Just curious... why doesn't the Radeon VII show up in this list?

https://milkyway.cs.rpi.edu/milkyway/gpu_list.php

---

Might not get exposed by the AMD API.

From recent comments on the AMD headers being included into the latest Linux kernels, AMD has not had a consistent naming structure for its gpus.

The Radeon VII might be listed as something different than its commercial name.

---

That's what I though Keith.
Surely it must be this then - AMD FirePro S9150 as the top card.
Doesn't really matter.... I was just curious. The page isn't that helpful if it can't list the correct GPU's. PG do a great job of that.
Thx

---

If I remember correctly when you look at the stderr.txt output of one of these Radeon VII cards, BOINC identifies it as a gfx11?something.
Not anything close to the popular product name like Radeon VII.
All that is coming from the AMD API call that the gpu_detect module does in the client.

---

Its listed now.

noob question: according to this list, why, for example, AMD Radeon Pro 580X (1.0) is faster than a AMD Radeon VII (0.506)? According to techpowerup, Pro 580X has 5.5 TFLOPS in FP32 and 345.6 GFLOPS in FP64 (https://www.techpowerup.com/gpu-specs/radeon-pro-580x.c3398), whereas Radeon VII has 13.44 TFLOPS in FP32 and 3.360 TFLOPS in FP64 (https://www.techpowerup.com/gpu-specs/radeon-vii.c3358). Am I missing something here?

noob question: according to this list, why, for example, AMD Radeon Pro 580X (1.0) is faster than a AMD Radeon VII (0.506)? According to techpowerup, Pro 580X has 5.5 TFLOPS in FP32 and 345.6 GFLOPS in FP64 (https://www.techpowerup.com/gpu-specs/radeon-pro-580x.c3398), whereas Radeon VII has 13.44 TFLOPS in FP32 and 3.360 TFLOPS in FP64 (https://www.techpowerup.com/gpu-specs/radeon-vii.c3358). Am I missing something here?

I am puzzled, too. I made myself a list of GPUs and theyr FP64 and FP32 computer power. Looking at the list, I find a lot of "weak FP64" cards at the top.

For example:

1. Place: Radeon PRO 580X has 5,5 TFLOPS FP32 and 345 TFLOPS FP64 .
15. Place: Radeon VII (13.4 TFLOPS FP32 / 3.3 TFLOPS FP64)

It almost seems that the list is summing up the total amount of GPUs of a specific type - if a lot more people use GCN4 Cards than Radeon VII, the total compute power in the grid can be higher.

If the list is meant to represent "fastest cards for MW@H, I am missing something obvious here..

noob question: according to this list, why, for example, AMD Radeon Pro 580X (1.0) is faster than a AMD Radeon VII (0.506)? According to techpowerup, Pro 580X has 5.5 TFLOPS in FP32 and 345.6 GFLOPS in FP64 (https://www.techpowerup.com/gpu-specs/radeon-pro-580x.c3398), whereas Radeon VII has 13.44 TFLOPS in FP32 and 3.360 TFLOPS in FP64 (https://www.techpowerup.com/gpu-specs/radeon-vii.c3358). Am I missing something here?

I am puzzled, too. I made myself a list of GPUs and theyr FP64 and FP32 computer power. Looking at the list, I find a lot of "weak FP64" cards at the top.

For example:

1. Place: Radeon PRO 580X has 5,5 TFLOPS FP32 and 345 TFLOPS FP64 .
15. Place: Radeon VII (13.4 TFLOPS FP32 / 3.3 TFLOPS FP64)

It almost seems that the list is summing up the total amount of GPUs of a specific type - if a lot more people use GCN4 Cards than Radeon VII, the total compute power in the grid can be higher.

If the list is meant to represent "fastest cards for MW@H, I am missing something obvious here..

With todays post from Tom the Admin this is probably moot

Thanx for mentioning the shutdown of the seperation tasks.....

I was just about to get more FP64 capable GPUs - seems I don´t have to any more :)

I have been reading the whole announcement thread, trying to find out if there are any other projects that would benefit from FP64 GPUs, but I can´t get a clear picture - seems FP64 is obsolete nowadays?

I loved the idea of putting old gear to good use and it was somewhat "sexy" to have a otherwise pretty useless old GPU contribute to something meaningfull, because it is good in FP64. For FP32/16 tasks, the situation is simpler, but way less fun: Simply get the biggest modern GPU...

Thanx for mentioning the shutdown of the seperation tasks.....

I was just about to get more FP64 capable GPUs - seems I don´t have to any more :)

I have been reading the whole announcement thread, trying to find out if there are any other projects that would benefit from FP64 GPUs, but I can´t get a clear picture - seems FP64 is obsolete nowadays?

I loved the idea of putting old gear to good use and it was somewhat "sexy" to have a otherwise pretty useless old GPU contribute to something meaningfull, because it is good in FP64. For FP32/16 tasks, the situation is simpler, but way less fun: Simply get the biggest modern GPU...

The good news is that Nvidia gpu's are fairly cheap right now with the 8gb 4060 Ti coming out at the end of the month for $299US

The good news is that Nvidia gpu's are fairly cheap right now with the 8gb 4060 Ti coming out at the end of the month for $299US

Probably typo. Should be non-Ti ;)

yes, for others this is an option - my intent was more like: "give old gear a new purpose" - sustainability thinking..
Of course, new GPUs still can share the FP32 compute power.

The good news is that Nvidia gpu's are fairly cheap right now with the 8gb 4060 Ti coming out at the end of the month for $299US

Probably typo. Should be non-Ti ;)

Yes it was, here's the latest headline:
Nvidia launches three new GeForce RTX 4060 GPUs, starting at $300
RTX 4060 Ti 16GB is $499, 4060 Ti 8GB is $399, and the vanilla RTX 4060 will be $300