Sign in with
Sign up | Sign in
Your question

Anandtech: New Small Woodcrest Review

Tags:
  • CPUs
  • Product
Last response: in CPUs
Share
July 13, 2006 4:27:02 AM

http://www.anandtech.com/IT/showdoc.aspx?i=2793&p=1

It's a more limited review running just 2 applications, but it's also more comprehensive since it examines both performance and performance/watt.

Also, to avoid criticism (some), they are now using a Tyan motherboard for the Opteron system with the top of the line 285. They are also using 8GB of RAM to more fully load the platforms, which should fully indicate the impact of the more power consuming FB-DIMMs. Also a RAID 0 setup was used, and they mention nothing about alleged RAID issues with Woodcrest and the Bensley platform. Also, they're using a 64-bit OS.

Some things to note, include the Woodcrest systems still be preproduction. It's also an Intel OEM reference system, it's up to you to decide whether you suspect Intel was sitting in on the benchmarking or not. Again, only DDR2 533 FB-DIMMs were used instead of DDR2 667 which means that the FSBs were not fully loaded and the latency was slightly higher.

More about : anandtech small woodcrest review

July 13, 2006 4:46:29 AM

pretty god damn impressive. for the time being, woodcrest seems to be a "no brainer" for companies who want/need bleeding edge.

i'm interested to see how it'll do in FPU intensive applications... that'll be the real test for cluster computing applications.

still, i fully expect to see a woodcrest cluster on the "top 5" list (Top 500 Supercomputers) next year.
July 13, 2006 4:54:33 AM

Quote:
http://www.anandtech.com/IT/showdoc.aspx?i=2793&p=1

It's a more limited review running just 2 applications, but it's also more comprehensive since it examines both performance and performance/watt.

Also, to avoid criticism (some), they are now using a Tyan motherboard for the Opteron system with the top of the line 285. They are also using 8GB of RAM to more fully load the platforms, which should fully indicate the impact of the more power consuming FB-DIMMs. Also a RAID 0 setup was used, and they mention nothing about alleged RAID issues with Woodcrest and the Bensley platform. Also, they're using a 64-bit OS.

Some things to note, include the Woodcrest systems still be preproduction. It's also an Intel OEM reference system, it's up to you to decide whether you suspect Intel was sitting in on the benchmarking or not. Again, only DDR2 533 FB-DIMMs were used instead of DDR2 667 which means that the FSBs were not fully loaded and the latency was slightly higher.

See they used 1T command rate for the Opteron. One less excuse for the horde. :wink:
July 13, 2006 5:47:35 AM

"Woodcrest outperforms all of the competition in 2-way configurations, and it does so with lower power consumption. This combination leads to better Performance/Watt in all tested situations. Woodcrest not only outperforms it's predecessor by over 100% in Performance/Watt, but it outperforms Opteron by as much as 60% in Performance/Watt."

Impressive
July 13, 2006 6:20:02 AM

Quote:
Just to make a quick note, and both companies are guilty of such sins, but AMD can produce a certain amount of marketing mis-direction as well as Intel:

http://www.dailytech.com/article.aspx?newsid=2971

AMD's claims --- they do not match reality.

Good one.[/Pink Panther] I was going to point that out.
July 13, 2006 6:58:51 AM

Quote:
Just to make a quick note, and both companies are guilty of such sins, but AMD can produce a certain amount of marketing mis-direction as well as Intel:

http://www.dailytech.com/article.aspx?newsid=2971

AMD's claims --- they do not match reality.


Its interesting that AMD uses TDP to determine the power consumed by Intel CPU's instead of power at the wall, which has shown that Intel uses less power than AMD. AMD also does not take into account the chipset required to communicate to the IOs.
July 13, 2006 7:29:46 AM

Quote:
Just to make a quick note, and both companies are guilty of such sins, but AMD can produce a certain amount of marketing mis-direction as well as Intel:

http://www.dailytech.com/article.aspx?newsid=2971

AMD's claims --- they do not match reality.


Its interesting that AMD uses TDP to determine the power consumed by Intel CPU's instead of power at the wall, which has shown that Intel uses less power than AMD. AMD also does not take into account the chipset required to communicate to the IOs.

Yeah, AMD is summing up the component level TDPs to report the power comparisions --- this assumes, of course, all components are operating concertedly at worst case full load conditions to add up to the total power, which of course has a very low probability of happening.

It is to there advantage to spin it this way because of the extra wattage that goes to the FB-DIMMS adds up quite quickly. The other point AMD would like to make, and is improper, is they scale up the CPU TDP to match their definition -- which is running at top voltage specs, this of course would only happen if the MB is malfunctioning :) 

The real point is --- if someone shouts "Intel spouts FUD", well AMD does it too.... it is marketing 101.

I definitely agree that both companies do it. Originally when AMD had a similar chart, they had Intel's system using 2x as many DIMMS (i think it had Intel using 8 FB-DIMMs and AMD using 4 DDR2-DIMMS).
July 13, 2006 10:05:41 AM

DUDE!!!! CANT YOU SEE THAT THE BENCHMARKS ARE RIGGED AND INTEL-TWEAKED??????????????????????????????????????????????

:lol:  :lol:  :lol: 
July 13, 2006 3:18:43 PM

ok... please let's not start this B.S. I know you're joking and being sarcastic, but this is (so far) a good thread. you've got the best minds in this forum commenting on this thread.

so let's try to leave sarcasim out of it.
July 13, 2006 3:26:25 PM

it is of my opinion that waiting 2-6 weeks after launch, even if you're excited and can't wait, is a good way to approach the coming products offered from any company.
July 13, 2006 3:45:19 PM

Quote:
it is of my opinion that waiting 2-6 weeks after launch, even if you're excited and can't wait, is a good way to approach the coming products offered from any company.
For sure, i'm never the first to buy anything. Can you imagine though, if even 1/2 the potential buyers did this? It would really hurt their bottom line. 8O
July 13, 2006 4:06:40 PM

I try to do the same but humans are prone to hypocrisy and i'm guilty of such =)

Quote:
it is of my opinion that waiting 2-6 weeks after launch, even if you're excited and can't wait, is a good way to approach the coming products offered from any company.
For sure, i'm never the first to buy anything. Can you imagine though, if even 1/2 the potential buyers did this? It would really hurt their bottom line. 8O
July 13, 2006 7:47:25 PM

AMD has been caught with their pants down right now and nothing but 95 watt QC opty's will bring their momentum back. It's nice to see an Intel server chip that is not a complete joke though :) 
July 13, 2006 8:30:01 PM

well in 2P yes its not a complete joke at all. Not many people use 2P setups though in "big business". I agree the small business server market could certainly use these machines. What I would have liked to have seen was a price comparison though of those systems.

I mean the ram alone might have pushed the Woody up a little over the opterons, and if by 20% then maybe its not worth the extra performance/watt.

meh, i'll wait for more conclusions and see why the gov thought there were issues enough to warrant not wanting to use them at this time.
July 13, 2006 8:35:08 PM

Quote:
"Woodcrest outperforms all of the competition in 2-way configurations, and it does so with lower power consumption. This combination leads to better Performance/Watt in all tested situations. Woodcrest not only outperforms it's predecessor by over 100% in Performance/Watt, but it outperforms Opteron by as much as 60% in Performance/Watt."

Impressive


I was hoping someone would mention that. 51xx scores 60 and Opteron scores 45 on Perf/watt. How is that 60% more? Hopefully it was a typo.
July 13, 2006 8:48:15 PM

actually there is one graph of the DVD store that shows the woody at 120 points and the Opties at 90...which is 30%... ;) 
July 13, 2006 9:31:12 PM

Quote:
"Woodcrest outperforms all of the competition in 2-way configurations, and it does so with lower power consumption. This combination leads to better Performance/Watt in all tested situations. Woodcrest not only outperforms it's predecessor by over 100% in Performance/Watt, but it outperforms Opteron by as much as 60% in Performance/Watt."

Impressive


I was hoping someone would mention that. 51xx scores 60 and Opteron scores 45 on Perf/watt. How is that 60% more? Hopefully it was a typo.

Well while we're at adding things we want thrown in an itanium 2 would make and interesting addition. I want to see if their mobile team inadvertently killed intel's next big crusade with their P!!! redesign.
!