Nvidia Pascal and AMD Polaris, starting with GTX 1080/1070, and the AMD 480
Language
Copyright © 2024 Einstein@Home. All rights reserved.
Copyright © 2024 Einstein@Home. All rights reserved.
Mumak wrote:I have just
)
Mines supposed to arrive Saturday. I also ordered one, the single fan EVGA SC 1050Ti. From what I read they can OC the cores pretty far...up to 1900mhz, and some are able to push the memory to 8Gbs. Ill run singles, duals and three units to see what fairs the best at the factory clocks, then Ill push the memory...in the past it always seemed to be the best improvement speed wise vs upping the Core speed.
I have ordered the GIGABYTE
)
I have ordered the GIGABYTE GeForce GTX 1050 Ti G1 Gaming 4G, which is dual-fan + 6pin.
Will swap a 750 Ti with it, so can make a direct comparison.
BTW, the 750 Ti currently consumes ~32W while running 2xBRP4G (~3550 secs).
-----
Some OpenCL Benchmarks see
)
Some OpenCL Benchmarks see page 5 NVIDIA GeForce GTX 1050 OpenGL/Vulkan/OpenCL Linux Performance
AgentB wrote:Some OpenCL
)
On the great majority of test cases mentioned on those pages, the 1050 just slaughters the 750. I'll be very pleasantly surprised if the advantage on Windows running Einstein is anything like that. NewEgg states my 1050 is in "packaging", but I took the cheapest shipping option, so most likely no results from me until well into next week.
Mumak wrote:I have ordered
)
Nice, I will be installing mine in a system currently with no discreet gpu. Will note the 1050Tis idle/crunching wattage use from wall.
My current stats for the
)
My current stats for the 1050Ti.
System: I5-4690K @ 4.0Ghz, Intel SSD, single EVGA 1050Ti SC single fan version, no extra power header. Running GPU tasks only. Enthoo Pro Case. EVGA 1000 P2 psu.
Power Consumption:
Idle 6.5W
Crunching 1-3 WUs 50W
Readings were taken from the wall on my UPS.
Ambient room temp:23.3C
GPU: EVGA 1050Ti SC running stock. Core Boosting to 1771mhz and staying there. Ram running @ 7000Gps. Fan 35%, temp 57C. Quiet.
BRP4G Crunch Times:
One Unit = 1500 secs.
Two Units = 2850 secs (1425)
Three Units = 4350 secs (1450)
https://einsteinathome.org/host/11997763
Thanks for the results
)
Thanks for the results Todderbert !
So a ~20% performance improvement, but for 60% more power consumption. That's disappointing.
My card should arrive in a few hours, so hope to post my results soon.
-----
Just arrived !750 Ti, 2
)
Just arrived !
750 Ti, 2 WUs:
GPU Power (reported by GPU): ~32 W
GPU Power as % of TDP (reported by GPU): ~50 %
Total machine power CPU+GPU (wattmeter at wall): ~130 W
Runtime: 3530 s
1050 Ti, 2 WUs:
GPU Power (reported by GPU): ~60 W
GPU Power as % of TDP (reported by GPU): ~50-60 %
Total machine power CPU+GPU (wattmeter at wall): ~130 W
Total machine power CPU only (wattmeter at wall): 80 W
Runtime: 2900 s
What's really odd and I can't explain yet is that the total machine power consumption (measured at wall using wattmeter) has not changed, while the difference in GPU power reported is 30 -> ~50 W. The only explanation would be that the 750 Ti power consumption is not 30, but 50 W too, but I remember that I measured this several times in the past (using power difference) and it was 30 W, which was pretty correlating with own GPU power reporting.
Host: https://einsteinathome.org/host/12435345
-----
Mumak wrote:What's really odd
)
Very odd. The 750 Ti does not have a 6 pin connector, and the 1050 does - maybe there is a difference there, the specs for the 750 Ti say maximum 60W.
AgentB wrote:Mumak
)
My normal 750Ti reports 45watt usage. My 750Ti FTW with the Six pin connector reports less but consumes the same power as my normal non six pin 750Ti.