Thinking about setting up a Woodcrest vs Opteron duel
What about an Opteron and Woodcrest duel?
There will be no SunGard ACR Intel Demo Version 0.1 guerilla benchmarketing stuff, no AMD crippling Intel compiler, just GCC, Linux (x86_64)+ PHP/Perl+Apache+MySQL+OpenSSL+Tomcat (add any open source stuff you want). I will set up the servers, connect them to the internet and allow people to login and check. People will sign up to get their time slots to run the benches, while others can observe the results via a web interfacce. I will invite the gurus visiting this blog to write the test scripts. Everything will be open, fair and reproducible. It will be an Open Source dual core duel.
The question is how do I pay for this. Maybe Intel and AMD can send their best, hyper tuned boxes. Or maybe we just buy an HP DL380 and an HP DL385. Or we setup a web page and put ads on it, and hope it will generate enough ad money to cover the cost...
The tests will be made as relevant as possible for real server usages: Web, FTP, MAIL, DNS, DB, Game Server, Video/Audio streaming...
This is serious business, folks. We can expand the idea from here to cover more products...
posted by Sharikou, Ph. D. @ 6/29/2006 03:57:00 PM
56 Comments:
I thought AMD had challenged Intel to a Dual Core Duel again, Woodcrest vs Opteron.
Nevertheless, I think it's a great idea. You should include power consumption benchmarks, and 1P, 2P and 4P configurations. It should be interesting to see how scalabe those processors are.
About the money, you could ask other reviewers how they manage to get their processors.
NICE! i like ubuntu linux!! the most polished distros. get intels so called WOODCREST and watch it get creamed by the opteron. or better yet get conroe aganst fx62 and get the $400 ram, you know corsare ultra low latency ddr2 800 memory.
madmodmike has a website called rubyworks and he might get it set up with your help.
you should pretend to be an intel pumper and an amd pumper so you can get the free stuffs for this duel. i got you got a great idea and i hope all those reputable sites will lend a hand. by those sites i mean the intel paid pumper sites. is an s939 opteron considered a server proc? too bad i don't have an opteron because i would gladly send it to you. i hope you have success with this.
Interesting.
But i think to buy these machines it's a lot of money, and too time to find it.
The only way it's to get some demo-test machine from some reseller adding his ads to the test page.
agree with the idea and I agree with the suggestion to use a Ubuntu base. It would be very simple on a Debian system to duplicate the tests, which was one of the complaints leveled against the Anandtech tests.
I would also venture that it would probably be a good idea to use the systems to run game servers like UT2k4, Quake4, or F.E.A.R.
That would put a crunch on the systems.
Funny, I would have expected those Intel employees and fanbois screaming that they will assist you 100% in getting the best Intel setup for this duel. Now the lamers are suddenly silent.
Come on you wankers, show us your damn faith! What are you afraid off?
*Queue echoo effect* The truth. The Truth. The Truth.
Sharikou is there any local PC shop that has stock for these chips? If so they could lend you for 1 day stuff needed to setup in their shop. And if they want get free plugs with a banner or a mention or whatever.
how do I know you Sharikou or any "experts" you invited are not AMD paid pumpers??
Now the lamers are suddenly silent.
No. They are not. They posted zillions of profanities to comment on this proposal. I had to delete those comments.
how do I know you Sharikou or any "experts" you invited are not AMD paid pumpers??
The whole thing will be open source. Every script, every configutation is open to examination via remote login. Also, we will only choose those reproducible benchmarks, nothing dependent on unknown issues .
Good idea. However I would suggest a server distro instead, e.g. Suse, RedHat, Debian.
I think HP Opteron vs Dell Woodcrest server would be nice.
Fantastic ideal, perfect timing; Dell and Intel have pumped millions of $$$ this week on TV adds highlighting their servers. I would think they would like to prove their ads aren’t just bull shit marketing, and then again this may turn out to be very expensive if the truth gets out. I don’t think their going to loan you one or even talk about it.
On the other hand, I’ll bet AMD is burning the rubber off their tires trying to get it to your test site.
If you pass a hat I’ll throw in a $100, just publish the website or address.
Honestly, you know what I think the real problem will be with this idea? Getting a hold of an actual Woodcrest CPU to test with. Nobody but Anand and his ilk seem to be able to get a hold of one right now.
Other than that, yes, this is a great idea... I can't think of anything fairer.
You will set up the servers and connect them to the internet and allow people to login and check. And should we *believe* that in your infinite fairness you will not try to bias the systems in AMD's favor in any way? Who is going to validate the setup? Who are your *gurus*?
Basically sharikou-no-phd, if you were unbiased and non-emotional in all your anti-intel posts perhaps people would listen to you more. Your duel will only draw the attention of pro-AMD fanboys who cannot accept what is staring them in the face.
Lol, I wonder if this post will even make it, or will you delete this too? Seriously what is it you have against Intel, why are ALL your comments so emotional? Oh and how about all the other websites (at least 10-15 which have favourable woodcrest reviews)? Did Intel pay them too?
Your duel will only draw the attention of pro-AMD fanboys who cannot accept what is staring them in the face.
Hmm. I wasn't even sure who would be the winner. But it seems Intel dudes aren't so confident after all. Guess they know those guerilla benchmarketing, hide-n-seek, hit-n-run stuff can't be subjected to open scrutiny.
I think it's a great idea!
If you want to sell ownership in this production, then I'm all in.
Say 1000 shares at $100 a pop sent to you in money-order by said date! That should be able to buy you two servers and a bit of marketing. Heck, I'm sure AMD will sponsor their side for free.
Next, sell tickets (say at $2 or $5 a peice) for which the winners (two of them) will win each a server. Should generate a shit load of money there too. Plus all the ads from visitors (and sponsors) will bring in lots more money.
Any surplus money will go to the investors!
I'm in for a couple of shares... let me know.
Note, the linux distrib should be as close to something sold by real server vendors. When I bought my Opteron, they were selling a 64-bit Debian.
Since most of the people who buy AMD/Intel chips run some variant of Windows, I think it would be more useful to do your benchmarking on a WIndows platform. IE, most readers will care since that is what most people run.
graham, I agree with that too... why not have both a 64-bit Linux and 64-bit Windows benchmarks!
Say 1000 shares at $100 a pop sent to you in money-order by said date!
Cool. That's some good business sense. We get paid for having fun. Why not? If there is enough interest. Let's do it. So we issue X number of shares, come up with the plan and do the work. Both AMD and Intel will be very nervous. Billions of dollars of server sales are at stake....
Most of the benchmarks we see are useless. I am think about what kind of tests we should consider. For example
1) anonymous FTP server
2) samba file server
3) DNS server
4) databases
5) mass web hosting (static pages, php, cgi...)
6) qmail/postfix server
7) mixed mode (everything on one server)
8) development box
For web server performance, instead of mindlessly retrieve the same page, we should emulate more realistic pattern, mixing static and dynamic pages....
" Most of the benchmarks we see are useless. I am think about what kind of tests we should consider. For example
1) anonymous FTP server
2) samba file server
3) DNS server
4) databases
5) mass web hosting (static pages, php, cgi...)
6) qmail/postfix server
7) mixed mode (everything on one server)
8) development box
For web server performance, instead of mindlessly retrieve the same page, we should emulate more realistic pattern, mixing static and dynamic pages...."
As it takes a lot of work to set up good benchmarks, I would take your list and focus on one area at a time.
This way you could present very useful information over a period of months and offer a "depth" level that really shows what is going on.
I'm sure you know this, but make sure for Linux that a good NUMA Linux is used. Not something that is popular because of hype but has next to zero real deployed servers.
Sharikou says...
Hmm. I wasn't even sure who would be the winner. But it seems Intel dudes aren't so confident after all. Guess they know those guerilla benchmarketing, hide-n-seek, hit-n-run stuff can't be subjected to open scrutiny
Dude... we all know that in your books only AMD can win. And you have not answered any other comments I made...
way to go:)
>>Did Intel pay them too?
The only server-like benchmarking I saw is Apache Web Server. Meanwhile the rest of the benchmark are SuperPI, LAME, DivX... ?! This is server?
Hi Sharikou!
Don't post this comment just wanted to get in contact with you! I found something very interesting! GeIL will soon offer DDR2 1066/1200 Modules (1GB & 2GB Kit) and thanks to the AMD Enhanced Parameter the modules will run with 1200 Mhz and CAS4; but ONLY on AMD Mobos!
"This setting is designed for AMD-based motherboards. The press released said the modules were pre-tested on the Asus AM2 M2N-32 Deluxe, using an AMD FX-62."
Couldn't find the press release on the GeIL-website, read it on www.planet3dnow.de
P.S. The Duel is cool and Woodcrest will be crushed!!!
Greetings from Italy
Van Dexter
Good idea!
Intel ignored that duel from AMD (maybe last year), and I'd really to see a fair competition between Intel and AMD chips, no paid pumpers and no more benchmark(s) optimized chips beating an one-year-old product.
Sadly, I do not have money to support your plan or buying any shares, but I will support you in spirit! Keep up the good work!
is done already... See spec.org.
Fujitsu Siemens PRIMERGY RX220 S1, AMD Opteron 280 processor Rock Web Server v1.3.3 (x86_64), Apache Tomcat 5.5.9
vs
IBM System x3650(3.0 GHz Intel Xeon 5160) Rock Web Server v1.3.3 (x86-64), Apache Tomcat 5.5.9
both 4 cores 2 sockets
Webperformance
Opteron: 8394
new Xenon: 9182
the Xenon is 9% faster. Should be available from 25 August. So Intel is closing the gap for some time.
I have been reading your blog for while and think this is a great idea.
You should be able to obtain hardware form SUN (Opteron), Dell (Woodcrest) for free for 30-60 days. To be honest depending on the outcome of Duel regardless who wins or loses these companies will likely let you keep the machines for free just for the generated publicity. Throw in some Solaris tests. Both Opteron and Woocrest actually perform better on Solaris 10 than any Linux distro.
Were burning day light; let’s roll. Get a PO Box today so we can start sending money.
Arlie, Chino Hills Calif.
Great idea!! I would love to see some benchmarks between Opteron and Woodcrest...
The idea of letting people remotely logging in to do the benchies, publish/question data, verify, and do whatever it is that is THEY do to represent how THEY use it... leaves the ?'s out of the equation.
To the poster that said there are 10-15 sites pumping the same results:
If 100 journalist points/stares at the SAME pile of dung/turd I shat (sp?) out, and says, "Yup, it smells/looks like a pile of dung... it must be true". (Analogy to Marketing via Internet websites with the SAME machine setup by INTEL, and says... please pust this "BUTTON", and record the results.)
^ ^ I dont' buy it ^ ^
I like the idea of fair and open competition! Wait, Intel declined the challenge... I wonder why?? :)
I'll chip in as an investor... ;-)
Why dont you call up your "friends" at dell to hook u up or maybe call up intel for it i am sure they are grateful for ur valuable advice that they didnt take and now their company is on the verge of bankruptcy like you predict in 7 Q.
Hahah you amd guys are stupid i like how everyone is wanting to pitch in but are posting under anonymous.
I should have a dual woodcrest system RSN. I'd be willing to run benchmarks on it, or possibly let a few others run them remotely.
We are mostly an opteron shop but like to keep and eye out for anything that improves price/perf or perf/heat.
Sounds great, but with your obvious bias noone will believe the results if AMD wins the benchmark.
Yes you say it is open source, but unfortunately the hardward is also 'open' to cheating.
Need an example? A little shot of spray insulation in the heat sink of either processor and - boom the processor throttles down.
How are your going to provide an audit trail that the hardware was not f*cked with in any way???
You say that some sites published results are purposely biased, I'm just saying that yours will be no different - in fact worse as you are an also ran in the community of those benchmarking.
Need an example? A little shot of spray insulation in the heat sink of either processor and - boom the processor throttles down.
No wonder Intel won't let people look inside the IDF boxes.
Yes you say it is open source, but unfortunately the hardward is also 'open' to cheating.
We will be using off the shelf stuff, no hidden doping or sabotage (like using an old BIOS). We will provide the exact hardware specification and our benchmark tools will be open source. Therefore people can reproduce the same result on their own identical machines. Anyone can challenge the results by sending his identical machine in for a drill.
" is done already... See spec.org.
Fujitsu Siemens PRIMERGY RX220 S1, AMD Opteron 280 processor Rock Web Server v1.3.3 (x86_64), Apache Tomcat 5.5.9
vs
IBM System x3650(3.0 GHz Intel Xeon 5160) Rock Web Server v1.3.3 (x86-64), Apache Tomcat 5.5.9
both 4 cores 2 sockets
Webperformance
Opteron: 8394
new Xenon: 9182
the Xenon is 9% faster. Should be available from 25 August. So Intel is closing the gap for some time."
Picking two different systems out of a SPEC list is NOT a benchmark.
The AMD system has the following for memory and disk:
Memory 16GB(8x2GB)DDR-RAM PC-3200R(CAS 3-3-3)
Disk Subsystem 2x250GB WDC WD2500JD, 30x36GB ST336753 CLAR36
Disk Controllers FastTrak S150 TX4(onboard), Emulex LP1050 (PCI-EX.)
Compare to the Intel system:
Memory 12 x 2 GB (DDR2 FBD 667MHz)
Disk Subsystem 146GB 15K Ultra320 SAS, 70 x 18GB Fibre Channel
Disk Controllers Onboard SAS controller
QLogic Fibre Channel QLA2342
Comparison:
The Intel system has 50% more RAM and far faster local drives, providing much faster disk I/O.
Additionally, the Intel system has RedHat's latest kernel -- 2.6.9-34.ELsmp x86-64. The AMD is setting one RedHat update package back -- 2.6.9-22 ELsmp x86_64.
It is NOT a fair comparison.
"Fujitsu Siemens PRIMERGY RX220 S1, AMD Opteron 280 processor Rock Web Server v1.3.3 (x86_64), Apache Tomcat 5.5.9
vs
IBM System x3650(3.0 GHz Intel Xeon 5160) Rock Web Server v1.3.3 (x86-64), Apache Tomcat 5.5.9
both 4 cores 2 sockets
Webperformance
Opteron: 8394
new Xenon: 9182
the Xenon is 9% faster. Should be available from 25 August. So Intel is closing the gap for some time."
Hey, guy. That is Opteron 280 at 2.5G. Should we try some thing faster like Opteron 284 at 2.8G (maybe in AM2 socket).
BTW. those Intel boys are funny. Even before the duel starts, they already have some many excuses for the result. Why can't they get the best guy to in charge of the Intel Box and make sure it is its best performance state.
Webperformance
Opteron: 8394
new Xenon: 9182
Clearly, the 2.4GHZ Opteron 280 is close to the 3GHZ Woodcrest. But it seems that the test was on JSP performance. Most web sites are not using JSP. Our tests will be more relevant, we should go to netcraft.com and see what's the most popular technologies and perform tests based on that.
One question: Do you really think even AMD takes you seriously?
One question: Do you really think even AMD takes you seriously?
I got emails from some Intel people commenting on this journal. Intel and AMD folks receive my weekly briefing. Willingly or unwillingly, some of my strategies were being adopted by Intel, or at least, I projected the moves of them. Such as
1) Intel to establish an uniform pricing scheme
2) DELL would spit at Intel and go AMD by the end of 2Q06
3) AMD should have a dual core sempron
Intel failed to follow my advice to
1) ramp down production and hike prices
As a result, Intel will pay a dear price.
What's yet to be proven off my journal:
1) DELL will go AMD across the board around the time of Conroe launch
2) Intel will suffer GAAP loss in 2Q06
3) Intel will suffer operating losses from 3Q06 onward
Anonymous said...
is done already... See spec.org.
Fujitsu Siemens PRIMERGY RX220 S1, AMD Opteron 280 processor Rock Web Server v1.3.3 (x86_64), Apache Tomcat 5.5.9
vs
IBM System x3650(3.0 GHz Intel Xeon 5160) Rock Web Server v1.3.3 (x86-64), Apache Tomcat 5.5.9
both 4 cores 2 sockets
Webperformance
Opteron: 8394
new Xenon: 9182
the Xenon is 9% faster. Should be available from 25 August. So Intel is closing the gap for some time.
Yeah, that would be true IF the AMD Opteron 280, a 2.4 GHz processor, were AMD's top chip.
But it isn't. If you're going to compare an Opteron server to an Intel server based on A CHIP THAT ISN'T EVEN AVAILABLE YET, you should at least pick the best AMD chip available TODAY, the 2.6 GHz 285.
As a side-note, just search "woodcrest opteron benchmark" in Google and see what comes up.
Good idea, hope you get it done.
"Yeah, that would be true IF the AMD Opteron 280, a 2.4 GHz processor, were AMD's top chip.
But it isn't. If you're going to compare an Opteron server to an Intel server based on A CHIP THAT ISN'T EVEN AVAILABLE YET, you should at least pick the best AMD chip available TODAY, the 2.6 GHz 285."
I believe the point was to compare it to an equivalently priced chip not the top of the line. So in price performance Xeon wins.
"I believe the point was to compare it to an equivalently priced chip not the top of the line. So in price performance Xeon wins."
Haha... That is not a duel, that is a price comparision. If that is what you want, get yourself Sempron and shut up. We are talking a duel of performance here.
Intel fanboys can’t believe anyone would want to have a fair test with rules, witnesses, multiple formats and online verifications. They are victims of Intel’s own corporate mind set. They Prefer Ignorant Super Spin or “P.I.S.S.” They have been at war with the laws of thermal dynamics for years and hate the thought of another disappointment. I personally feel the Woody will win a fair share of the tests.
"I believe the point was to compare it to an equivalently priced chip not the top of the line. So in price performance Xeon wins."
This is only true when taking into account only the processors. Something people tend to forget is that Xeon DDR2 FB-DIMM memory is 50% more expensive than Opteron's registered DDR400 memory. So actually a top-of-the-line Opteron 285 ($1050) with 4 GB of registered DDR400 ($400) is as expensive as a top-of-the-line Xeon 5160 ($850) with 4 GB of DDR2 FB-DIMM ($600).
I believe the point was to compare it to an equivalently priced chip not the top of the line. So in price performance Xeon wins.
http://www.monarchcomputer.com/Merchant2/merchant.mv?Screen=PROD&Store_Code=M&Product_Code=120303&AFFIL=FRG&NR=1
Guess not.
Heres the real problem, where can you actually buy a woodcrest server? When AMD announces RHT and dell goes AMD all around its the beginning of the end for intel. Conroe is the best they can do? Trust me with RHT .65nm new fabs and K8L intel loses bigtime.
There is no doubt that Wouldcrust is better than that suckage of the ages, Pentium 4.
And there is no doubt that AMD's next generation offerings will be better than Wouldcrust.
Intel and their endless pumpers, paid and not, keep up with the fraudulent comparisons because they know they have made progress, but they are still at the little kids table.
Sharikou,
You did not trust Intel when they showed bench mark comparisons of their Woodcrest vs the Opteron.. What makes you think that an AMD extremest like you would be trusted to be fair in banchmarking?
Sorry body, you lost fairness credibility long time ago.. Good scheme though in trying to get free servers!
One more thing, deleting Intel fanboys comments will definetly hurt your blog! It just shows that you are not very open minded like other bloggers.. Imagine how you feel if the Voodoopc blog decided to block your comments!
What makes you think that an AMD extremest like you would be trusted to be fair in banchmarking?
Intel boys are scared of the open duel I proposed. This has made it more interesting to see what Intel guerillas are hiding from us...
You're going to find it hard finding such a setup until Intel release the chips officially and you fund it yourself.
Why would a company want to send out chips to you for testing if you know how to beat them?
Would send Intel into the crapper.
sharikou , sir i came across a zdnet review where u actually talked about TDP . AMD calculates TDP as the V(max) X I(max) . while intel sort of takes the typical values . Sir , leaving the server segment and going to the mobile segment , it is of common knowledge that AMD based notebooks become very hot (very very very hot !!)comapared to their intel centrino counterparts and the woodcrest in many ways has derived it's architecture from the dothan class of architecture (which also is true for centrino) . Therefore don't you think woodcrest processors will have a decent advantage over AMD opteron processors .
nyx, when amd release it's future stuff, Intel fanboys will shout about nehalem, even now, from time to time I here them saying: It will be fun to watch how nehalem will beat everything what amd has to ofer at that moment [[;
it is very funny to here those things [[;
also they don't understand sharikou's devotion to talk badly about Intel. Who can blame him???
The other day I saw the ad on the net saying, that woodcrest is 60% faster than the stuff from the other guys(AMD). Seriously, what are they thinking? Even those so called pre-release benchmarks don't show that advantage ;/
In my opinion woodcrest is faster than opteron, and conroe will be faster than a64, but not by much. Surely, let's wait for nehalem :D
as far as I remember nehalem was supposed to be that 10ghz chip [[[[[;
tho what do I know, maybe woodcrest is really 60% faster than opty?nice [[;
even if it were true(which it isn't) woodcrest never be so appealing as opty for me. It doesn't look serious if it cannot scale as proper server chip.
"I admire your neverending quest to shed light on the ignorance presented by those who actually believe INTEL has something viable to compete with, but in the end some might just not want to face the truth."
So how long have you two been dating?
I'm planing on a benchmark dual Core-2 duo e6400 vs an AM2 athlon 64 x2 4600. the systems are price compatable with 2GB of ram and WinXP and linux benchmarking. the systems are going to be about 1K a Peice, another reson i'm doing it is for an sinance experament at school. sharkarou my email is atbusbook@aol.com .
Post a Comment
<< Home