Folding Forum

Wow got some catching up to do, nice pacing ( and articles ).

Looking good!

toTOW wrote:First review is finally here ... it's not a revolutionary piece of hardware, but we'll use your feedbacks to improve future work : 9800GTX+ hardware review.

Interesting little read. Be interested to see more on the increasing RAM speed, tried giving it a nudge on my 9800GTs and seemed to give a little more, no effect on my GTX 260's that I could notice.. How much should one give them?
kiore.

Kiore please post before and after ppd just like in the article, I don't think the 'general consensis' is wrong, and believe the author of this article might be judging a broader issue with a to narrow data set to justify giving any conclusions.

Ram speed should not matter was not only an observers pov, but also one semi handed down by the developers. Latency was more important then raw mhz in some cases ( hw specific, specific wu's? ) was an argument I heard more then once and there should be some data on this on the forum I believe.

There has been more testing done showing memory to be much less effective than shader oc, and I don't believe this will ever change due to how the hw works and how the software needs to implement things.

I'm trying to remember what we said before about RAM overclocking, but I think I've always advised that memory overclocking gave too less benefit compared to the issue it can rise ... most memory chips used on video card are near their maximum clock capacities, so very often, overclocking them gave more instabilities than performance improvements ... I guess that's why we usually advice to not overclock memory.

Yeah that two, mem doesn't really produce errors at x speed indicating it's not advised to run it there, it can run at a to high frequency and can die much more suddenly then a gpgpu core.

I been doing other things but if this really has as many doubts as it seems when it's being adressed by an FA article it might be prudent to offer some more testing, not only with memory speed but also latency. I could use nibitor and try to tighten my stock timings abit and see the impact. Usually tightning timings will not damage hw, just prevent xxxx clock to be reached.

Not entirely true about most ram being at their max from the factory, it all depends on the manufacturer. Some gso's for instance where/are fitted with the same ram ic's as 8800gt's just less volts, and upping volts gave a real nice speed increase ( though those cards also failed quick as I heard, probably same batch ic's but lower binned hence the pairing with the gso's and it's slower clock? ).

I think ihaque also said things about ram speed ( in regards to folding performance ) in his thread about the g80 cuda ram tester?

From what I got in quick tests on the GTX+, performance increase from 1100 to 1200 and also from 1200 to 1300 ... and at 1400 I got artifacts and immediate Unstable machine errors.

I'll try to confirm that in the next review planned around Christmas.

toTOW wrote:From what I got in quick tests on the GTX+, performance increase from 1100 to 1200 and also from 1200 to 1300 ... and at 1400 I got artifacts and immediate Unstable machine errors.

I'll try to confirm that in the next review planned around Christmas.

Why mention arti's? Shouldn't you test it with the stand alone memory checker

Or occt's implementation? Or, both?

You got arti's in furrmark? Atitool? Game?

Easier to test with dedicated tools no?

I got artfacts in Windows/VNC ... it was empiric tests : I choose arbitrary frequencies to quickly release this "dummy" review ... and I didn't push the hardware to the maximum as stated in the review ... I guess it's not necessary to start a long chat about this in the current review, as I intentionally did something simple and quick.

Next reviews will be more complete, with measures at various frequencies, and with the search of the highest stable frequencies for FAH.

Hehe sorry I just like pushing people to work harder ( eg. give me the info I would like to see asap )

Will be waiting on the review that's for sure, wouldn't mind running some tests if you specify a formula to use for consequent testing results!

Latest news :

The Protomol core unveiled in Advmethods

New project for nVidia GPU2: p10101

Statistics system webpage optimisations

Folding@Home in 2010

Server went down at 7:45 UTC ... an intervention is planned ... I don't know what's wrong with the server.

edit 10:23 UTC : server went back 30 minutes ago

Motherboard died and has been replaced

I signed up last Sunday, but it never sent me an acknowledgment. I went ahead just now and tried to sign up again. It said the entered username is already used !, the entered e-mail is already used !
So, they are entered. Have been in there a week. They just don't work. Never received an email from FAH-Addict.
David

Did you look in your spam folder ? The mail might have arrived here by mistake ...

If you don't find it, I found your account, so I'll approve it manually.

I did not find it in my spam folder, so if your could manually approve it, that would be great.

This is the message I am getting now when I try to log in:

Warning: You still have %d attempt(s) remaining after that you will have to wait 5 minutes to obtain 2 new attempts (10mins for 5) !

David

Done.

That's still weird that you didn't get the mail ...