First "real" Nocona vs. Opteron review?

[rck01]

Justly,

As I mentioned before, this is a project for the InfoWorld Test Center. A private study, for one of our corporate clients, would be more specific in nature. However, given the data at hand, and taking into account the usage patterns observed at various customer sites, I'd be hard pressed to recommend Opteron as a workstation solution for my customers.

Also note that, despite the objections noted here, the vendors themselves are not disputing our findings. IBM feels they got a fair shake, as do the other particpants. In the end, that's all that really matters...

RCK

 

[NewBlackDak]

The part that has everyone up in arms is that your "findings" are opposite of what everyone else if showing.

 

[clarkey01]

rck01 really works for Intel ;-)

 

[rck01]

Wrong. Though I must admit, I'm surprised it took this long for someone to trot out the "works for Intel" accusation. Must be a slow day...

RCK

 

[Klixxer]

Originally posted by: rck01
Wrong. Though I must admit, I'm surprised it took this long for someone to trot out the "works for Intel" accusation. Must be a slow day...

RCK

Honestly, can you claim to be unbiased?

 

[mamisano]

"Randall C. Kennedy is the Director of Research and co-founder of Competitive Systems Analysis, Inc.... Later, as a contract testing and development engineer for Intel Corporation"

http://www.csaresearch.com/about.asp

 

[Klixxer]

Originally posted by: NewBlackDak
The part that has everyone up in arms is that your "findings" are opposite of what everyone else if showing.

Anyone could design a set of benchmarks to show either CPU's inferior, but try to do this: Take both of them, at the same cost and compare them, the Nocona is nowhere near the Opteron when it comes to price/CPU.

Well, why not throw Ultrasparc or PowerPC into the equation? The price range is so different that it would make as much sense to do so.

I read some BS that you didn't want to use the built in RAID function, could that be because the RAID functions in those boards would actually lower the net computing power as it only contains software RAID through BIOS and uses the CPU as a RAID controller?

 

[Klixxer]

Originally posted by: mamisano
"Randall C. Kennedy is the Director of Research and co-founder of Competitive Systems Analysis, Inc.... Later, as a contract testing and development engineer for Intel Corporation"

http://www.csaresearch.com/about.asp

Nice find, he's a sham.

 

[rck01]

Said I don't work for Intel. Now. Present day.

Never said I *didn't* work for them in the past... and for AMD... and for Cyrix... and for VIA... Microsoft... Compaq... HP... IBM... Dell... Citrix...

Experience. There is no substitute.

RCK

 

[clarkey01]

rck01, you take the fun out of forums.


in the last post directed @ me you Imply that im slow, what the hell im 18, funny though at my ripe age I still know Opteron is a better choice then Xeon currently.

Hold a poll

 

[Vee]

rck01, - do you have any preliminary suggestion (technical) as to why Xeon, in this instance, performs better under many concurrent workloads?

 

[greyhoundguy]

It all depends on what you are doing. A fairly simple simulation shows that the AMD solution has a fairly severe I/O limitation that the Intel solution does not have. So if you want I/O throughput then the processor/memory speed is not your gating item. Intel used to suck horribly on CPU/memory throughput. Nocona and Lindenhurst makes it competitive to the Opteron (maybe better, maybe not, but at least in the same ball park). But if you need bus based throughput of any kind then Intel has always been the performance choice. So I hope that rck will be either benchmark the application his clients have, or he should be running a broad range spanning from CPU intensive to I/O throughput intensive.

 

[Vette73]

^

What are you basing that off of? In fact there have not even been any reviews yet of the Nocona and Lindenhurst. Heck even Intel has said there are problems.

Fanyboy rck01?

 

[mamisano]

I said it once, I'll say it again... use a NUMA aware OS like Win2K3 and then get back to me on Memory performance.

Also, what Video cards were used in each system? If you pulled the SCSI drive from the IBM system because the rest were SATA, the you should have pulled the Video too and baselined to one card.

What was the purpose of this "review" anyway? To test as-is workstation performance (obviously not) or to simply test Xeon vs. Opteron?

 

[greyhoundguy]

I do my simulations based on statistical models of multiple requestors and average sustained throughputs. Mostly because the work I currently do is involved in high rates of network traffic, I focus on the PCI throughputs. With mulitple requestors, peak throughput is reduced to somewhat less than half to a third, depending on the chipset design and the numbr of requestors. The AMD 8131 supports only 2 independent PCI buses and the Intel solutions support 4. In real world tests we easily reach bus saturation on the AMD platforms. On older Intel platforms we easily reach CPU/memory saturation.

The new Lindenhurst chipset has a much better memory architecture than Intel's old chipsets. This will make it competitive in processor/memory bandwidth. It all depends on what you are looking to do with your machine. If I were specifying a high end single user workstation then I would go AMD. If I wnated a high end network server, then I would spec out an Intel solution. When the Nocona/Lindenhurst boards are here and stable, maybe they will be just as good as the AMD solution for workstation/gaming systems. Maybe.

 

[mamisano]

Originally posted by: mamisano
I said it once, I'll say it again... use a NUMA aware OS like Win2K3 and then get back to me on Memory performance.

Also, what Video cards were used in each system? If you pulled the SCSI drive from the IBM system because the rest were SATA, the you should have pulled the Video too and baselined to one card.

What was the purpose of this "review" anyway? To test as-is workstation performance (obviously not) or to simply test Xeon vs. Opteron?

Still waiting for a response...

 

[Zebo]

I guess in this age of political correctness "fairly well" means it get it's ass handed to it?

http://www.gamepc.com/labs/view_content.asp?id=noconaopteron&page=13

Also , everyone, this is a commerical site selling- hardware. would you trust your GM dealership to do car reviews? Fine cancel my consumer reports subscription.

Lets all get our reviews from "CSA research" or "GAMEPC" who have a financial incentive to show certain "results"

 

[Zebo]

Originally posted by: Klixxer

Originally posted by: mamisano
"Randall C. Kennedy is the Director of Research and co-founder of Competitive Systems Analysis, Inc.... Later, as a contract testing and development engineer for Intel Corporation"

http://www.csaresearch.com/about.asp

Nice find, he's a sham.

Certainly sounds full of sh1t with lots of buzz words...reminds me of a Don Lupree presentation..check this out:

Randall C. Kennedy( why not add esq. in there with the pompus middle initial lawyers love to use?..but then I guess you'd have to graduate college for that.) is the Director of Research ( notice the high falooting title he has...how many employees you got, 1? You forgot "Janitor" "Secretary" and "labrat" wonder why?) and co-founder ( Whos the other your mom or wife?) of Competitive Systems Analysis, Inc. A 20 year veteran of the IT industry, Mr. Kennedy is known world over (never heard of him, moreover who gives a sh1t? Think Bill gates says "he's known world over"???!!? Hell no, he does'nt have to toot his own horn..sales and product movement do it for him. ) as an analyst, developer and freelance writer. Prior to heading CSA's Research organization, Mr. Kennedy worked for Giga Information Group as the Sr. Industry Analyst covering Windows NT and related operating system issues. Later, as a contract testing and development engineer for Intel Corporation, he led the effort to create tools and resources to articulate the company's performance initiatives surround high-end desktops (Constant Computing) and Gigabit Ethernet networking. ( blah blah blah... any HS kid with a A+ certificate can do the same..we should call Intel and see if this joker really ever worked there..BTW I love the capitalization of "gigabit ethernet" most impressive must be important... and who specializes in "low end" desktops?)

Mr. Kennedy is an internationally recognized expert ( ohoo first "world over" now "internationally" there you go again, according to whom??!?! I bet if you had a three paragraph BIO you'd say "the universes expert" just to mix buzz words up eh??) on personal computer operating system technology, having authored and co-authored numerous OS-related books, including: "Migrating to Windows NT" (Brady); "Ultimate Windows 3.1" (NRP); and several DOS and Windows titles from New Riders Publishing. He is also a frequent contributor to numerous Ziff-Davis, IDG, McGraw Hill, Duke Communications and Miller-Freeman publications where Mr. Kennedy?s insights have been read by millions of IT professionals world-wide. ( why are you talking in third person..I doubt anyone wrote this bio for you??!?!)

 

[sandorski]

Not wanting to get in the middle of this interesting conversation ( ), are there any links to new real reviews/comparisons yet? Other than the GamePC review which is Old News.

 

[DonPMitchell]

Speaking of NUMA aware OS, I got the impression the GamePC article was saying they *did* use one. Didn't they run the benchmarks on the beta 64-bit WinXP? And he says that is NUMA aware. I went back and checked, but maybe I'm reading it wrong.

This was interesting, because it really let the Opterons go to town when benchmarks benefited from very independent parallel speedup.

 

[clarkey01]

Zebo I LOVE your last post.

I didnt have balls to do it .

 

[mamisano]

Originally posted by: DonPMitchell
Speaking of NUMA aware OS, I got the impression the GamePC article was saying they *did* use one. Didn't they run the benchmarks on the beta 64-bit WinXP? And he says that is NUMA aware. I went back and checked, but maybe I'm reading it wrong.

This was interesting, because it really let the Opterons go to town when benchmarks benefited from very independent parallel speedup.

They were able to install XP64 on both machines, but without 64-bit PCI-Express drivers for the Intel board, both machines ran standard Windows XP.

 

[rck01]

Honest debate isn't working so you resort to personal attacks. Typical fanboy behavior - and a waste of my time.

Disengaging...

RCK

 

[mamisano]

Originally posted by: rck01
Honest debate isn't working so you resort to personal attacks. Typical fanboy behavior - and a waste of my time.

Disengaging...

RCK

Well, that's an easy way out of answering NUMEROUS unanswered questions...


come on RCK, you can do it...

 

[clarkey01]

it would be an Honest debate if you answer the questions posed.

But you wont, becaues you cant.