GeForce RTX 3080 and 3090 support enabled !

[gunnarre]

1) Let me see how many platforms I have access to

Does this mean that the upcoming core also solves the problem where FAH couldn't cope with multiple OpenCL platforms on the same computer - e.g. if you had both an Nvidia and AMD card in the same computer, or you had both ROCm and AMDGPUPro OpenCL installed?

[bruce]

I do not know whether that problem has been fully solved or not, but I do know it's being worked on.

[PantherX]

[...Does this mean that the upcoming core also solves the problem where FAH couldn't cope with multiple OpenCL platforms on the same computer...

Great question! Assuming you have access to such a system, once FahCore_22 version 0.0.13 is released to FULL, you can give it a shot and see what happens. I would personally consider these edge cases which can be very hard to replicate in the lab due to hardware constraints. However, if you're willing to test it out and provide feedback, we might be able to support it in future.

[foldy]

multiple openCL platforms should be the default when CPU has integrated GPU too and a dedicated GPU is also attached.

[PantherX]

I am not sure if multiple OpenCL platforms can be used simultaneously. However, I do know that GROMACS does support GPU off-loading when performing CPU tasks. Let's wait and see what happens with FahCore_22 and FahCore_a8.

[robertmiles]

[snip]

Maybe not I was thinking the same thing I posted a link to here maybe he will answer what projects/PPD

says he can never post here on the forum says authorization needed, here is a log on one..

Has he noticed that a SEPARATE account must be created for the forum before any posts become possible?

index.php

[flarbear]

I just noticed my stats for my 2080Ti showing >4m PPD for the past 2 days on extreemeoverclocking. The PPD were 4.5m yesterday, 5m for the entry just added in the past few minutes for today and the core it is running now is estimating 4.7m PPD at over 50% completion.

The current work unit is a new 0.0.13 core_22. I didn't scour my logs for stats on the cores, but my first 0.0.13 core downloaded on the 24th and now my log is full of them (180 occurences according to grep|wc).

(I mention this because these were close to the figures that were predicted in an earlier post for a 3080, but I'm seeing them on a 2080Ti. Most of mine are project 13426, though. Wondering what the newer cards will get now that 0.0.13 is out there...?)

will be very interested to see what 0.0.13 can do with a 3080

Some quick numbers from Project 11765 in Linux:

TPF 73s - GTX 1080Ti running OpenCL/ 1.554 M PPD
TPF 57s - GTX 1080Ti running CUDA / 2.253 M PPD
TPF 49s - RTX 2080Ti running OpenCL/ 2.826 M PPD
TPF 39s - RTX 2080Ti running CUDA / 3.981 M PPD
TPF 36s - RTX 3080 running OpenCL / 4.489 M PPD
TPF 31s - RTX 3080 running CUDA / 5.618 M PPD

I do expect that the numbers might potentially be better once the drivers have matured a bit, generally in about 6 months. By that time, we might have a new version of FahCore_22 that can unlock more performance too!

[PantherX]

The figures I posted were from a traditional F@H Project where CUDA brought in 15% to 30% improvement (reduction in TPF). However, for the Moonshot WUs, that can range from 50% to 400% and Project 13426 is a Moonshot Project.

EDIT, here's data for 3090 too:
Project 11765 on Linux
TPF: 73s - 1080 Ti OpenCL
TPF: 57s - 1080 Ti CUDA
TPF: 49s - 2080 Ti OpenCL
TPF: 39s - 2080 Ti CUDA
TPF: 36s - 3080 OpenCL
TPF: 31s - 3080 CUDA
TPF: 30s - 3090 OpenCL
TPF: 26s - 3090 CUDA

[toTOW]

All known Ampere GPUs have been added to GPUs.txt :

Code: Select all

2204	GA102 [GeForce RTX 3090]	
2206	GA102 [GeForce RTX 3080 10GB / 20GB]	
222b	GA102 [GeForce RTX 3080 Ti Engineering Sample]	
222f	GA102 [GeForce RTX 3080 11GB Engineering Sample]	
2230	GA102GL [Quadro RTX 48GB]	
223f	GA102GL	
2482	GA104 [GeForce RTX 3070 Ti]	
2484	GA104 [GeForce RTX 3070]	
2486	GA104 [GeForce RTX 3060 Ti]	
249c	GA104M [GeForce RTX 3070 Mobile]	
249d	GA104M [GeForce RTX 3070 Mobile / Max-Q]	
24af	GA104 [GeForce RTX 3070 Engineering Sample]	
24bf	GA104 [GeForce RTX 3070 Engineering Sample]	
252f	GA106 [GeForce RTX 3060 Engineering Sample]

[HaloJones]

interesting confirmation of the TI models

[toTOW]

2230 / GA102GL [Quadro RTX 48GB] updated with commercial name : 2230 / GA102GL [RTX A6000]

[MeeLee]

Anyone running a large atom WU on a 3090 and on a 4 to 5Ghz CPU, would want to run HWMonitor (in Windows) or HTOP in Linux, to see how much percent the GPU takes of one CPU core? (If you have a multi core CPU, perhaps it's best to disable HT/SMT in the bios, and only run between 2 to 4 cores. Should make determining the max CPU frequency easier).

[JohnG__]

What could we expect from the 3060Ti?

[kiore]

What could we expect from the 3060Ti?

Hard to say looks by cores that it could be similar to a RTX 2070/80 but lots of variables, a promising new member of the elusive 3000 series family.

[ajm]

What could we expect from the 3060Ti?

First results can be seen here: https://folding.lar.systems/gpu_ppd/bra ... tx_3060_ti