Folding Forum

i currently have a 5700XT, actually 2 but one for my gaming pc.

the Vega64 has 4096 cores and 791 Gflops
the 5700XT has 2560 cores and 609 Gflops

which one would you guys get ? because i could trade my 5700XT for the Vega64

Guy is selling the Vega64 for 400$ CAD and i bought the 5700XT for 400$ CAD

That is a difficult question. Looking at pure FP32 TFLOPS the Vega 64 should have the edge. However, if the numbers from this https://docs.google.com/spreadsheets/d/ ... tput=html# are accurate, then it seems like there is some factor beyond just TFLOPS. Keep in mind that that is self-reported data with only 1 sample for the 5700 XT...

Personally, I would go with the Vega 64 for the TFLOPS numbers, but at this point I can't tell you for certain what performs better for FAH. Whatever you do, consider submitting some samples to the database to help people in the future viewtopic.php?f=38&t=34510

@r0k0:
FAH uses mixed precision calculations which means that it uses FP32 most of the time with a bit FP64.
Therefore FP32 performances are more important for folding, while your numbers show FP64 performances.
Still, 5700 XT shows also lower FP32 theoretical performance compared to Vega 64.

As you own a 5700XT already you know how much PPD it makes. I don't have proper numbers, however I expect it to be >1MPPD

Most charts show 800kPPD for the Vega 64, however those numbers are based mostly on Core21 numbers. Currently most projects are based on Core22 which is faster by ~10%.
Therefore I expect it to be ~1MPPD.

I would assume that both perform similar with a slight benefit for 5700XT.
Additionally, TDPs of 295W for Vega 64 compared to 225W for 5700XT let assume that 5700XT is more engery / cost efficient.
--> Therefore I would opt for a 5700XT

PS.
I don't know if you use HFM.NET to log your folding results.
If you do I would appreciate if you could provide an export.
In this topic I describe that I want to create a PPD hardware database as most databases are based on older/finished projects.
viewtopic.php?f=38&t=34510
You have a 5700XT and those numbers are rarely available.

i have 2 5700XT actually lol, i dont know how to get that info from the logs. as soon as i do, i will

r0k0 wrote:i have 2 5700XT actually lol, i dont know how to get that info from the logs. as soon as i do, i will

You need HFM.NET to do it. I created a guide viewtopic.php?f=16&t=34516

Vega64 needs work units with high atom count to efficiently use all those shaders. There are work units with 25k atoms, 100k atoms or even more and Vega64 sucks on those low atom count work units. Stay with the RX 5700XT

NoMoreQuarantine wrote:

r0k0 wrote:i have 2 5700XT actually lol, i dont know how to get that info from the logs. as soon as i do, i will

You need HFM.NET to do it. I created a guide viewtopic.php?f=16&t=34516

awesome, i wil go read on it

I would choose the 5700XT too.

foldinghomealone wrote:@r0k0:
FAH uses mixed precision calculations which means that it uses FP32 most of the time with a bit FP64.
Therefore FP32 performances are more important for folding, while your numbers show FP64 performances.
Still, 5700 XT shows also lower FP32 theoretical performance compared to Vega 64.

As you own a 5700XT already you know how much PPD it makes. I don't have proper numbers, however I expect it to be >1MPPD

Most charts show 800kPPD for the Vega 64, however those numbers are based mostly on Core21 numbers. Currently most projects are based on Core22 which is faster by ~10%.
Therefore I expect it to be ~1MPPD.

I would assume that both perform similar with a slight benefit for 5700XT.
Additionally, TDPs of 295W for Vega 64 compared to 225W for 5700XT let assume that 5700XT is more engery / cost efficient.
--> Therefore I would opt for a 5700XT

PS.
I don't know if you use HFM.NET to log your folding results.
If you do I would appreciate if you could provide an export.
In this topic I describe that I want to create a PPD hardware database as most databases are based on older/finished projects.
viewtopic.php?f=38&t=34510
You have a 5700XT and those numbers are rarely available.

currently my asus rog 5700xt is pushing 1 0027000 PPD Est.

I would keep the 5700. The performance difference isn't that much. But the 2 things the 5700 has over the vega 64, is it uses less power (225W stock vs 290W on the Vega).
But what's more, the 5700XT supports PCIE 4.0, which is interesting if in the future you want to upgrade your motherboard.
The 5700XT could run from a PCIE x1 slot (at 4.0 speeds), without major performance penalties (both in gaming and folding).
The Vega 64 is limited to PCIE 3.0.

MeeLee wrote:I would keep the 5700. The performance difference isn't that much. But the 2 things the 5700 has over the vega 64, is it uses less power (225W stock vs 290W on the Vega).
But what's more, the 5700XT supports PCIE 4.0, which is interesting if in the future you want to upgrade your motherboard.
The 5700XT could run from a PCIE x1 slot (at 4.0 speeds), without major performance penalties (both in gaming and folding).
The Vega 64 is limited to PCIE 3.0.

really ? i can run a GPU out of a pcie 1x port ? huh i did not know that, do i need like a riser or i can slap that in the slot and screw it down like any regular card ?

5700xt is the way to go, much more efficient architecture does not depend on atom count lottery to get PPD. Doesn't vary much from its average (1.2m PPD).
Vega64 is either very good (1.5m PPD) or extremely bad (400-500k PPD).
Same with Radeon 7 - either exceptionally good (2m+ PPD) or exceptionally bad (600-700k PPD)

r0k0 wrote:

MeeLee wrote:I would keep the 5700. The performance difference isn't that much. But the 2 things the 5700 has over the vega 64, is it uses less power (225W stock vs 290W on the Vega).
But what's more, the 5700XT supports PCIE 4.0, which is interesting if in the future you want to upgrade your motherboard.
The 5700XT could run from a PCIE x1 slot (at 4.0 speeds), without major performance penalties (both in gaming and folding).
The Vega 64 is limited to PCIE 3.0.

really ? i can run a GPU out of a pcie 1x port ? huh i did not know that, do i need like a riser or i can slap that in the slot and screw it down like any regular card ?

You'll need:
1- An AM4 or TR4 motherboard (usually 550/570 boards), that support PCIE 4.0.
2- You'll need to run ONLY PCIE 4.0 GPUs on that motherboard
3- There is a possibility that the PCIE 4.0 compatibility only works on the upper 16x slot with the first motherboards. Since they only came out a few months ago, not a lot of info is known about if ALL their slots work at 4.0 speeds; but if they do, see #2 above if you want to run at 4.0 speeds.
4- You'll need a POWERED RISER, one end plugging into a PCIE x1 slot, the GPU fitting in the x16 slot. Both the riser and GPU need to be powered. The best risers are those that have PCIE 6-pin connectors. If you have a SATA connector, make sure it's connected to a lead that goes to the PSU. Don't use PCIE 6-pin to SATA converter cables. They might work on an empty cable, but if anything else is plugged into that lead to the PSU, chances are your GPU will fail.
7- GPUs can't be mounted in the chassis in a standard way with a PCIE riser. You'll have to find alternative mounting points for it.
8- Considering the lower PCIE bandwidth on x1 slots, and the higher bandwidth usage of Core22, it's recommended to run FAH from Linux (not windows, no VM); to have the best performance.


The issue I think is most important to know, is that not a lot is known about PCIE 4.0 x1 performance, due to their newness.
You'd be pioneering it. You might end up spending money on hardware that doesn't work or doesn't work on your motherboard.
If you do decide to go down that path, let us know your findings!

Forgot to mention, your CPU must be a Ryzen 9 3000 series, or threadripper 3000 series. No other CPUs support PCIE 4.0 yet.