First DDR SETI benchmark!

[sciencewhiz]

AMDZONE has a review of a 1.2ghz athlon on a DDR mobo and only showed a 3% decrease in SETI WU times

See the details here: http://www.amdzone.com/articleview.cfm?articleid=418&page=6.

Notice the RC5 and OGR benchmarks also, the DDR system is slightly slower than a SDR system

 

[TuffGuy]

sweet!!! thanks sciencewhiz.

 

[CurtOien]

sciencewhiz,

That really sucks.
I was hoping for better than that.

Thanks for the info.

 

[JWMiddleton]

I agree CO, I had figured that there would be more of a difference. Well, at least it is not enough to cause me to buy! I was really worried that DDR would obselete my stuff.

They may be able to tweak more out of it with driver/BIOS improvements as it is pre-release.

 

[cory]

i bet it would have had more of a difference with
a 2.04 client then the 3.0.

 

[artemedes]

That is disappointing, but actually makes me feel better. I just ordered a t-bird and a msi pro2 on saturday. Then found our last night that the DDR boards would be here soon. I was thinking of canceling my order. Not anymore, I'll just be happy with what I am getting. It is replacing a k6, so I should be much better off.

 

[BoberFett]

It's possible that there will be a more pronounced difference with the Duron, since it has a significantly smaller cache than the Thunderbird.

 

[blade47]

I've always heard that ddr has more latency than sdr which may be affecting the seti times alot. I don't know but it's just a thought.

 

[cory]

no i think ddr has pretty good latency.
you can get 2cas ddr ram.
they used 2.5 cas ddr ram for that ddr seti test.
ddr ram has much better power usage than rambus or even
normal sdram i think.

rambus has a very bad latency. in fact
when you add rambus modules the latency increases.

 

[cory]

one thing to remember is that most of these benchmarks
are running on the amd reference board.

i am sure once asus and abit and the others are shipping boards
and improving the bios, that the times will change.
it went from 4.29 hours to 4.12 hours with ddr
(and that was with 2.5 cas ddr)
(the speed differences from a good bios can really be big)

 

[Assimilator1]

Also don't forget that v3.0 seems to be less dependant on memory bus ,it seems that 256k cache is enough for SETI to run in ,as far we can tell at the mo.

Anyone running Xeons on v3.0?

 

[office boy]

I have to say that question the validity of this benchmark due to the lack of details given.
The time is really very slow. And no talk about using the same WU, or even a standard WU.

 

[Sukhoi]

Well, maybe future boards will be faster.

 

[artemedes]

After reading the article, I have to agree with Office Boy. The review lists times for seti in SECONDS. The numbers listed don't tell me how long an average WU would take or anything. Makes me think they did not even finish a single WU, but just ran it long enough to get an idea of the speed. We all know how the speed can vary throughout the WU. I think I will wait until someone can give a better description of how the tested, and get tests for boards you can actually buy before I trust the seti statistics for the DDR boards.

 

[BadThad]

Assim1 - I read that the new client uses 4% more memory. That's on the seti website. Wouldn't that make it MORE dependent on memory bus?

 

[cory]

it uses more ram but has a smaller
working set so it fits in the cache better.

 

[sciencewhiz]

Here is the link to the program AMDZONE used for their benchmark. http://130.113.184.129/sp/seti.htm

It isn't an indication of how long a particular workunit will take but a much more general comparison.

Hopefully in the coming days there will be more definitive evidence as to whether DDR helps SETI alot or not.