3 CACHE!!!

Wow, you missed all of the drama. I am still confident that AMDs projected numbers will be hit. These chips - if they're the same ones that Fuad had - were B1 stepping according to CPUz.

Looking at the time it takes to get a chip into a socket, I would say that if they are using non-packaged in some type of "test socket," they can get bugs worked out quicker.

I just hope they come out soon. I really want to see final silicon running 8P. I would also like to see if they solved the odd multiplier issue that AM2 has. I mean I woulod hate to have my spanking new Phenom X4 running my low latency 1066 RAM at 988 or something.

As Ninja said last year when Anand did the first AM2 tests, they were getting killed by 939 by a lot and that was about 2 months before the launch. When all was said and done low latency DDR2-800 gave 20% over 939.

Barcelona is alot more complex than AM2 and the cash flow problem certainly isn't helping things. In the end, August I think will be a good month for the industry.

I am curious though why they didn't use 790G for this test since it is AM2+. I know I saw some mobos. That may make a difference. But we'll see.


ALL HAIL THE DUOPOLY!!!!
June 6, 2007 4:28:38 PM

Quote:

This is a huge market segement, with virtualisation and server consolidation being the key words in most medium to large organizations.


Nonsense. More than two socket servers is a small segment in the overall market.

Small? Its huge.
Every major retailer, bank, oil company, call centre, manufacturing based company or utlitiy compmany have vast amounts of 4+ socket sytems and even more 2 socket based systems and are looking to consolidation and virtualisation.
Anybody with a major database with millions of customers run on large backend databases just for ERP based processing, then you have CRM based systems, large web bases systems (at the backend is a usually a bloody large database).
I am currently doing work for a single medium sized frozen goods company. Just for the development environment (ERP system, Datawarehouse, Portal, Supply chain and a data broker) we are using a complete IBM p570 machine with 16 dual core processors. We then need to build, Test, Training, User acceptance, Production and Disaster recovery systems. With the production and Disaster recovery system being at least 3 times the size. And that is only a small propotion of the total number of systems the customer has.
The likes of BP or Exon probably have 1000s of large databases running worldwide.
June 6, 2007 4:37:34 PM

Do you have numbers to prove otherwise? Last number I remember is 2 socket comprises about 90% of server market.
June 6, 2007 4:41:09 PM

Quote:


Small? Its huge.


:lol:  ... google up some server market segmentation reports and learn something.

I guess AMD had to borrow $1 billion for day to day cash flow because someone hacked their bank accounts :roll:
June 6, 2007 4:41:27 PM

Quote:
Dailytech (Anandtech) is as credible (pffft!!) as this site (under Toms Hardware) - both are Intel sponsored sites ... both sidestep the NDA's with manufacturers accordingly.

Both have been actively creating FUD for AMD.

The single socket comparisons are useless from a server perspective anyway. http://www.overclockers.com/tips01162/

Opteron still rules the roost in that high end area.

Has Cray approached Intel ?? Yeah ... right ... I forgot ... you ppl here are noobs and teenyboppers who believe what they read from forums ...

Mebbe some of you should have stayed at school a bit longer.

The maths debate was interesting ... I can see none of you work in accounting either ... thankfully.

Baron... you should be glad...
You finally have a successor to your legacy :lol:  :lol:  :lol: 
June 6, 2007 4:48:31 PM

Thats the thing though. People want to get rid of lots of two socket servers.
They take up valuble real estate, take more people to manage, more cabling, etc.
I don't have figures. But I do work for a major IT consultancy and about 50% of our work at the moment is about consolidation and virtualisation up from about 10% 3 years ago.
We have one utilities company currently running 200 odd 2 Socket 1U rack machines. Most of them arn't even heavily utilised (any architect who wants to keep his job will always specify a minimum of 2 sockets for a production environment for redundency).
They are migrating to two 64 way HP superdomes. At the same time getting rid of about 20 staff.

Be it intel/AMD/IBM/SUN they would all rather sell 10 High end and expensive server CPUs with nice expensive warrenties than 100 single socket low cost low profit margins. This is what I think AMD is aiming for and the architecture is perfect for that market.
!