Folding@Home: THGC Needs You -Team 40051

[Sakkura]

Wow, I just realized that when I migrated to a new main desktop I forgot to open up and leave this thread open in a tab. Looks like there is a lot more activity here now 😀 I should have thought to check what was up when I saw our team PPD start ramping up. So what's all the excitement these days? I saw some mentions of new hardware, AMD and NVidia problems. Anything in particular giving you guys problems? Such as slowdowns, Bad State errors, weird PPD things like that? I'll try to read back through some of what I missed too. I've been active in folding just inexplicably dropped off from this thread.

AMD problems? No such thing. 😀

Looks like a lot of people picked up GPU upgrades. Personally I "only" got an R9 290, and not one with a particularly good cooler at that. But it's still a nice upgrade over my previous 270, and it doesn't have any flaky behavior.

 

[EXT64]

That is probably right - I think the effective speed might be 6-7k and the actual clock might be 3-4k but I'm not sure.

I don't have a 980 on hand but I think some people said you had to use NV Inspector to lower their memory speeds.

Also, are you on the most recent driver?

 

[pimpanther]

Just a quick warning - with some of the new WUs (Core 21) you might see Bad State errors, even at stock, on GTX 980s and possibly other Maxwells. If you do ever see these errors somewhat often, try reducing your memory clocks. NVidia already does this for compute (dropping the 7000 effective memory speed to 6000). However it looks like it might need even more of a drop (perhaps 5800). I don't know if this is really a hardware issue or something that can be fixed in software, but just wanted to let you guys know in case you are seeing any errors.

Nice to have you back..

Nice to be back..

I have had many bad states on 0x21 wu's, thanks for the info.

Don't seem to be able to reduce my memory speed via Afterburner, maybe I need to stop folding before it will take effect.

Also my memory speed reports 3004mhz?!?

http://techarx.com/kfa2-presents-geforce-gtx-980-hof-8pack-edition/ check this out beano their is a button on the outside and when youpress it it goes into like a boost mode.,its a monster

 

[pimpanther]

Wow, I just realized that when I migrated to a new main desktop I forgot to open up and leave this thread open in a tab. Looks like there is a lot more activity here now 😀 I should have thought to check what was up when I saw our team PPD start ramping up. So what's all the excitement these days? I saw some mentions of new hardware, AMD and NVidia problems. Anything in particular giving you guys problems? Such as slowdowns, Bad State errors, weird PPD things like that? I'll try to read back through some of what I missed too. I've been active in folding just inexplicably dropped off from this thread.

AMD problems? No such thing. 😀

Looks like a lot of people picked up GPU upgrades. Personally I "only" got an R9 290, and not one with a particularly good cooler at that. But it's still a nice upgrade over my previous 270, and it doesn't have any flaky behavior.

 

[pimpanther]

nothing flaky here

 

[beanoslim]

That is probably right - I think the effective speed might be 6-7k and the actual clock might be 3-4k but I'm not sure.

I don't have a 980 on hand but I think some people said you had to use NV Inspector to lower their memory speeds.

Also, are you on the most recent driver?

Ok, I'll try it later.

I'm on the latest drivers 358.50, I had tried some others to try and solve the issue but it was the same so I stuck with the latest, its currently crunching through a 0x21 6206 with no issues.

Its the 96** WU's that I have problems with, some just crash the minute they finish downloading.

 

[beanoslim]

Actually I just tried it before I go to work.

Memory will only reduce 500mhz so as low as it will go.

Default 3505mhz / current 3005mhz, we will wait and see.

 

[beanoslim]

Squash that, the memory doesn't change, its pre-set to -500.

I'll have another look later.

 

[EXT64]

Yeah, unfortunately your rapidly outrunning my knowledge on this, but some people saw much improved results with certain projects by reducing the memory clocks. It hasn't been tested on 96xx yet.

What I do remember is that the card has different performance states. P0 (gaming speed) uses a memory speed of 3500 (7000 effective) by default. P2 (compute speed) uses a memory speed of 3000 (6000 effective) by default. Do you have the ability to change P2 memory speed? Of course there is a good chance this is dependent on card BIOS too, but I can ask the guy who got this to work exactly what he did for his card.

 

[Sakkura]

Well, GDDR5 memory has three different clock signals. The command clock signal, as well as two write clock signals that run at twice the frequency and offset. That means the data rate ends up 4 times the command clock frequency, or twice the write clock frequency. Normally the command clock is considered the "real" clock frequency of the memory, but maybe some programs report the write clock frequency for some reason.

 

[pimpanther]

my god the things you learn on this thread.[its my main thread atm and prob will be for some time] i am pimpanther on most of the oc sites but this folding room is pretty cool.

 

[pimpanther]

im doing 0x18 wu it seems,how do you get a 21

 

[Sakkura]

im doing 0x18 wu it seems,how do you get a 21

The assignment servers decide, I don't think there's anything you can do to influence it. I'm doing an 0x17 unit right now. Low PPD. 🙁

 

[pimpanther]

im doing 0x18 wu it seems,how do you get a 21

The assignment servers decide, I don't think there's anything you can do to influence it. I'm doing an 0x17 unit right now. Low PPD. 🙁

i really feel bad for beano,i do not thinkits his cards or his system i think he has just DL some moody work units.surly that is the most logicalexplination.and my god he has tried everything else.

 

[beanoslim]

Yeah, unfortunately your rapidly outrunning my knowledge on this, but some people saw much improved results with certain projects by reducing the memory clocks. It hasn't been tested on 96xx yet.

What I do remember is that the card has different performance states. P0 (gaming speed) uses a memory speed of 3500 (7000 effective) by default. P2 (compute speed) uses a memory speed of 3000 (6000 effective) by default. Do you have the ability to change P2 memory speed? Of course there is a good chance this is dependent on card BIOS too, but I can ask the guy who got this to work exactly what he did for his card.

Haha, 4 different performance options, see below

]

P2 allows me to drop the memory speed, have done so, what speed do you think is ideal??

Also while I have your attention I don't know how many of my issues you have tracked back and read about but what do you think to an FX6300 @4.5ghz effecting the folding performance on a 980 and a 980ti, CPU only reports 40-50% load with both cards running.

Also do you think it could be down clocking due to lack of psu power, in particular spilt 12v lines with 40A on each.

Cheers EXT 😉

 

[beanoslim]

Just checked log, my old friend the 96** wu

12:47:36:WU00:FS01:0x21:Saving result file log.txt
12:47:36:WU00:FS01:0x21:Folding@home Core Shutdown: BAD_WORK_UNIT
12:47:37:WARNING:WU00:FS01:FahCore returned: BAD_WORK_UNIT (114 = 0x72)
12:47:37:WU00:FS01:Sending unit results: id:00 state:SEND error:FAULTY project:9631 run:1 clone:22 gen:7 core:0x21 unit:0x00000009ab436c9b5609bee2967e32f5

 

[beanoslim]

Here we go a 9637 wu, memory set to 2500mhz, so I'm hoping that should be 5000mhz effectively.

Its funny, you can only adjust gpu clock speeds on P0 setting and only adjust memory on P2 setting?!?

 

[beanoslim]

EXT, I also had the same exact issue with th 96** wu's on the 980ti, could it be the same thing do you think?

Sorry people I seem to be taking over the whole thread atm....

 

[pimpanther]

-you got probs.thats what the thread for

 

[Mousemonkey]

EXT, I also had the same exact issue with th 96** wu's on the 980ti, could it be the same thing do you think?

Sorry people I seem to be taking over the whole thread atm....

You may not actually have a problem mate, there could be more to this than first meets the eye.

19:29:50:WU00:FS01:0x21:Folding@home Core Shutdown: BAD_WORK_UNIT
19:29:50:WARNING:WU00:FS01:FahCore returned: BAD_WORK_UNIT (114 = 0x72)
19:29:50:WU00:FS01:Sending unit results: id:00 state:SEND error:FAULTY project:9205 run:4 clone:65 gen:6 core:0x21 unit:0x00000027664f2dd055d4ca225460c2dc

I just saw that pop up on one of the 970's logs.

 

[beanoslim]

First bad state @40%, turning mem down to 2000mhz.

 

[beanoslim]

@80%, no errors since dropping memory speed.

Anyone know if running the bus width @256 bit / 192GBs vs 256 bit / 128GBs would make any difference to Folding??

That's the only real difference I can see between P0 and P2, P0 also defaults to 3500mhz mem vs P2 3004 but I've down clocked that anyway.

 

[Sakkura]

Anyone know if running the bus width @256 bit / 192GBs vs 256 bit / 128GBs would make any difference to Folding??

That's the same bus width, 256 bit? The lower bandwidth would be from lower data rate/clocks.

I don't see how you could run at lower bus width to the VRAM. That's AFAIK not configurable, it would be a very low-level driver thing if it was even possible. Unlike for example the PCIe connection, which is configurable, in fact the card will scale down that connection for power savings when idle.

 

[pimpanther]

EXT, I also had the same exact issue with th 96** wu's on the 980ti, could it be the same thing do you think?

Sorry people I seem to be taking over the whole thread atm....

You may not actually have a problem mate, there could be more to this than first meets the eye.

19:29:50:WU00:FS01:0x21:Folding@home Core Shutdown: BAD_WORK_UNIT
19:29:50:WARNING:WU00:FS01:FahCore returned: BAD_WORK_UNIT (114 = 0x72)
19:29:50:WU00:FS01:Sending unit results: id:00 state:SEND error:FAULTY project:9205 run:4 clone:65 gen:6 core:0x21 unit:0x00000027664f2dd055d4ca225460c2dc

I just saw that pop up on one of the 970's logs.

i DO NOT belive all your cards have developed a problem all at the same time.i think as mousemonkey has pointed out with that last post there may be something coming in wrong from Stanford itself that only affects Nvidia cards.what do we think of that suggestion peeps? virus?

 

[pimpanther]

JUST ASKED THE GUYS ON ANOTHER FOLDING TEAM [I KNOW I KNOW] OC UK the online shop asked if anyone had had probs with x20 WU.so lets ee