Vega 7 (Ryzen Cezanne 5600G) when you get a chance

[l_gel]

No rush. It's already folding, and folding quite well once I overclocked the system memory a bit. It just identifies as "Radeon TM Graphics" in the client.

I know it's getting this identification from AMD's drivers, but when it's ID'd like this I can't use it for competition since like 30 other Vega-based APUs are identified this way. I've already complained to AMD driver team about this.

If there's a way to force the the proper name of the GPU, that would be quite awesome. Hopefully I'm not creating too much work for you guys. I find you all very helpful.

Thanks.

Luke

[JimboPalmer]

Welcome to Folding@Home!

Because the CPU and GPU share a path to memory, imbedded GPUs are often too slow to fold.

If you want to try you would need to give us an exact Device ID, finding this is OS specific and you did not give us the first 100 lines of your log so we would know your OS.

Assuming it is Windows, GPUZ would show your Device ID. https://www.techpowerup.com/gpuz/

[l_gel]

Device ID from GPU-Z:

Windows 10 Pro

1002 1638 - 1002 1636

I know this won't complete too many work units per day, but the competition between my buddies has a handicap and is more weighted on overclocking. This particular iGPU is folds pretty slow at stock settings, but the Cezanne architecture has quite a lot of tuning headroom on the iGPU as well as the memory controller. It gets more PPD than my HD 7850, and at lower power consumption

[Joe_H]

There already is an entry in the GPUs.txt file for this iGPU and its device ID. At most someone may update the description text, but the iGPU is used in more than one AMD processor. So it will likely stay somewhat generic. But that will not affect folding speed or assignments, the description text is only used for display purposes.

As for comparisons with the HD 7850, your Vega 7 has about the same rating in GFLOPS. But the HD 7850 is based on a first generation GCN chip, the folding core is more efficient on later versions of that architecture.

[l_gel]

Yeah it's pretty good. Pulling over 100K ppd.

The description, yes that's what I'm talking about. It needs to be differentiated from the other Vega-based iGPUs.

[MeeLee]

Yeah it's pretty good. Pulling over 100K ppd.

The description, yes that's what I'm talking about. It needs to be differentiated from the other Vega-based iGPUs.

Seems to fall between a GT 1030 (80k PPD overclocked) and a GTX 1050 (130k PPD stock).
The 1030 uses 25-30W max, the 1050 uses about 75W.
Would be interested to see how much more your system uses on power with the IGP folding?

[l_gel]

It looks like the graphics side of the APU consumes about 50W or so while running at 2400MHz. The massively overclocked Quadro M2000 also folding in this system is consuming about 75W but gets about twice as much PPD. I've yet to tune up the CPU-side of the APU to see if the CPU cores will be worthwhile to fold. It seems like the automatic boosting behavior of of this Ryzen is a little too aggressive and will crash when folding. What I'm thinking is just set all of the CPU cores to a flat 4GHz or something at a low-moderate voltage. It's all under water using less than $75 worth of cooling parts from ebay/ali express.

[PaulTV]

You should change your profile to 'Dr. Frankenstein'

[l_gel]

lol right I made that test bench out of aluminum angle from the hardware store

[BobWilliams757]

Yeah it's pretty good. Pulling over 100K ppd.

The description, yes that's what I'm talking about. It needs to be differentiated from the other Vega-based iGPUs.

I think the descriptions for a number of the iGPU's was changed at some point. Due to the device ID, hardware being very similar, etc, they just don't all get a unique description for each identifier. Mine used to identify as a Vega 8 mobile IIRC, and now is identified as Raven Ridge [Radeon Vega Series / Radeon Vega Mobile Series], but it still folds just fine.

They are reasonably efficient for an onboard GPU. I'm pulling less average PPD on my 2400G, but work units vary quite a lot regardless. And in my case, if it meets deadlines I'll let it fold.


Folding both CPU and GPU at the same time they do take a hit, so I tend to just do one or the other. But even with my older 4 thread CPU, they churn those work units along reasonably well all things considered.

[toTOW]

I edited the GPU description : 0x1638 / Cezanne