Intel Xeon 5570: Smashing SAP records.

[Donnie27]

Donnie27 I don't agree with your history lesson.

On servers i7 need to get the power down

No, it's NOT my history lesson at all. It's more like you not knowing computer history. If you think it is merely me fabricating it, prove it? You can't do that of course.

http://www.intel.com/products/processor/pentium4HTXE/
http://www.intel.com/products/proces...v+list_itanium

Intel used L3 on its 130nm processors long before AMD even thought about using a L3.

The Alpha architecture was sold, along with most parts of DEC, to Compaq in 1998. Compaq, already an Intel customer, decided to phase out Alpha in favor of the forthcoming Intel IA-64 "Itanium" architecture, and sold all Alpha intellectual property to Intel in 2001, effectively "killing" the product. Hewlett-Packard purchased Compaq later that same year, continuing development of the existing product line until 2004, and promising to continue selling Alpha-based systems, largely to the existing customer base, until October 2006 (later extended to April 2007).[2]

As part of the purchase agreement Intel had to promise to not cut AMD off from the EV6 license Dirk got from DEC/Compaq.

http://www.alasir.com/articles/alpha...pha_21264.html

There were 2 system logic sets designed initially for the 21264 processors: DEC Tsunami (21272; also known as Typhoon) and AMD Irongate (AMD-751), though could be many more if to take into account that both 21264 and Athlon utilised almost the same system bus licenced by DEC to AMD.

http://www.eetimes.com/news/semi/sho...cleID=10817401

So yes, disagree with real history all you like. Remain blind if it makes you feel good. I'll at least back up my beliefs with facts instead of misguided feelings for some large corporation like AMD or Intel for that matter.

[Donnie27]

Yup, I understood. Just saying they are a sweet little system even though on a crippled platform.

Cool! Yes they were! They should have given some here a foreboding feeling LOL!

[NaeKuh]

bah.... gainestown..

oh... headache's coming up again...



dave knows why..

[Movieman]

bah.... gainestown..

oh... headache's coming up again...



dave knows why..

Maybe not..Got time for a call?

[Donnie27]

Maybe not..Got time for a call?

Some usual suspects are conspicuously missing from this thread, what's up that?

[Zucker2k]

Some usual suspects are conspicuously missing from this thread, what's up that?

You noticed too? Some don't like figures from legit sources. Now, if only I could get my filthy paws on one of those chips.

[Drwho?]

Do those Xeos have the turbo feature? How many multis are possible if? Sorry I'll be impressied if they show those results in the spec_power benchmark. I'm sure those results are tuned for marketing reasons as they always are.
Once those chips are available amd can close the gap with six core instanbul chips.

The results of spec are not tune for "marketing reason", but for "performance"

The rules of specs are very clear, and every industry player can join. Intel has one vote out of many many, the source code for SPEC goes through a process where SUN, AMD, IBM are all represented, plus many other companies. AMD has been recommanding SPEC FP RATE for more than 3 years, will it become a bad benchmark? If so, why? Same for the SAP benchmark, it was one of their favorite bench 1 before Conroe showed up ...

Nobody can say that SPEC is not fair, just come and join and see, it is running very democratically, and very fairly. If you have any proove of unfairness about spec, please let me know, I got to feed back to the SPEC Gurus at intel.



SPEC is a good benchmark, I don t like SPEC_FP_RATE, because I don t think a machine run many of the same task in parrallel, starting at the same time, and if it does , it is very often in a Innerthreading way.

This is my personal opinion, my employer is not responsible for my post on xtremeSys

[Donnie27]

You noticed too? Some don't like figures from legit sources. Now, if only I could get my filthy paws on one of those chips.

Yea LOL, a thread they thought I would jump in was close in twenty posts because same guy/s went loony I stayed out of it because I already knew what was about to happen.

I'll not call him out by name but I gave him 20 links to info he requested. 5 of the links were AMD leaning sites. Yet the very next thread, he quickly forgotten them all and acted as if I'd never posted them. Dewd, even Alzheimer's patients have longer short term memory than that.

It is also funny how anyone not posting Positive PR becomes Anti whatever

[savantu]

After Intel resolve some of the C0 stepping erratas and the cache coherency issue.

What issues ? Please expand.

[Movieman]

The results of spec are not tune for "marketing reason", but for "performance"

The rules of specs are very clear, and every industry player can join. Intel has one vote out of many many, the source code for SPEC goes through a process where SUN, AMD, IBM are all represented, plus many other companies. AMD has been recommanding SPEC FP RATE for more than 3 years, will it become a bad benchmark? If so, why? Same for the SAP benchmark, it was one of their favorite bench 1 before Conroe showed up ...

Nobody can say that SPEC is not fair, just come and join and see, it is running very democratically, and very fairly. If you have any proove of unfairness about spec, please let me know, I got to feed back to the SPEC Gurus at intel.



SPEC is a good benchmark, I don t like SPEC_FP_RATE, because I don t think a machine run many of the same task in parrallel, starting at the same time, and if it does , it is very often in a Innerthreading way.

This is my personal opinion, my employer is not responsible for my post on xtremeSys

Francois:
Sometimes here the talk goes round and round and misses the point.
Give me a dual socket Nehalem board with a bios that will let me work it like your single socket nehalem board and thats all I need.
unlocked multi cpu's, adjustments in bios for memory and voltages.
The rest is small stuff.
With that the cpu's will prove themselves just as they did in the single socket board.

[gosh]

No, it's NOT my history lesson at all. It's more like you not knowing computer history. If you think it is merely me fabricating it, prove it? You can't do that of course.

You missunderastand what I was writing, I am not that good at english and didn't want get involved in totally different discussion.

http://www.xtremesystems.org/forums/...3&postcount=65

[duploxxx]

I can guarantee you Xeon 5500 series servers uses LESS power than 5400 series.
And when you can cool 4 and 8 barcelona chips you can cool anything

you can't stop posting crap can't you, perhaps check what tigertown and dunnington consumes and the performance lack they have against 45nm shanghai and that power consumption before you make such a post, after all those are all 45nm parts out in the field all the rest is vaporware for another 3-4 months and 8-9 months.

The results of spec are not tune for "marketing reason", but for "performance"

The rules of specs are very clear, and every industry player can join. Intel has one vote out of many many, the source code for SPEC goes through a process where SUN, AMD, IBM are all represented, plus many other companies. AMD has been recommanding SPEC FP RATE for more than 3 years, will it become a bad benchmark? If so, why? Same for the SAP benchmark, it was one of their favorite bench 1 before Conroe showed up ...

Nobody can say that SPEC is not fair, just come and join and see, it is running very democratically, and very fairly. If you have any proove of unfairness about spec, please let me know, I got to feed back to the SPEC Gurus at intel.



SPEC is a good benchmark, I don t like SPEC_FP_RATE, because I don t think a machine run many of the same task in parrallel, starting at the same time, and if it does , it is very often in a Innerthreading way.

This is my personal opinion, my employer is not responsible for my post on xtremeSys

right and wrong.

from the early benchmarks it is clear that in SPEC benchmarks the HT functionality is often able to cope with the synthetic load providing higher scores while in real world server loads it was better to shut it down.

However nehalem HT can be a better implementation then netburst implementation so I give it a sceptic at this point in time.

Francois:
Sometimes here the talk goes round and round and misses the point.
Give me a dual socket Nehalem board with a bios that will let me work it like your single socket nehalem board and thats all I need.
unlocked multi cpu's, adjustments in bios for memory and voltages.
The rest is small stuff.
With that the cpu's will prove themselves just as they did in the single socket board.

early platforms are getting prepared now, so start pushing your contacts. For me few weeks to arrive, first toy after holiday....

[T_Flight]

One point is abundantly clear...lemme break it down in real simple english, and in a geek level language that everybody can understand...


This friggen CPU owns! It really is that simple. 2 months ago there were posts that Intel was gonna take the server market by storm. Now check out what's happening. They ain't selling by marketing gimmicks, and publicity stunts. They merely have a company showing the facts, and that speaks volumes about what is about to happen.

Anybody with a clue realizes this new archtichture is a technological monster, and even that is an understatement. I have never before seen these kinds of leaps in every area, and at the same time saw a CPU that excelled at almost everything.

This is unprecedented. I've built and done OC'ing for 20 years and have not seen this kind of a leap.

I also remember the Conroe timeframe, and man 'o man did that ruffle the feathers of the AMD faithful. That really got their blood boiling, but guess what? It was AMD who called out Intel in those years? Anybody remember?

Let's take a trip back in time to a website called "AMDZone", where a lone Athlon was standing in the ring up against the ropes all lonely, like it was waiting for a fight...

Well, they got one didn't they? Nobody can blame anybody else but themselves for that one.

Now here we are after the Desktop and Gaming PC market has been taken over, and Intel is going for the Server market. Any guesses who my money is riding on? I'll give a hint...it's Money in the bank.

[DoubleZero]

HT is great... for benchmark apps. The guys drooling over the sap records must have missed the reviews of I7 where there it gets amazing results with HT on in... benchmark apps, and then in real word apps, it gets the same or even worse results then with HT off.

[Stukov]

I love how "Intel used L3 first on the P4" yet don't recall the K6-III that had L3 and that was where the TLB error came from and was never fixed.

[accord99]

I love how "Intel used L3 first on the P4" yet don't recall the K6-III that had L3 and that was where the TLB error came from and was never fixed.

Of course, the L3 cache on the K6-3 was on the motherboard and not on the CPU die itself.

[gosh]

Anybody with a clue realizes this new archtichture is a technological monster, and even that is an understatement. I have never before seen these kinds of leaps in every area, and at the same time saw a CPU that excelled at almost everything.

I think that AMD tried to say that when the where out with phenom, but they don't have the market share that intel has and not that much money to do marketing. Now when i7 is out then people starts to understand

[Stukov]

Of course, the L3 cache on the K6-3 was on the motherboard and not on the CPU die itself.

Ahh, you are right. I was thinking it was on die, that was the L2 that did that and made the chip too big.

[Shintai]

This is why Dunnington shows a near-1.5x scaling compared to Tigerton.

I would look on the cache architecture if I was you

[Shintai]

I think that AMD tried to say that when the where out with phenom, but they don't have the market share that intel has and not that much money to do marketing. Now when i7 is out then people starts to understand

Problem is, unlike Nehalem. Phenom didnt deliver. Phenom was technological "pure". But performance was nowhere to be seen.

[justapost]

The results of spec are not tune for "marketing reason", but for "performance"

True, accord had posted an link whom explains the differences between throughput and response time optimisation. For a good picture we'd need both extremes and maybe one in the middle with 1,5s average response.

The rules of specs are very clear, and every industry player can join. Intel has one vote out of many many, the source code for SPEC goes through a process where SUN, AMD, IBM are all represented, plus many other companies. AMD has been recommanding SPEC FP RATE for more than 3 years, will it become a bad benchmark? If so, why? Same for the SAP benchmark, it was one of their favorite bench 1 before Conroe showed up ...

Nobody can say that SPEC is not fair, just come and join and see, it is running very democratically, and very fairly. If you have any proove of unfairness about spec, please let me know, I got to feed back to the SPEC Gurus at intel.



SPEC is a good benchmark, I don t like SPEC_FP_RATE, because I don t think a machine run many of the same task in parrallel, starting at the same time, and if it does , it is very often in a Innerthreading way.

This is my personal opinion, my employer is not responsible for my post on xtremeSys

German heise magazine uses spec in a fair way. They normaly get comparable systems from intel and amd and then run spec gcc compiled, icc compiled with and without optimisations. Here also spec_power results are usefull but those results tuned for a single benchmark are not helpfull if you run broad range of different workloads in production.

BTW: Saw they changed the diagramm on Anandtech, so both results had HT enabled.

[gosh]

Problem is, unlike Nehalem. Phenom didnt deliver. Phenom was technological "pure". But performance was nowhere to be seen.

They was early yes! And because of that the advantages wasn't that obvious. You need to understand how the processor work in order to understand the advantages. Applications wasn't that multithreaded also, so few applications was able to show the performance gain using this type of processor. Also developers need to adapt to the main market and that means that they need to adapt to Core 2. Developing for a processor (even if that processor has a more modern design) that has a tiny market share isn't economically smart.

[savantu]

I love how "Intel used L3 first on the P4" yet don't recall the K6-III that had L3 and that was where the TLB error came from and was never fixed.

BS.

There is no relation between TLB bugs on similar designs ( K10/Nehalem) , not to mention CPUs designed 10 years.

Each patient with its own disease.

[Stukov]

BS.

There is no relation between TLB bugs on similar designs ( K10/Nehalem) , not to mention CPUs designed 10 years.

Each patient with its own disease.

Uh, the TLB was a bug from previous designs of L3 cache that was never fixed. It's Fudzilla but it was the first in google search http://www.fudzilla.com/index.php?op...=6242&Itemid=1

We learned that the good old K6-III had a similar L3 cache errata to what we now know as the TLB bug, and the K10 simply inherited this almost a decade (nine years) old problem. After more than a year AMD finally found a solution for TLB and the new B3 CPUs will finally fix this raging issue.

The K6-III was the first CPU to have L3 cache from AMD and it had some similar issues. At that time AMD didn’t make a big deal out of it and it all went by quietly, but at that time AMD didn’t really have a strong server market presence, as TLB mostly affects servers, not desktops.

It is good to know that AMD solved this issue, but the issue itself ended up costing AMD billions. What can you do, sooner or later you have to slip.

[Macadamia]

I would look on the cache architecture if I was you

The AMD chips suffer from none of that, so core scaling would be as simple as i7's etc. Just add cores and L3 (if you wish).


Once CHTT 3+ gets onto Istanbul I don't think you'd see real problems with even 24 cores on 4P.