![]() |
|
|
#276 |
|
∂2ω=0
Sep 2002
República de California
22×2,939 Posts |
@RE: you're right, I was checking latest used-R7 prices on Amazon, saw "Radeon Pro" and the $1499 price and early-August, and assumed it was the Pro VII, figuring if anyone could get it cheaper and earlier than other retailes, it would be mighty Amazon.
Thx also for the clarification re. the liquid-cooled one. |
|
|
|
|
|
#277 |
|
"TF79LL86GIMPS96gpu17"
Mar 2017
US midwest
11110100011112 Posts |
Despairing of finding Radeon VIIs at acceptable prices used, or available at all new except by the $1000+ scalper crowd, I've begun to look at Vega 64 etc. Vega 64 are being bid up above $300 used on eBay, which seems a premium above their GhzD/day abilities.
Last fiddled with by kriesel on 2020-08-02 at 16:53 |
|
|
|
|
|
#278 |
|
"Viliam Furík"
Jul 2018
Martin, Slovakia
2·401 Posts |
What's the expected increase in performance when moving from Windows 10 to Linux with ROCm, in percents?
|
|
|
|
|
|
#279 |
|
"TF79LL86GIMPS96gpu17"
Mar 2017
US midwest
172178 Posts |
Needs testing on same system, clock rates, fft, and assignment, but a rough estimate is at 5M fft, George Woltman's 510 GhzD/day linux vs my 480 on Win10, ~6%. I haven't seen any posted figures elsewhere. George has reported being able to run 1200Mhz memory clock. I think my results were from somewhere in 1100-1160. Other gpu tuning should also be documented and comparable; gpu clock, fan curve, power limit etc. Please make your own comparison tests and let us know what you find.
Last fiddled with by kriesel on 2020-09-28 at 17:41 |
|
|
|
|
|
#280 |
|
Mar 2022
Earth
2×32×7 Posts |
Picked up a Radeon Pro VII for ~$400.... excited to put it to work!
|
|
|
|
|
|
#281 |
|
P90 years forever!
Aug 2002
Yeehaw, FL
17×487 Posts |
|
|
|
|
|
|
#282 |
|
Mar 2022
Earth
2×32×7 Posts |
|
|
|
|
|
|
#283 |
|
"Mihai Preda"
Apr 2015
26548 Posts |
Other GPUs from the RadeonVII family, just as old, that with a bit of luck may be had for cheaper are "Radeon Instinct" MI60 and MI50. They are about as powerful as R7, maybe a bit faster even. They have 32GB of RAM (MI50 also has a 16GB variant) and full ECC.
The main drawback is that they have no fan! (but they absolutely need one) But with a bit of tinkering a fan contraption can be added to one end of the GPU, with some sort of adapter. (also needed: fan, and fan power supply e.g. from the motherboard) For GpuOwl ATM we don't need either the large VRAM (32GB) or the ECC. Maybe in the future. The ECC is useful for the parts of computation that are not error-checked, which as an example would be second-stage P-1 (which BTW GpuOwl does not run anymore). PS: If anybody is trying to boot such GPUs (MI50, MI60) in Linux, you may need to add the "pci=realloc" kernel argument if the GPU is not initialized properly. |
|
|
|
![]() |
| Thread Tools | |
Similar Threads
|
||||
| Thread | Thread Starter | Forum | Replies | Last Post |
| Vega 20 announced with 7.64 TFlops of FP64 | M344587487 | GPU Computing | 4 | 2018-11-08 16:56 |
| GTX 1180 Mars Volta consumer card specs leaked | tServo | GPU Computing | 20 | 2018-06-24 08:04 |
| RX Vega performance | xx005fs | GPU Computing | 5 | 2018-01-17 00:22 |
| Radeon Pro Duo | 0PolarBearsHere | GPU Computing | 0 | 2016-03-15 01:32 |
| AMD Radeon R9 295X2 | firejuggler | GPU Computing | 33 | 2014-09-03 21:42 |