• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

ATI GPU2 - Lambda Repressor

E

EvilAlchemist

2[H]4U
Joined
Jan 11, 2008
Messages
2,732
Some good news on the ATI GPU2 Preformance

Running HD 4850 @ 675mhz.

Avg. Time / Frame : 2mn 37s - 3015.75 ppd

ATI Team is now working on getting the CPU usage down and PPD will only go up with planned optimizations.


Looking better all the time for ATI.
48XX cards are going to start to see some major PPD boosts are WU are being configured to take advantage of the added SP's.

nVidia clients don't seem to handle Lambda as well as ATI (1500+ ppd drops have been reported on some nVidia cards)... but there PPD is still higher then ATI at the current moment. until the optimizations can be implemented.

Going to start heating up here soon .....

 
That's awesome news! Nice to see some parity finally coming into play. I think there's a lot of pent up frustration out there with the ATI crowd that have patiently waited for this to come about. Nice to they're finally having some success utilizing that potential.


 
Great news! I myself have a 4870 for the killer gaming bang/buck, naturally I'd love to see it fold harder too. :cool:

 
Taken from the Folding Forums.
mhouston said:
The goal of this core was to stabilize a few issues we had picked up. It took longer than expected, hence the jump from 1.04 to 1.10. Once this core checks out fully, my stack of optimizations can start going in.
Click to expand...

I know many were getting upset with ATI because it seemed no progress was being made , but finally we have some good progress.

Keep in mind that the extra SP of the 48XX series have not been introduced, so they still have some work to do .. but it is good news all the same.

 
So, building the 4870X2x2 big brother to my current 98gx2 boxen is now (or soon to be) worth while?


Hey look, I just broke 500k!
 
WOOT! thanks EvilAlchemist for the "heads up" post :eek: I had about given up any hope of mentioning ATI and nVIDIA in the same sentence regarding anything to do with folding. (I didn't want to be laughed at for bein' a complete fool) :(

Not bein' a fan boy of either ATI or nVIDIA (well, I do like ATI better :p) I'm glad to see some competition from ATI. ;)

FOLD ON!

 
I am glad to hear that there will be some parity going forward.

 
All I can say is finally, Ati really deserves this boost after all the tech they poured into their new generation of GPUs. They earned my respect with the great pricing model and amazing gaming perfoirmance; this was the last obstacle Ati needed to overcome. :cool:

 
Indeed. I found it kinda funny how they completly caught nvidia off guard, and Nvidia had to immediatly slash prices. Competition is great and Nvidia's prices were getting out of hand.
 
Anyone running 3000-series cards? Very interested to know what to expect with their production now and in the future. TIA
 
APOLLO said:
Anyone running 3000-series cards? Very interested to know what to expect with their production now and in the future. TIA
Click to expand...

At the moment, there is no real PPD difference between the 3XXX series and 4XXX series cards.

The 3XXX series have faster GPU's so their PPD is slightly higher on the current work units.

The 4XXX series's has 800 stream processors vs the 3XXX series's 320 stream processors.

When new work units come out (Time Frame unknown), some will be designed to make sure of all the extra SP's on the 4XXX series, but right now, all the WU only use 320.


 
Currently up to ~3800 PPD on my HD4870 at 790/1100 clock speeds. Much better than the ~2200 I was at.
 
Actually, lambda WUs are in the wild and scale up on the 48XX reasonably well, but not quite 2X, but you do get more PPD on those WUs with a 48XX than you do with 38XX. Basically lambda has more atoms so it's easier to scale with the current code. Back to optimizations...
 
mhouston said:
Actually, lambda WUs are in the wild and scale up on the 48XX reasonably well, but not quite 2X, but you do get more PPD on those WUs with a 48XX than you do with 38XX. Basically lambda has more atoms so it's easier to scale with the current code. Back to optimizations...
Click to expand...

Mike Houston .. shocked to see you over here ... lol ....
Glad to know you get around ...

Good Information to have ... thanks
Since you have some "inside info" we will take your word as gospel.

And yes .. keep optimizations going!
 
I keep an eye on more things than one would think. ;-) I've been a HardOCP forum member for awhile now, more than twice as long as you. :-P I'm more of a lurker though. However, since I've gotten many emails from the [H]orde, I thought that I'd chime in since I had said lambda would run better on ATI as compared to the smaller WUs and would also scale up better on 48XX and that's now coming to fruition.

Sadly, getting stability back after the v1.04 core and moving into production took longer than everyone would have liked and put optimization work on hold. We are gearing back up, but remember that it will be awhile before optimizations make it public since things have to be thoroughly tested in beta before they go out to the public. Also note that the 1.10 core has a slight performance regression for the smaller WUs that we are also looking into.
 
mhouston said:
I ;-) I've been a HardOCP forum member for awhile now, more than twice as long as you. :-P I'm more of a lurker though.
Click to expand...

Hey now ....
I went out at purchased another 4870 today, just for you to help test the new cores.
+ I think I sent you some bios info ... so maybe we are about even ....

Seriously, thanks for all your hard work. I know it was frustrating during 1.05 -1.09.

 
I was just teasing. ;-) I pop in and out regularly. And yes, thank you for more information as it helps narrow down the problem as we hunt for an exact reproduction case and having more folks providing as much detail as possible as we test new things is really useful.
 
One more thing Mike, and I know the community wants to know as well.

In the future, will it be the core that can make use of the 48xx series extra sp's or specific work units or a combination of both?
 
Combination of both. We will naturally do better with larger proteins, as you are seeing now. Performance of small to mid sized proteins will relate more to core updates and optimizations. The core is the only thing I can control directly, so that is where we spend our time. What work units are sent out is up to the Folding@Home guys.
 
Sssshhhh! Less posting, more working on making my HD4870 put out 6000+ PPD! ;)

Thanks for your dedication, Mike. We're all eagerly (and at least somewhat patiently) waiting for more improvements to come.
 
My 4870 is still putting out 2223PPD. Must not have gotten one of the new WU yet. Oh well, I bought it for the gaming performance, not the PPD.
 
My 4850 shows 3300 PPD
Mike did say over on the home forum hes has a pile of optimizations to put in as soon as this new 1.10 core get the nod from stanford.
 
Does ATI card let you keep Xfire on when your folding?

I know you have to take Sli off with Nvidia cards.

 
So is ATI catching up to Nvidia ?

I had read GTX 260 were doing almost 3x PPD then 4870s.
 
Performance as compared to Nvidia will depend on WUs. Currently, I don't believe Nvidia is doing lambdas, but I expect we close a bunch of the performance gap, at least until Scott from Nvidia takes another optimization pass. But, at least both cores (ATI and Nvidia) seem to be stable, so now it's a battle of wills and resources for Scott and I to battle it out in optimizations. ;-)
 
Kendrak said:
You need to grab FAHMon or FAH Spy

Or you can check your stats @ http://www.hardfolding.com/index.php

Click to expand...

The points on FAHmon are way off when comparing to EOC, it seems to get worse with Faster frame times, FAH Spy is even farther off, I mean my FAhmon showed 40,000PPD but EOC only showed 30,000 when I have everything running, Stanford did make some changes to the uploading and download to make it faster so the amount of error,may be way less now,than it was a few weeks Ago !
 
Given enough time for EOC to average properly, I have had under a 5% (or so) variance in comparison to FAHMon.

EOC does need a week with no changes to the farm.

 
Me to on SMP work ect,but with the GPU , I have 6 8800Gt's it was different.

I always go to team user list at EOC and it shows the points today, I wasnt talking avg.
 
You must log in or register to reply here.
Back
Top