Istanbul vs. Nehalem clck-4-clock (almost) in Linpack

[Movieman]

Looking forward 2.8Ghz six-core Lisbon with DDR3 in Q1

Perkam

and then we can put up a dual 6 core gulftown against it and see a more realistic comparision as long as the prices are close.
That's a comparision I'm looking forward to and god's truth, I hope the AMD is neck and neck with the Intel..
Good competition is good for everyone including the two companies.

[Nedjo]

It is unfortunately.

Another good try Nedjo, as always.

try?

I just posted bloody link on to some HPC company site that apparently is building AMD and Intel HPCs and obviously have means to test server processors from both companies in clock-4-clock fashion! in HPC environment synthetic benchmarks.

[villa1n]

Then read the rest of that company blog/s, you'd see that it's almost void of anything positive about Intel. Go ahead, spend a little time doing what you expect others to do

http://www.advancedclustering.com/company-blog/

It seems their clientele list is fairly extensive, and it would seem that they would probably not take kindly to any sort of playing favourites in highly competitive fields, in the corporations dept...

Most benchmarks show nehalem out ahead, but it may be that in LinPack type tasks are its achilles heel, kind of like superpi for amd. Doesn't seem that hard to believe given the design strategy nehalem had on cost and efficiency, maybe it wasn't cost effiecient to design the processor to dominate at a niche task, instead it takes overall effieciency for the most used tasks in their biggest market , small business's.

It feels like some people take things a little personal, they are benchmarks, and losing 1 benchmark doesnt make or break any architecture..its the big picture. These guys are worried about Price / Gflop ratio's.. given they have highly specific tasks as cluster computing...i m sure that is the bottom line for them...as for us, that use cpu's for other tasks... its a nice piece of info to know.. but not to bank a decision on.

And to be fair, Nedjo is allowed to be excited about the company he works for, just like DrWho is... I envy them actually, my work is boring and tedious. lol.

Company list here http://www.advancedclustering.com/customers/

NASA , NOAA, Department of National Defense(Canada), M.I.T., Harvard, Rice, Lockheed MArtin, Boeing, National Institutes of Health (i m pretty sure it d be hard to fool these guys),

[STaRGaZeR]

try?

I just posted bloody link on to some HPC company site that apparently is building AMD and Intel HPCs and obviously have means to test server processors from both companies in clock-4-clock fashion! in HPC environment synthetic benchmarks.

Maybe if their testing metodology wasn't biased people would be more interested. Also, as you perfectly know because you use them for your own convenience as an AMD employee (yay finally confirmed!) only when they make AMD look better, synthetic benchmarks are good for nothing. You can continue with your hypocrisy, but don't be surprised if someone calls BS on you.

[initialised]

What does Istanbul get on the Windows 7 index?

[accord99]

Most benchmarks show nehalem out ahead, but it may be that in LinPack type tasks are its achilles heel, kind of like superpi for amd.

It's not really an Achilles heel; it's just that LinPack is generally well-optimized and since Nehalem and K10 have the same 4 DP FLOP/cycle theoretical peak, whoever has the most CPU cores*MHz will end up being faster. Much like how Xeon Paxville were faster than K8 Opteron DCs in LinPack because they both had the same theoretical peak DP capabilities but the Xeons had more MHz. Of course, that didn't translate very into the vast majority of applications.

[Nedjo]

Maybe if their testing metodology wasn't biased people would be more interested.

biased... using Intel compilers, configuring workload to fits 100% of memory on both systems... well I guess they're willing to ruin their required objectivity that is assuring them accounts like NASA , NOAA, Department of National Defense(Canada), M.I.T., Harvard, Rice, Lockheed MArtin, Boeing, National Institutes of Health

Also, as you perfectly know because you use them for your own convenience as an AMD employee (yay finally confirmed!) only when they make AMD look better, synthetic benchmarks are good for nothing.

I was under impression that those benchmarks are used when that company sells systems (including Intel based) to accounts like: NASA , NOAA, Department of National Defense(Canada), M.I.T., Harvard, Rice, Lockheed MArtin, Boeing, National Institutes of Health

You can continue with your hypocrisy, but don't be surprised if someone calls BS on you.

Nothing can't surprise me from Intel fanboys like yourself! Even when I don't write single comment in a post on someones benchmarks you guys get nasty

[villa1n]

It's not really an Achilles heel; it's just that LinPack is generally well-optimized and since Nehalem and K10 have the same 4 DP FLOP/cycle theoretical peak, whoever has the most CPU cores*MHz will end up being faster. Much like how Xeon Paxville were faster than K8 Opteron DCs in LinPack because they both had the same theoretical peak DP capabilities but the Xeons had more MHz. Of course, that didn't translate very into the vast majority of applications.

Ahh, that makes sense. And it also fits why that would be important for a cluster provider, considering their code is generally well-optimized as well, for specific tasks. ^^

[STaRGaZeR]

biased... using Intel compilers, configuring workload to fits 100% of memory on both systems... well I guess they're willing to ruin their required objectivity that is assuring them accounts like NASA , NOAA, Department of National Defense(Canada), M.I.T., Harvard, Rice, Lockheed MArtin, Boeing, National Institutes of Health

In case you don't know, the Intel compilers are faster on AMD processors too. Also, if you have ever run Linpack you'd know that this is not an apples to apples comparison. You even hide it in the thread title (with the "almost" ). And yeah, I'm sure the NASA is running Linpack all day long.

Synthetic benchs are good for nothing. I really think you do understand what this is all about but instead you have forgotten what a reference application is just to make AMD look better. I wonder why you have not spammed SPECfp_rate numbers in the forum since Nehalem came out like you used to do. It's like the SPEC benchmarks are no longer used as reference in the server world! What a change, don't you think?

Nothing can't surprise me from Intel fanboys like yourself!

That's funny coming from AMD people. Now go find another apples to oranges comparison so we can make fun of you again.

[poke349]

Why is there a flame war over every single benchmark?

Benchmarks are benchmarks. Some favor Intel, some favor AMD, live with it!

[Hans de Vries]

http://www.advancedclustering.com/co...eron-2400.html

Well, The peak GFLOP performance of a 2.60 GHz Istanbul is about equal to
that of a 4 GHz Nehalem so a 34% lead for Istanbul is not really unexpected.

The nice surprise (for Istanbul that is) is that this happens with the data-
sets in external DRAM instead of the caches as is sometimes the case with
Linpack.

Many benchmarks used lately are very bandwidth sensitive favoring the
high bandwidth Nehalem, specially in Server applications were the Opterons
are limited to dual channel 800MHz DDR2. The Maranello platform should
improve the bandwidth with about 66% to dual channel 1.33 GHz DDR3 like
the current Phenom II parts.

Regards, Hans.

[Nedjo]

In case you don't know, the Intel compilers are faster on AMD processors too.

Sooo? Does this make that article biased?

Also, if you have ever run Linpack you'd know that this is not an apples to apples comparison. You even hide it in the thread title (with the "almost" ).

I've hide it?! How? buy supplying that info in the TITLE??

And yeah, I'm sure the NASA is running Linpack all day long.

mature, very mature

Synthetic benchs are good for nothing. I really think you do understand what this is all about but instead you have forgotten what a reference application is just to make AMD look better.

Let me remind you, since obviously you don't have clue what this thread is about: I've provided single link, to a single article about single benchmark that compares Istanbul and Nehalem! That's it! No conspiracy theories no suggestive messages, just one benchmark! And if it's troubling you please complain to the author of that article (mail is provided over there), and don't shoot the messenger !

I wonder why you have not spammed SPECfp_rate numbers in the forum since Nehalem came out like you used to do. It's like the SPEC benchmarks are no longer used as reference in the server world! What a change, don't you think?

errr...
cos I didn't spammed anything, I just provided link to an article about some other benchmark, that doesn't have SPEC in its name!

That's funny coming from AMD people. Now go find another apples to oranges comparison so we can make fun of you again.

as I wrote Intel fanboys comments don't touch me, and reading all your comments that don't have anything to do with this thread and have all to do with me and imaginary conspiracy theories I wonder of whom should fun be made

[hyc]

In case you don't know, the Intel compilers are faster on AMD processors too. Also, if you have ever run Linpack you'd know that this is not an apples to apples comparison. You even hide it in the thread title (with the "almost" ). And yeah, I'm sure the NASA is running Linpack all day long.

That's funny coming from AMD people. Now go find another apples to oranges comparison so we can make fun of you again.

Actually, the matrix operations used in Linpack *are* run all day long in a lot of HPC work. My former group at JPL was totally dedicated to array processing. This is the stuff that space engineering is built on, and benchmarks like this are the main criteria upon which big computing purchase decisions are based.

[stangracin3]

can someone ban everyone with an intel system in there sig from an amd thread? cause THIS GETS ING ANNOYING AFTER A WHILE GOD DAMNIT.

[hyc]

So is Linpack the Super Pi of benches in the Server world?

That would be a fair question, if the code used to do the computations in Super Pi were routinely used by computer professionals to get their real work done day in and day out. Or if the output of the code was ever actually used for something. I guess, if one of you ever bothered to memorize 32 million digits of pi, and crosschecked SuperPi's results, then one might be able to consider your use of Super Pi to be comparable to using Linpack.

[poke349]

That would be a fair question, if the code used to do the computations in Super Pi were routinely used by computer professionals to get their real work done day in and day out. Or if the output of the code was ever actually used for something. I guess, if one of you ever bothered to memorize 32 million digits of pi, and crosschecked SuperPi's results, then one might be able to consider your use of Super Pi to be comparable to using Linpack.

Some of the algorithms used in computing Pi are applicable to other fields. But aside form that, Pi is really only useful for benchmarking and stress-testing.

[Movieman]

can someone ban everyone with an intel system in there sig from an amd thread? cause THIS GETS ING ANNOYING AFTER A WHILE GOD DAMNIT.

Does that mean I have to ban myself?

Short version guys, stop the baiting here or leave and I mean that.. Enough ok?
All the guy did was post a damned link.
No need to go after him like he insulted your mother!

[perkam]

*snip*

Ty

So many threads end up this way Why can't we all just love each other('s workstation rigs)

Perkam

[Donnie27]

It seems their clientele list is fairly extensive, and it would seem that they would probably not take kindly to any sort of playing favourites in highly competitive fields, in the corporations dept...

Most benchmarks show nehalem out ahead, but it may be that in LinPack type tasks are its achilles heel, kind of like superpi for amd. Doesn't seem that hard to believe given the design strategy nehalem had on cost and efficiency, maybe it wasn't cost effiecient to design the processor to dominate at a niche task, instead it takes overall effieciency for the most used tasks in their biggest market , small business's.

It feels like some people take things a little personal, they are benchmarks, and losing 1 benchmark doesnt make or break any architecture..its the big picture. These guys are worried about Price / Gflop ratio's.. given they have highly specific tasks as cluster computing...i m sure that is the bottom line for them...as for us, that use cpu's for other tasks... its a nice piece of info to know.. but not to bank a decision on.

And to be fair, Nedjo is allowed to be excited about the company he works for, just like DrWho is... I envy them actually, my work is boring and tedious. lol.

Company list here http://www.advancedclustering.com/customers/

NASA , NOAA, Department of National Defense(Canada), M.I.T., Harvard, Rice, Lockheed MArtin, Boeing, National Institutes of Health (i m pretty sure it d be hard to fool these guys),

Please go back and look at what I said prior to that? I said NO ONE expects Nehalem to be faster on all benchmarks.

I've not said anything negative about Nedjo. My one comment was in relation to Nedjo and Dr Who?. I said I've read the good, the bad and the ugly from both of them. Meaning I've agreed and disagreed with both of them. Lately I've disagreed with Dr. Who? more. I respect folks who come here and Identify themselves as Employees of any company. I wouldn't piss on a shrill to put them out if they were on fire.

How something is tested is more important than the results. I said that when Extreme Tech showed Athlon64 winning 10 out of 11 tests and the same thing when Conroe won 10 out of 10. Without meaningful consistent tests, folks get mislead and end up with egg on their face.

The blogger and the Company aren't the same. I'm not sure if that blogger is the head PR guy for them or one AMD biased fan within the company. I stand by what I said looking at the rest of the blog. Intel is hardly mentioned. The test was poorly performed and that weakens the results. But that's just my opinion, more power to those who thought is was good or OK.

Last but not least, the other poster could have omitted that crap about "retards" and etc.......

[hyc]

Some of the algorithms used in computing Pi are applicable to other fields.

Sure, but I wasn't talking about the algorithms, I asked about *the code*. Linpack is an industry standard library that is used everywhere in real HPC work. Is the SuperPi code available in a library that anybody else in the world uses for any practical purpose?

[poke349]

Sure, but I wasn't talking about the algorithms, I asked about *the code*. Linpack is an industry standard library that is used everywhere in real HPC work. Is the SuperPi code available in a library that anybody else in the world uses for any practical purpose?

Maybe, it depends on how it's written. I haven't seen it so I wouldn't know.
But superpi's code won't be very useful because it's too ridiculously slow compared to virtually every other pi program out there.

If there's any "large-number" program that could have significant use, that would be Prime95... I'll spare the details here. Too technical...

[stangracin3]

Does that mean I have to ban myself?

Short version guys, stop the baiting here or leave and I mean that.. Enough ok?
All the guy did was post a damned link.
No need to go after him like he insulted your mother!

no because your old and your an admin

lol j/k j/k

[Movieman]

no because your old and your an admin

lol j/k j/k

I'm not that old.. can still play tackle football with the young guys if needed..
Of course the next day you might not hear from me..

[villa1n]

Please go back and look at what I said prior to that? I said NO ONE expects Nehalem to be faster on all benchmarks.

I've not said anything negative about Nedjo. My one comment was in relation to Nedjo and Dr Who?. I said I've read the good, the bad and the ugly from both of them. Meaning I've agreed and disagreed with both of them. Lately I've disagreed with Dr. Who? more. I respect folks who come here and Identify themselves as Employees of any company. I wouldn't piss on a shrill to put them out if they were on fire.

How something is tested is more important than the results. I said that when Extreme Tech showed Athlon64 winning 10 out of 11 tests and the same thing when Conroe won 10 out of 10. Without meaningful consistent tests, folks get mislead and end up with egg on their face.

The blogger and the Company aren't the same. I'm not sure if that blogger is the head PR guy for them or one AMD biased fan within the company. I stand by what I said looking at the rest of the blog. Intel is hardly mentioned. The test was poorly performed and that weakens the results. But that's just my opinion, more power to those who thought is was good or OK.

Last but not least, the other poster could have omitted that crap about "retards" and etc.......

Donnie, I was commenting on your quote of Salad, and in particular, how you said and i will requote

Then read the rest of that company blog/s, you'd see that it's almost void of anything positive about Intel. Go ahead, spend a little time doing what you expect others to do

If you had actually read the company blog, you would see the last 2 posts identify the author, as an engineer for the company, you would also notice this... stream processing compared source, and i m not sure you can construe these results as being favourable to amd...

Your earlier part where you "pointed out the obvious flaw in the testing" which biased the results to being unfair towards intel, but it wasn't clear what that was, since it was addressing a memory difference explained in the blog as to why it was tested that way. although i m not sure bandwidth wise that triple channel ddr3 is fair to compair against dual channel ddr2. I think they were trying to reflect as best as possible real work situations, with the platforms having their strengths catered to, like they would in the realworld, and then having all other things be equal. Fairly logical, and applicable since the blog is for people in the HPC community i m sure. Again, you in the same post where you said it wasn't fair for amd to use "features" like its ability to properly utilize 16gb of ram, you had stated "

Yet, those results shown looks just like only 4 cores with HT turned off. If folks are going to use these features, they should be turned on (if they have a negative affect, then so be it).

This OBVIOUS flaw was addressed, and it was for the reason you said it was biased ironically, they were playing to each architectures strengths.(see FEATURES)

The amount of RAM varies due to the Nehalem providing the best performance when using its tri-channel memory architecture versus the Opteron's dual channel. Since HPL performs best when using as much memory as it can, we adjusted the problem size (N in the HPL configuration file) to use as close to 100% of the RAM on the system as possible.

Meaning they adjusted the problemsize to fill up amd's ram to compensate for the difference.

So i m not clear on where the OBVIOUS error was. You seem to try and make yourself feel unbiased by saying, look at the source, which you obviously didnt yourself, and "you ll see there are no intel positive results in it blah blah blah." Unless ofcourse, 1 specific test, yeilding a positive AMD result, makes something biased....lol and that was why i said it was just one benchmark, and that people should seperate their identity from their processors a touch more, your processor losing a bench, is something temporary, it only matters today, as new architecture will usurp the old..etc.. So to cut this intel trolling short, i wanted to point it out. Let people have their small victories, its not like anyone doesnt know intel and nehalem currently is a superior architecture in most situations, and this isnt directly at you donnie, this is also other fanboi's that come out the woodwork with their KEYBOARDS FLAMING, ready to degrade an interesting thread and results into my Puter, is better than your Puter!

And to whoever was mocking linpack, someone in the thread, i beleive addressed why those calculations are relevant further clarifying what the article stated, and further more we had 2 other people clarify why AMD got these results based on how those calculations are computed, so to those in the know, it was no surprise. I believe it was Hans De Vries.

While many different benchmarks have been published comparing both systems, they all have seemed to fall a bit short in showing an accurate comparison of the two platforms.. The HPC industry standard for benchmarking is HPL or High Performance Linpack, on which the "Top 500" list is based. Many industry insiders look to the results of this benchmark as an insight into how their systems will perform.

- This to me seems highly relevant in HPC computing.




For all of us that are not Compsci phd's and systems engineers, our energies might be best served listening to those who do have knowledge, and asking relevant questions, instead of what i can no better describe than a pissing contest happens.

And thanks movieman for trying to stop the inevitable spiral downwards ^^

[Dante80]

The fact that AMD countered Intel on the servermarket i suppose

AMD counters (on performance/watt/$) Intel on the servermarket for many years now (partly due to bad choices from Intel in the past)...Nehalem EP though is truly an amazing server chip, that offers as a platform performance on a scale unheard of before. That doesn't mean though that it will win all benchmarks @ all conditions @ all setups.