Gpu2 Client Is Here!!!! This Is Not A Joke!!

[Marvin_The_Martian]

What I am wondering, I haven't read the official forums so maybe I'm so far off track I can't be saved no more, but seeing the low temps and all it seems like the cards aren't really working hard. I don't have an r6xx gpu but do they have 2d clocks still? Could it be their not engaging those because I can't believe load temps into the 50s if it's really flexing it's muscels :S

[MikeB12]

I read 3d is forced... kinda like sse is on smp...

but, yeah, I agree the 97 point wu's are not pushing the cards at all. even ocing doesn't increase frame times. so it's almost like these wu's are beta or dummy wu's; not really processing anything, just going through the motions and counting steps. otherwise oc's would improve frame times and the temps would be climbing and hitting load temps.

[sc00p]

Maybe it's then my time to make pro-motivation post and encourage you all keep it going with this very first open beta test phase, and for god's sake report problems at official forums (oh well could stretch that to any client). It's very important I think and thanx 7im that you're keeping eye on us
So I do not have anything else to do since don't own any ati card

Fold on.

[Kunaak]

Just an Observation outta curiousity.

the current ATI cards run at much much slower MHz when in 2D mode.
meaning, if the 3d switch isnt flipped, you can run as many 3D Mhz as you like, but your probably still running in 2D mode.
which for ATI is usually around 300 mhz.

you might try forcing 3D Mode to work, by monitoring your MHz with say Rivatuner.

see if that explains why some people are getting such low points.

of course, this is just a theory, since I dont actually know how this is working.

[[XC] DragonOrta]

When I OC in either ATI OverDrive or in Rivatuner, Rivatuner reads my clocks as the ones I set the OC as when F@H is running.

[[XC] riptide]

Just an Observation outta curiousity.

the current ATI cards run at much much slower MHz when in 2D mode.
meaning, if the 3d switch isnt flipped, you can run as many 3D Mhz as you like, but your probably still running in 2D mode.
which for ATI is usually around 300 mhz.

you might try forcing 3D Mode to work, by monitoring your MHz with say Rivatuner.

see if that explains why some people are getting such low points.

of course, this is just a theory, since I dont actually know how this is working.

Good point Kunaak. How about GPUz?? Also I notice that on same volts a severe hike in clocks ALONE will put the temps on the core up a little. Immediately. Thats the Acid test if you will.

[Marvin_The_Martian]

Good point Kunaak. How about GPUz?? Also I notice that on same volts a severe hike in clocks ALONE will put the temps on the core up a little. Immediately. Thats the Acid test if you will.

Yeah same thing I said

But DragonOrtha already answerd

[Jaivan]

Are you guys using Cat 8.3 or 8.3 hotfix?

[Lurker246]

Hi guys,

Here's an early run on my sig rig. Cat 8.3 (No hotfix) Vista X64 SP1. I've HR-03's on the 2900's and they rarely go over 50c when gaming so 44,45c seems right. Q6600 @ 40% load. Azureus likes to swallow 1.5GB of RAM on me so the RAM usage isn't caused by the client.

[[XC] riptide]

Right. Lurkers clocks are 3Ghz. I reckon if you got them up, that time would change between steps./

PS: What part of Ireland are you from?

[Lurker246]

Right. Lurkers clocks are 3Ghz. I reckon if you got them up, that time would change between steps./

PS: What part of Ireland are you from?

I wish I could up the clocks, anything over 333 FSB causes it to become unstable. Definitely not stable enough for folding. Curse this 975x chipset I'm in Dublin BTW.

I've just noticed that it saves the logfile to C:\Users\<user>\Appdata\Roaming\Folding@home-gpu. - That's FAHmon back up and running for me. Now to disable crossfire and run 2 gpu clients.

Edit: 2 clients up and running, it's only running 2d clocks apparently. No increase in 3d clock usage according to Riva.

[sc00p]

@Lurker: have you tried some 1.4 volts for GO and run mems 1:1? ~380 fsb depends of course your cooling cpu&nb

[Lurker246]

@Lurker: have you tried some 1.4 volts for GO and run mems 1:1? ~380 fsb depends of course your cooling cpu&nb

Yeah, I've tried just about every bios/setting/voltage combo possible. This board doesn't like quads running over 333fsb. I can bench at around 360 x 9 1:1 but as I said anything above 333 is unstable, especially for folding . Core2's can reach 400-420fsb on the mobo. Though I think I'd be better off upgrading my mobo to an X48 rather than slapping an E8500 in my current board.

[Stijndp]

I think an x38 with your current DDR2 will be sufficient :p

[Mechromancer]

I get more PPD when I set the FahCore_11.exe affininty to a SNGLE CORE. Unfortunately you have to redo that at the beginning of every work unit (an affinity changer program for that would be handy). Try it out and see what type of PPD increase you get.

[Lurker246]

I think an x38 with your current DDR2 will be sufficient :p

Asus Rampage X48 is ready to roll with DDR2

Looks like I'm staring down the barrell of a CPU bottlekneck. The default 2d clock on my cards is 507Mhz. I overclocked the 2d clocks with AMD GPU tool to 750Mhz and it didn't make a bit of difference to PPD.

Also can't use ATI tool on Vista x64, it's seriously fooked. The gpu client gets an EUE if I force 3d clocks in ATI tool with Vista- It's crashed my system a couple of times too.

This Beta certainly looks promising though... seen this on the folding forums

http://foldingforum.org/viewtopic.php?f=10&t=2020

toTOW wrote:
Number from HD3870.

Q6600 @ 2.4 GHz (266 * 9), GPU @ 780 MHz : 1310 PPD
Q6600 @ 3.4 GHz (485 * 7), GPU @ 780 MHz : 1860 PPD
Q6600 @ 3.4 GHz (485 * 7), GPU @ 850 MHz : 1860 PPD

More numbers with the 2900XT ...

Dual Opteron 2212 @ 2.7 GHz : 1350 PPD (with GPU @ 739 and 847 MHz)
Q6600 @ 3.4 GHz (485 * 7), GPU @ 739 and 847 MHz : 1860 PPD
Q6600 @ 3.8 GHz (475 * 8), GPU @ 739 MHz : 1860 PPD
Q6600 @ 3.8 GHz (475 * 8), GPU @ 847 MHz : 2100 PPD
Q6600 @ 3.8 GHz (475 * 8), GPU @ 860 MHz : 2100 PPD

[Bigchrome]

Using cat 8.3s on XP SP2 with an E6850 @4ghz I get roughly 2090ppd from my 2900XT set to 820/945 in ati tool

[Marvin_The_Martian]

I wish I could up the clocks, anything over 333 FSB causes it to become unstable. Definitely not stable enough for folding. Curse this 975x chipset I'm in Dublin BTW.

I've just noticed that it saves the logfile to C:\Users\<user>\Appdata\Roaming\Folding@home-gpu. - That's FAHmon back up and running for me. Now to disable crossfire and run 2 gpu clients.

Edit: 2 clients up and running, it's only running 2d clocks apparently. No increase in 3d clock usage according to Riva.

I was right lol

I get more PPD when I set the FahCore_11.exe affininty to a SNGLE CORE. Unfortunately you have to redo that at the beginning of every work unit (an affinity changer program for that would be handy). Try it out and see what type of PPD increase you get.

I'll try and put something together.

Had a more optimistic message here but

The code looked simple but I'm getting permission errors here when trying it on my smp client so that will happen with fahcore_11 to I guess.

[Kunaak]

Edit: 2 clients up and running, it's only running 2d clocks apparently. No increase in 3d clock usage according to Riva.

thats exactly what I thought was happening.

since these arent 3d applications, they run the same speed as the desktop.
so they arent getting thier full potential.

things like GPUZ arent good for varifying core speeds.
neither is catalyst control center, or ATITool.
you need something that does that in real time, which is why I suggest rivatuner.

there are various methods for overclocking ATI Hardware.
but few I can recall well for the 2900xt.

[[XC] DragonOrta]

I've been using Rivatuner to OC my 3850 and it keeps 2d and 3d clocks the same for me.

[beard]

I just flashed higher 2d (slightly higher then default high 3d) clocks into my 3870x2. Of course now my pc consumes more power in desktop use even without gpu client.

[Lurker246]

Hi folks,

Been running 18 hours stable with 2 gpu clients and a console client. It's putting out 1935 PPD, which isn't too bad I think for Vista 64.

I left the gpu client running for a few hours with 2d clocks @ 700Mhz but it doesn't make a blind bit of difference to PPD.

Hopefully someone out there is smart enough to get 3d clocks running in Vista 64 and the PPD should hopefully improve a bit.

[linflas]

I crunch on wcg on my q6600's but both my rigs have radeon 3XXX series cards, 2 x3870's and one 3850, how will running this affect my WCG points as it is so cpu dependent?

[sc00p]

I crunch on wcg on my q6600's but both my rigs have radeon 3XXX series cards, 2 x3870's and one 3850, how will running this affect my WCG points as it is so cpu dependent?

I guess your fears will become true, but then again you would receive f@h points too?

[Marvin_The_Martian]

I've been using Rivatuner to OC my 3850 and it keeps 2d and 3d clocks the same for me.

So do you have some clock speeds and frametimes?

I just flashed higher 2d (slightly higher then default high 3d) clocks into my 3870x2. Of course now my pc consumes more power in desktop use even without gpu client.

Did it improve frame times?

Hi folks,

Been running 18 hours stable with 2 gpu clients and a console client. It's putting out 1935 PPD, which isn't too bad I think for Vista 64.

I left the gpu client running for a few hours with 2d clocks @ 700Mhz but it doesn't make a blind bit of difference to PPD.

Hopefully someone out there is smart enough to get 3d clocks running in Vista 64 and the PPD should hopefully improve a bit.

Read the posts above, flash 2d clocks the same as 3d clocks or use RT to set them the same, and please post your frametimes before/after and maybe temps before/after?

I crunch on wcg on my q6600's but both my rigs have radeon 3XXX series cards, 2 x3870's and one 3850, how will running this affect my WCG points as it is so cpu dependent?

Gpu client needs a dedicated cpu core, so wcg and fah gpu client might work but both projects will take a big hit as supposed to running them seperatly.