Woodcrest officially fragged: performance 10-20% slower than Opteron
Performance, Price/Performance, Performance/watt
In March, I projected that Woodcrest should be 10-20% slower than Opteron in server performance. Now GamePC has done some Apache web server* benchmark for Woodcrest 5140, and Woodcrest is a dog. The Woodcrest 5140 is 14% slower than Opteron 280. In fact, Woodcrest 5140 is slightly slower than Opteron 270 (2GHZ).
Unlike those review sites which live on ad money, GamePC is a system vendor that sells servers, workstations, desktops and notebooks, both AMD and Intel. In fact. you can pre-order Woodcrest CPUs from GamePC.com. GamePC's customers include big names such as Dolby, Hitachi, Lockheed Martin, FAA.
The Woodcrest 5140 is to be priced at $470. Opteron 270 is available for $441.
The key here is 64 bit server performance. Woodcrest is simply no match to Opteron. Now, you understand why Intel had to pre-arrange benchmarks for people to push buttons.
In 64 bit file compression/decompression tests under Linux, a 2.6GHZ Woodcrest is 40% slower than a 2.4GHZ Opteron. In 64 bit ScienceMark MolDyn test, the Opteron 275 (2.2GHZ) is 31% faster than Woodcrest Xeon 5140.
According to this report, Conroe/Woodcrest's enhancements such ops-fusion won't work under 64 bit.
As for power consumption, Woodcrest is getting close to socket 940 Opteron levels. The Woodcrest 5140 system consumes 23 watts more than the Opteron 270. I noticed that the test systems only had 2GB of RAM, so the extra power introduced by FB-DIMMs was small. In a fully loaded system, the Woodcrest system will definitely consume more power.
*Apache runs 61% of world's web sites (Microsoft IIS runs 29%). Apache is like the swiss-knife of web servers, it has a highly sophisticated design that allows almost infinite features and flexibility.
posted by Sharikou, Ph. D. @ 6/25/2006 09:12:00 AM
49 Comments:
dear shakirou,
it seems you missed the conclusion of the test you quoted:
... We’re only seeing a first glimpse here, too. The Xeon 5140 and 5130 processors we tested are the mid-range $300-$500 Xeon variants, whereas Intel will be rolling out the big-gun 5150 (2.66 GHz / $850) and 5160 (3.0 GHz / $1,000+) models in the next few weeks. From the numbers we’ve seen thus far, these high-end models should be able to re-capture the workstation/server performance crown for Intel, and will put AMD’s Opteron on the defensive in a major way...
it seems your march statement
In March, I projected that Woodcrest should be 10-20% slower than Opteron in server performance.
is true, if you compare top-range amd with mid-range woodcrest. not that impressive...
In March, I projected that Woodcrest should be 10-20% slower than Opteron in server performance.
The only meaningful comparison is at the same clockspeed. That was what I projected. Intel claims that their top speed will be 3GHZ. AMD's top speed is now 2.8GHZ. 3GHZ is only 7% clockspeed increase from 2.8GHZ, not enough to overcome the 10-20% lag in performance.
Also, remember that Rev F will boost Opteron performance by about 1-10% at the same clock.
The only meaningful comparison is at the same clockspeed.
do you compare amd opteron and itanium that way?
Also, remember that Rev F will boost Opteron performance by about 1-10% at the same clock.
i agree, that this could be possible. but it will take some time for ref f to arrive.
The only meaningful comparison is at the same clockspeed.
The point is this: Woodcrest and Opteron are expected to have very similar clockspeeds. Unlike Itanium which is only capable of running at 1.6GHZ. We will have 3GHZ Woodcrest and 3GHZ Opteron, 2.6GHZ Woodcrest and 2.6GHZ Opteron...
From the numbers we’ve seen thus far, these high-end models should be able to re-capture the workstation/server performance crown for Intel, and will put AMD’s Opteron on the defensive in a major way
Nowadays, high end workstations are armed with 4 Opteron processors, total 8 cores. Woodcrest is not in that market, because it is only capable of 2P.
"Nowadays, high end workstations are armed with 4 Opteron processors"
Are you quoting a market research study or just making generic statements? Can you tell us how the Opteron workstation sales are split between 2P and 4P?
"Woodcrest is not in that market, because it is only capable of 2P."
I think you are refering to the Blackford chipset. Can you tell us for a fact than no OEM will use Woodcrest above 2P?
The only meaningful comparison is at the same clockspeed.
This is completely ridiculous. Why stop at clockspeed? It's totally unfair that Woodcrest has 4MB cache, let's cut that down.
Look, the market cares about price and performance. Now come back with something that shows Opteron has the lead.
I don't really see why the only meaningful comparison should be clockspeed, even if the clockspeeds of Conroe and K8 are pretty similar. Seems to me that all comparisons should practically come down to performance/price and performance/watt. Both of which AMD is beating Intel in at the moment, and it looks like... in the future as well. Also, AMD will has greater availability (now), and that should continue during some time after the initial launch of these new top-range Conroes.
Nowadays, high end workstations are armed with 4 Opteron processors, total 8 cores. Woodcrest is not in that market, because it is only capable of 2P.
your definition of "highend workstation" and the definition of say, hp or fujitsu siemens seems to differ. their highend workstations have to processors. which manufacturer sells quad processor workstations?
The point is this: Woodcrest and Opteron are expected to have very similar clockspeeds. Unlike Itanium which is only capable of running at 1.6GHZ. We will have 3GHZ Woodcrest and 3GHZ Opteron, 2.6GHZ Woodcrest and 2.6GHZ Opteron...
a clock to clock comparison makes no sense in my opinion just because "woodcrest and opteron are expected to use the same clockspeeds". In my opinion it makes more sense to compare the top processors, because they are the best, what the manufacturers are able to produce.
Bravo. Nice job finding the AMD silver lining in every cloud.
In only very few tests does the Woodcrest at 2.33GHz fall behind the Opty 2.4GHz. In most tests, the Woodcrest 2.33GHz is faster than even the Opty 2.6GHz.
And, the Woodcrest 2.33GHz system consumes less power than the Opteron 2.2GHz system at load.
In 1P and 2P segments, the Woodcrest is going to be untouchable till K8L. Except in Apache; or very few other areas.
So, people wanting to use Apache will use Opterons, while for almost ALL other uses, the Woodcrest is better; price-wise, performance-wise and power-wise.
Great; so the Opterons run Apache very well.
Look at Encoding performance. Even the 2GHz Woodcrest is faster than the FASTEST Opeteron, at 2.6GHz in the tests...
Great; so the Opterons run Apache very well.
Look at Encoding performance. Even the 2GHz Woodcrest is faster than the FASTEST Opeteron, at 2.6GHz in the tests...
I think 64 bit is the key to true AMD64 performance. Those photoshop stuff are 32 bit, and Opteron is basically handicapped. Today, more and more server application are moving to 64 bit. AMD64 has no competition there.
If you need 64 bit server performance, Opteron is the answer. If you wants to do HPC scientific stuff, Opteron is the answer. If you wants to do media encoding, in my opinion, the only real answer will be a Torrenza card which boosts performance 100x. Today, general CPUs are all too slow for encoding video -- unless you goes 4P.
Why stop at clockspeed? It's totally unfair that Woodcrest has 4MB cache, let's cut that down.
When I projected Woodcrest performance compared to Opteron, I was assuming both have about the same clockspeed. We know Woodcrest has 4MB, and Opteron has 1MB cache. Of course, if Intel can bring out a higher clock Woodcrest, good for them. But it's more likely AMD will bring out a higher clocked Opteron as it moves to 65nm.
Give Intel credit for the Woodcrest improvements.
But if AMD's claim about its quads will have the same wattage ranges as its present day duals, then future duals should have even better performance/watt. And if AMD can ramp production successfully, then the numbers will improve for the peformance/watt/price too.
"Update: OVH contacted us to inform us that it had been mistaken and Woodcrest had not tested, but Dempsey, the generation of processor not being built on new Core architecture, but old Xeon architecture. One thus includes/understands well better the results obtained. The tables were updated with the true data, and this time the scores are without call!"
The Woodcrest is a killer!
Woodcrest love Integer.
Compression = Integer.
You need to update your title now:
Opteron officially fragged...
Update: OVH contacted us to inform us that it had been mistaken and Woodcrest had not tested, but Dempsey
Did you see that 3.73GHZ Xeon 5080 there in GamePC's benchmarks? That's the Dempsey Intel launched just a couple weeks ago. But Dempsey 3.73GHZ was fragged by Opteron 265 (1.8GHZ). Right now, Dempsey is being sold at $199 on newegg.com, but few wants that 170watt beast.
Is socket F any good for rev G CPU'S?
Yes. I think K8L will be socket F too.
If your clients want servers now, just choose whatever is the best. If you wait, you will never get anything. There is always a better chip later.
"Nowadays, high end workstations are armed with 4 Opteron processors, total 8 cores. Woodcrest is not in that market, because it is only capable of 2P"
The 4P market is viable option now thankx to AMD with IM numa, HT and cost.
Previously, no one in their right mind would pay for 10% improvment in thru-put going from 2P to 4P with Intel let alone going with dual core P's thanks to FSB effects.
2P is yesterdays market.
Servers are about memory speed, FPU speed, data thru-put and scalability
dude.. its not cool to just spam someone's wall if you disagree with them.
"The only meaningful comparison is at the same clockspeed."
the comparison is meaningful ONLY if Core and K8 have similar microarchitecture, which they're not in this case.
you should do the comparison based on pricing, since performance/price is what consumers really care about.
"dude.. its not cool to just spam someone's wall if you disagree with them."
lol, I don't how he managed to hack Shariku's blog...
>>the comparison is meaningful ONLY if Core and K8 have similar microarchitecture, which they're not in this case.
To see who has a better microarchitecture, comparing on the same clock is the best bet. e.g. if 3GHz ABC vs 3GHz XYZ, and XYZ having a better score, we could know XYZ have a better microarchitecture.
>>Eat shit and die AMD fanboy! Woodcrest crushes Opteron!
>>>>dude.. its not cool to just spam someone's wall if you disagree with them.
These tell us the maturity of an Intel fanboy.
lol, I don't how he managed to hack Shariku's blog...
You think those Intel fabnois are smart enough to hack a web site? If they are, they won't be dancing for Intel's crap and fraud.
I approved the posts just let him expose his kind...
What's funny about that article is the underlying Intel fanboy remarks the author makes, such as on the power consumption "Woodcrests new TDP of 65w means it will be undercutting AMD and taking back the performance per watt crown" (paraphrased). Uhm...is this guy for real? Does he really think 65w is full load? Last time I checked, an Intel TDP was when the CPU was at "normal load" (try 50% and you're near).
Isn't AMD making a Quad-Core CPU w/ 95w TDP and less? What? Clovertown and Kentsfield are over 100w for idle? Oh no! AMD lost the PPW lead of 50w over Intel's load for their idle! It's now 49w! The world is ending! (laughs at stupid Intel fanboys)
Mad Mod Mike, the article showed actual power consumption numbers, and Intel was slightly better than AMD in that respect. Also, could you provide some links to back up your TDP claims?
AMD and Intel would both have already developed quad core cpus. Intel at least have and one was exposed in a quad SLI setup however, for various reasons benchmarking results weren't posted.
It was posted on www.coolaler.com
Just need to keep in mind that software still hasn't been utilised for 2 cores, let alone 4 and 8.
Servers of the high end which use this stuff use relatively different software too.
Any consumer could get their hands on this stuff as well, but for $10,000+ for a computer? I think not.
Mad Mod Mike, the article showed actual power consumption numbers, and Intel was slightly better than AMD in that respect. Also, could you provide some links to back up your TDP claims?
he can't. but that will not stop him posting nonsense...
Sharikou,
Do you think the inquirer has been compromised by Intel. I see a lot less AMD ads and more of Intel ads. What is your take on this?
Do you think the inquirer has been compromised by Intel.
I don't think so. But, in general, any for ptofit IT news site can be easily compromised. All IT sites will go up and down, they first look good and just, then got bought off by struggling companies because it's cheap to buy them off, then they go down. Then another will come...
Have you seen press lately?
Woodcrest rocks! (anandtech)
Woddcrest is God! (sako)
AMD doomed for the year (pcblitz)
I think you should acknowledge the fact that Intel will turn the tide now.
"To see who has a better microarchitecture, comparing on the same clock is the best bet. e.g. if 3GHz ABC vs 3GHz XYZ, and XYZ having a better score, we could know XYZ have a better microarchitecture."
not necessarily true. performance/price and performance/ watt are more suitable for comparison. according to most benchmarks done to date, woodcrest have shown an advantage over opteron in both of these fields. even if woodcrest has higher clockspeed, it consumes less power than opteron.
"Mad Mod Mike, the article showed actual power consumption numbers, and Intel was slightly better than AMD in that respect. Also, could you provide some links to back up your TDP claims?"
Ask any respectable IT individual who's had experience in the business and you'll know the truth and Intel themselves have stated that it's more correc to label the TDP at nominal use since most CPU's aren't 100% all the time...which is the stupidest statement I have ever heard.
I also bookmarked an article where the author gives numbers and proof about the TDP of Intel CPU's, cept I just reformatted 3 hours ago so I'll head back to google and find it for you, not that your fanboy @$$ would even believe it.
"I think you should acknowledge the fact that Intel will turn the tide now."
I think you should acknowledge the fact they all got their information from the same machine setup by Intel, that, and the fact they're all paid off by Intel.
Slightly off track but this is what I think of the whole Woodcrest/Conroe VS Opteron/AMD64 fiasco...
(Warning you may disagree)
---------------------
Possible AM2/SKT-F platform updates over the next 1-12mths.
> Reverse-HT (who knows what effect this may have assuming it even exists/works?)
> Rev G (possible minor IPC improvements?)
> Improved 65nm process (lower power + higher clock speeds)
> 4 X 4 (Multi socket enthusiast platform)
> Shared L3 cache? (bigger caches?)
> K8L (big IPC boost?)
> Z-RAM maybe? (5x dense cache? uber big L3 caches?)
> HTX expansion slots for co-processors? (Torrenza)
> Probably other cool stuff too like bells & whistles etc... lol
---------------------
My point?...
Should AMD be scared of Conroe/WC? maybe, maybe not? I have no idea... I just know that I won't be jumping straight onto the Conroe bandwagon/hype until I at least see what AMD's next card is.
Something to seriously consider is that Opteron was released more than three years ago (April 2003) i.e. if Intel's latest & greatest 65nm chip wonder wasn't able to match or beat current 90nm Opterons (particularly concidering Conroe/WC has a huge 4mb L2 cache) I'd be pretty worried...
Now stop and actually think about that last paragraph guy's...1,2,3
Woodcrest/Conroe are good no doubt, but dang, three years is a long, long, loooong time for AMD to be sitting still wouldn't you think?
....food for thought.
"Can you Intel fanbois slamming Mad Mod Mike link some articles to back up your claims that he's full of it? If not, then shut up. Seriously."
just to make things clear, i'm not an intel fanboy.
i've posted numerous sources to confirm that MMM has nothing but BS.
when he argued about how AMD was @ disadvantage from using 5-5-5 timing RAM, i gave him pcper's review. it changed nothing
when he argued about how Conroe XE's score is lower than E6600, i proved that he mistakened XE with an ordinary E6600.
i even argued on sharikou's blog, that clockspeeds cannot be a clear representative for comparing the different models from AMD and Intel. performance / watt and performance / price are much clearer representatives, and they both are more accepted in this world than clockspeeds.
are my results posted? no. are my postings even being looked at before deleting? i doubt it. i'm not saying MMM has nothing but BS, but i seriously recommend him to do some spelling and fact checks before posing his blogs.
no one gets banned from a notable, well-respected forum for no reason.
"Have you seen press lately?
Woodcrest rocks! (anandtech)
Woddcrest is God! (sako)
AMD doomed for the year (pcblitz)
I think you should acknowledge the fact that Intel will turn the tide now."
Haha, that proves nothing other than what Intel's best at, MARKETING.("paid off" marketing more like it for Anandtech I'm sure) Nothing more nothing less. As Sharikou has already stated, 64-bit is the future, Woodcrest/Conroe/Merom is a 32-bit hold over from a fading era in computing that Intel is fast clinging to. Once 64-bit Vista is here, it's all over but the crying for Intel, for they know they will get thier ass handed to them across the board when it comes to performance. 32-bit performance will quickly become moot when this happens as it already has for most server operators anyway. I expect Intel will once again be sluggish to respond(more like caught with it's pant's down) to it as well and end up right back in the same situation during the K8 vs. P4 days.
AMD64 > EM64T
AMD64 > IA32
The only performance advantage Intel has currently in the 32-bit realm will fade faster than your whites thrown in with the load of colored clothes. Rev. F/G and K8L will be here faster than most people think, Opteron's advantage at 4P and up is here to stay, indefinately. Enjoy your brief respite in the sun while you can Intel fanboys, it's going to get ugly for them next year when 64-bit Vista arrives, that you can count on.
"AMD64 > EM64T"
of course, at least this is the fact for now.
however, according to hkepc, Fx-62 only gets an average boost of 2~3% under 64bit, where intel gets none. putting that into real world scenario, Fx-62 still cannot match the performance/price and performance/watt of NGMA. it needs to fill the gap of 5~25%, and a merely 2~3% is just not enough.
64bit won't come that quick either. Vista is 32bit compatible, and nothing is really lost by using the 32bit Vista. most programs will come in 32/64bit after Vista's release. it will take at least months to years to switch from full 32bit to full 64bit.
No macro-op fusion in 64-bit environment:
http://www.hkepc.com/bbs/itnews.php?tid=621654
The only performance advantage Intel has currently in the 32-bit realm will fade faster than your whites thrown in with the load of colored clothes. Rev. F/G and K8L will be here faster than most people think, Opteron's advantage at 4P and up is here to stay, indefinately. Enjoy your brief respite in the sun while you can Intel fanboys, it's going to get ugly for them next year when 64-bit Vista arrives, that you can count on.
i don't care what's next year, when i'm looking for hardware now. right now to the end of year it seems the intel processors are superior. if i have to buy new processors next year, i will have a look at the market again.
""Mad Mod Mike, the article showed actual power consumption numbers, and Intel was slightly better than AMD in that respect. Also, could you provide some links to back up your TDP claims?""
How about using Google?? lol
I think silentpcreview as a article about it.
Intel TDP is typical.
AMD TDP is max.
It would be funny if Intel coudn't make CPU's with typical TDP lower then AMD max TDP.
"when he argued about how Conroe XE's score is lower than E6600, i proved that he mistakened XE with an ordinary E6600."
I'm sorry but you're a moron and the only thing you've proven is that you don't know anything about anything.
"The only meaningful comparison is at the same clockspeed."
This is a nonsensical statement by Sharikou and he knows it. Doesn't Sharikou know that comparing at the same clockspeed is like comparing two cars based on wheelbase? The correct comparison for the customer is by price and when comparing comparably priced Conroe processors, Conroe exceeds the competition.
Hmmm, the chart shows Woody running at lower total power and out-benching Opteron in most of the tests. I can't see how that is bad for Intel and good for AMD. Shall we just call Sharikou an AMD fanboi and be done with it?
"Can you Intel fanbois slamming Mad Mod Mike link some articles to back up your claims that he's full of it? If not, then shut up. Seriously."
Uhhh, dumbass, if you had read the original article that Sharikou (idiot) posted, you would see that the total power numbers are right there on page 3 of the article. SInce you didn't read the gamepc.com review, maybe you should just sit this one out until you do.
"(Warning you may disagree)"
Yes indeed. ;)
"Reverse-HT (who knows what effect this may have assuming it even exists/works?)"
Disagree. It's a rumor. No matter how likely it seems true in some's eyes, IT'S STILL A GxD DxMNED RUMOR.
"Rev G (possible minor IPC improvements?)
Improved 65nm process (lower power + higher clock speeds)"
Very likely. Should be where the major performance improvements come from.
"4 X 4 (Multi socket enthusiast platform)"
Isn't this already happening right now on AM2?
"Shared L3 cache? (bigger caches?)
K8L (big IPC boost?)"
AMD already confirmed the shared L3 cache in K8L, but it's not clear how 'big' the IPC boost would be.
"Z-RAM maybe? (5x dense cache? uber big L3 caches?)"
5x dense and 8x slower - sounds more like a on-chip main memory than a cache. It doesn't seem to apply to K8L's L3 (which is less than 4MB in size), either.
" HTX expansion slots for co-processors? (Torrenza)"
Not that soon. It takes more than a year to build up such (co-processor) market chain & space.
" Probably other cool stuff too like bells & whistles etc... lol"
and being useless, too.
Hi!
I have seen the GamePC test, very interesting.
But... The Apache test as i see, it's a 32 bit version charged on a Windows 64!!
This is what the site say for the test:
(http://www.gamepc.com/labs/view_content.asp?id=woodcrest&page=4)
Operating System Windows XP Professional x64 Edition - Default Install
Benchmarking Software
Apache Web Server 2.0.55 Win32 (!!!!!!!!)
Adobe Photoshop CS2
Alias Wavefront Maya - Version 6.5
Discreet 3D Studio Max - Version 7.0
LAME MT Multi-Threaded MP3 Encoder - Version 3.97 MMX
Macromedia Flash MX2004
Microsoft Windows Media Encoder x64 - Version 9.0
SiSoft Sandra 2005 SP3 x64 - Version 2005.10.10.69
Sciencemark 2.0 x64 - March 21st Build
If this is a 32 bit test with Apache (on Windows...), your consideration for the superior performance in 64 bit env, i think that it's not true.
Or i'm wrong with the test configuration noted on the site?
Thanks for your attention Sharikou, i like a lot your blog! :-D
Daniele from Rome - Italy
What a joke!
They have tested server software on Windows XP PRO?!
And as far as i know Windows XP is limited in user connections.
Benchmarks suck!
Bring real world scenarios, and i will take a look.
Thanks Daniele from Rome - Italy
"If this is a 32 bit test with Apache (on Windows...), your consideration for the superior performance in 64 bit env, i think that it's not true."
If that was a Win32 Apache, than it just further proves how much better Opteron64 is than Woodcrest. If that's Win32, expect Win64 to further increase AMD's lead over Woodcrest.
Quote: "If that was a Win32 Apache, than it just further proves how much better Opteron64 is than Woodcrest. If that's Win32, expect Win64 to further increase AMD's lead over Woodcrest. "
And i agree. But Sharikou said that this was a test to prove the 64bit superior performance of the AMD64, and that's not true, simply.
Me too i suppose that the 64bit architecture of the AMD is far superior, but i have to admit that this test on gamePC sucks a little bit to prove this point, and in the other tests the Woodcrest was superior.
To be trusted, we have to said always the true, i think.
Friendly ;-D
Daniele from Rome - Italy
Post a Comment
<< Home