what is the current state of gpu folding?

jebo_4jc

[H]ard|DCer of the Month - April 2011
Joined
Apr 8, 2005
Messages
14,566
I am apparently out of the loop. Word on the street is that qrb is back in public beta. What is the ppd?

I also heard AMD cards are viable again. I would love to switch back.

Tell me everything.
 
I am apparently out of the loop. Word on the street is that qrb is back in public beta. What is the ppd?

I also heard AMD cards are viable again. I would love to switch back.

Tell me everything.

MMMM I'm so glad I stopped by, my 5870 has been lonely for SO long! I need to warm it up with some folding! I assume the new gui client will be the one to go download? I've been rolling the old school client SMP forever. I'm probly leaving loads of perts on the table between my 2600k and 5870 not folding the new shizzle.
 
There is a new core using openCL that works well on AMD cards. fahcore 17. Download the latest v7 client and put on the beta flag. my 7970 is showing 30-40k ppd at stock clocks
Link for details
http://folding.typepad.com/news/2013/03/introducing-foldinghome-core-17-gpu-zeta-core.html

and look! a couple of you [H] guys got a shout-out!
Extra special thanks to our testers:
Jesse_V, k1wi, artoar_11, bollix, ChelseaOilman, bruce, Demonfang, Grandpa_01, EXT64, Flaschie, HayesK, jimerickson, mmonnin_, P5-133XL, Patriot, rhavern, sc0tty8, SodaAnt, SolidSteel144, EvilPenguin, art_l_j_PlanetAMD64, thews, cam51037, Pin, muziqaz, baz657, PantherX, QINSP, Schro, and hootis.
 
Gonna try this out on my 7950..I noticed that this new cored runs on OpenCL, so Nvidia hardware (fermi and newer) is fully supported...anyone tried it out yet? I am trying to decide if I should hang onto my 480 (since it already has a full cover block) and fold on both..
 
The [H] guys listed there participated in the internal beta testing of the new core. Because we're [H]ard like that ;-)
 
Are NVidia or AMD GPUs going to be the hot ticket item for QRB coming up here?
 
From what I'm seeing, it sounds like the 7970 outperforms the 680 by a good margin as a result of the 680's comparatively neutered compute performance. I assume Titan would kick ass, but it isn't supported by the core due to an OpenCL driver update being required (per the linked blog post). I'm not sure what the word on the higher-end Fermi's (580) is.

40k ppd for $350-400 isn't terrible these days, might be a "good" excuse to get myself a second 7970 for crossfire action :D

EDIT: This from EOC forums

johnerz said:
QRB is out on beta for both AMD and NV, it's a unified core 17, on open CL - that means that AMD now outshines NV in a very big way

a 7970 is almost double the ppd of a 680 - the 680 is now producing high 20's K ppd the 7970 high 40's k ppd



Good luck and its about time the red had its day (he says folding on 2 green GPU's)

The 5 series is also out performing the 6 series by a lot

a 570 is about 38k ppd

a 670 is about 27k ppd

both overclocked of course
 
So any idea what we're looking at for 5870 PPD yet? I will probly set it up this evening but a baseline would be useful to know if i set it up correctly. I'd guess from activate AMD's last post i should be seeing at least 20k out of it compared to a 7970?
 
So any idea what we're looking at for 5870 PPD yet? I will probly set it up this evening but a baseline would be useful to know if i set it up correctly. I'd guess from activate AMD's last post i should be seeing at least 20k out of it compared to a 7970?

This is just a beta, so production PPD is still up in the air.

You can run the benchmark program..

http://foldingforum.org/viewtopic.php?f=38&t=23440

Which will give you an estimate (estimate will be better for AMD than NVidia) of how any card will perform, relative to another. Remember that with QRB, PPD is not linear. In the link above, a 7970 scores 2.5x higher than a 5850, so PPD would be significantly less than 7970/2.5 (because of QRB), so well under 20k. I didn't see any posts for 5870, but I'd guess low teens...shrug.
 
This is just a beta, so production PPD is still up in the air.

You can run the benchmark program..

http://foldingforum.org/viewtopic.php?f=38&t=23440

Which will give you a very accurate idea of how any card will perform, relative to another. Remember that with QRB, PPD is not linear. In the link above, a 7970 scores 2.5x higher than a 5850, so PPD would be significantly less than 7970/2.5 (because of QRB), so well under 20k. I didn't see any posts for 5870, but I'd guess low teens...shrug.

Yeah let me see if i can run the bench here during lunch. brb
 
Looks like my stock clocked 5870 is
explicit: 9.875
implicit: 73.32

So i'm just under a 6950 per that forum link. Time to turn the clocks up to 11, mmm watercooling. I think i can get close to that 6950 score. I might need to update my catalyst drivers too. Might help.
 
So judging by this, the 570 is a steal right now. Why so high on AMD if 7970 only gets ~10k ppd higher and costs twice as much as a 570? That's roughly 25% upgrade in ppd for 200% of the cost. Am I missing something?

PS. Anyone have any numbers on a 580?

Originally Posted by johnerz
7970 high 40's k ppd
570 is about 38k ppd
670 is about 27k ppd
 
this is the thread I was looking for

soon time to turn off the bitcoin miners with the ASICs coming out
 
So judging by this, the 570 is a steal right now. Why so high on AMD if 7970 only gets ~10k ppd higher and costs twice as much as a 570? That's roughly 25% upgrade in ppd for 200% of the cost. Am I missing something?

PS. Anyone have any numbers on a 580?
Not wrong. From a dedicated folding perspective that does look like a potential ppd/$ sweet spot at this time. I think AMD's performance is notable because its finally where it should be, but also that its significantly better than nV's flagship 680. I'm just happy because I chose the 7970 over the 680 for cost and availability reasons when I was looking for a gaming gpu, and that decision has paid off for folding as well.
 
Ohh goody, GPU folding right before summer time.

everything here has been running flat out for 2 summers already

the downside for me...I was kind of looking forward to actually playing a game
 
So judging by this, the 570 is a steal right now. Why so high on AMD if 7970 only gets ~10k ppd higher and costs twice as much as a 570? That's roughly 25% upgrade in ppd for 200% of the cost. Am I missing something?

PS. Anyone have any numbers on a 580?

The main difference at this point is, due to driver issues, NVidia cards require a full CPU core , whereas AMD 7 series only require 1-2% of a core. So, with AMD you can get faster SMP times, or alternatively if not folding on CPU, get less total power draw.
 
Ohh goody, GPU folding right before summer time.

Right. I can't afford to run the GPUs and cool the house in summer, so I'm trying to talk myself out of selling my 460s and 470s to trade in for 570s, 580s, or 7900s.
 
So yeah, the 7.3 client is controlled via web interface? that's a new one. Hope it works out as I expect.

I'm concerned it doesn't allow me to be as flexible. There's virtually no settings at all beyond the slider and the user/team/passkey?

edit: core is only using 40% of my gpu :( The test used 100% of it per msi afterburner but this isn't loading it up. I updated to cat 13.1, anyone got any hints?

edit2: well i'm an idiot, found the systray tool, that's helpful :)
 
Last edited:
You don't have to use the web interface. Select your folding icon in your task-bar and select "advanced control." Heck of a lot more options there.

PS. You don't even need to keep the web interface open.
 
You don't have to use the web interface. Select your folding icon in your task-bar and select "advanced control." Heck of a lot more options there.

PS. You don't even need to keep the web interface open.

Yeah I found the systray icon. I'm still concerned about the low GPU utilization. Do i need to add some client or core options?

Edit: bingo client-type = beta should do it. I'll see what it picks up tomorrow after it finishes this current one.
 
Last edited:
Yeah I found the systray icon. I'm still concerned about the low GPU utilization. Do i need to add some client or core options?

Edit: bingo client-type = beta should do it. I'll see what it picks up tomorrow after it finishes this current one.

also hit the slider all the way to "FULL THROTTLE YO" or whatever it's called, otherwise it'll only hit 100% on your gpu when your pc has been idle for XX minutes
 
That's a little bit of a debate from what I can tell. I've read several threads talking about that subject and no one seems to have an answer. Some people have fixed it by using different drivers. The only flags I have on my 560ti 448 are the "client-type"/"beta" and "next-unit-percentage"/"100" flags. Another thing to check is the slider on the main FAHControl screen. Be sure that the slider is set to FULL if you want to use all your potential power.

edit - bigted beat me by a minute :)
 
FWIW, I enabled beta-flag, woke up in the middle of the night and saw my PPD cut in about half, my CPU pegged, and the wattage from the box go from 183w-190w up to 203. I deleted the beta-flag and went back to sleep.

Dual GTX 650Ti rig running on an AMD A2Llano dual-core CPU. Normally, CPU utilization is nominal, but both CPU cores were pegged. If I have time in the next few days, I might swap in a A6 quad-core CPU and see what the difference is, but for now, I'll just have to wait and see.
 
You can see some PPD numbers from the core_17 internal beta testers at [keeps munching the URL] goo(dot)gl/fYx2d (read only).
 
agreed, my gts 450's went from 13k down to 6k overnight.
 
Any word on if the beta flag is working as intended for NV GPU's. I read this last night and threw on the beta, and 100 flag's on my 670. Noticed all 6 cores pegged to 100%. Estimated PPD was at like 40k, woke up and total PPD had dropped to below 26k. GPU+CPU.
 
lol

BTW Ted, you are showing up on my threat list now...jerk

I've got the same problem. I'm currently getting around 81k ppd with the following equipment:

i5-2500(boxen) - 12k ppd
i5-3550(boxen) - 12k ppd
i7-3770k - 29k ppd
gtx 560ti 448 - 28k ppd

EOC shows he will take me over in 3.6 days, but it also shows my average as only 40k, so I should be able to hold him off for a week or so :D
 
Sorry for the n00b questions here. Figured if I had the question someone else might too, and I spent about an hour looking around for the answer and couldn't find it here or on foldingforum (or google). Maybe I'm looking in the wrong place.

Using v7.3.6

1. I'm assuming I add the " -beta" flag to the end of my shortcut filepath, but it just breaks the link and won't run. Which file exactly am I looking to add this to? FAH Control or Client?

2. If I run in -beta mode, will my GPU receive only Core 17 WU's, or will they be mixed in with other, lower PPD units? If the latter, it may not be worth the extra power for me.

System: Intel 2500k + AMD 6870.
 
Sorry for the n00b questions here. Figured if I had the question someone else might too, and I spent about an hour looking around for the answer and couldn't find it here or on foldingforum (or google). Maybe I'm looking in the wrong place.

Using v7.3.6

1. I'm assuming I add the " -beta" flag to the end of my shortcut filepath, but it just breaks the link and won't run. Which file exactly am I looking to add this to? FAH Control or Client?

2. If I run in -beta mode, will my GPU receive only Core 17 WU's, or will they be mixed in with other, lower PPD units? If the latter, it may not be worth the extra power for me.

System: Intel 2500k + AMD 6870.


I'm not at home so I am going off of memory here, but in fah control you need to click on your GPU slot and go up to Configure. Then go to the Slots tab and select your gpu. Scroll to the bottom and select Edit. Then scroll to the bottom again and select add. Set the Name to "Client-Type" and the Value to "Beta".

There is no guarantee that when you put the Beta flag in your client that you will only receive high PPD beta units. I have received a few smaller units since adding the beta flag.

THIS is a great guide on setting up the v7 client.
 
So it looks like i'm netting about 12k more ppd running the 5870 along with the smp client. I was getting about 29k out of the 2600k previously. That TPF went up due to some cpu usage for the gpu client, but overall not a bad bump. I need to run the gpu alone for a bit to see what its total ppd is without any interruptions.
 
The main difference at this point is, due to driver issues, NVidia cards require a full CPU core , whereas AMD 7 series only require 1-2% of a core. So, with AMD you can get faster SMP times, or alternatively if not folding on CPU, get less total power draw.

Wasn't it only 6-12 months ago that this was the other way around??

I currently have a Core2Quad and Radeon 7750... a second Radeon 7750 for crossfire would be the most cost-effective upgrade for gaming... it's beginning to sound like it'll be good for folding again, too!

Yahoo!

How do you think a Radeon 7750 will perform? It's the fastest card out there that doesn't need extra power, so it may be an excellent choice for low-watt folding!

BTW - does downloading the new 7.3.6 give me this new performance? :confused:

EDIT: Hmmm... According to AMD Catalyst, only 50% of the GPU is being used, even on the "FULL" setting. I wonder what I'm doing wrong? The Core2Quad is running my 4 cores at about 90% each which is nice when still browsing the net, etc.
EDIT2: When bringing up the pretty spinning blobbies graphic, it mentions the core is 0x16... don't I need this new x17 core? How do I get/apply it??

Thanks!
 
Last edited:
Wasn't it only 6-12 months ago that this was the other way around??

I currently have a Core2Quad and Radeon 7750... a second Radeon 7750 for crossfire would be the most cost-effective upgrade for gaming... it's beginning to sound like it'll be good for folding again, too!

Yahoo!

How do you think a Radeon 7750 will perform? It's the fastest card out there that doesn't need extra power, so it may be an excellent choice for low-watt folding!

BTW - does downloading the new 7.3.6 give me this new performance? :confused:

EDIT: Hmmm... According to AMD Catalyst, only 50% of the GPU is being used, even on the "FULL" setting. I wonder what I'm doing wrong? The Core2Quad is running my 4 cores at about 90% each which is nice when still browsing the net, etc.
EDIT2: When bringing up the pretty spinning blobbies graphic, it mentions the core is 0x16... don't I need this new x17 core? How do I get/apply it??

Thanks!

Add client-type = beta to the flags under the expert options tab, client options. That allowed me to start picking up core 17.
 
Add client-type = beta to the flags under the expert options tab, client options. That allowed me to start picking up core 17.

you can also add it per slot if you don't want CPU or NV cards to pick up beta, go to "Configure" then "slots" tab, choose your ati gpu and hit Edit, scroll down, click "Add" and you'll have this screen to add the flag:

geQ3ML6.png
 
Back
Top