I'm thinking of buying another GPU. Does anyone know the running times for the gravity wave WUs on a rtx 2060 and rtx 3050. Other models are possible but I like the relatively low power (heat) and used price for those two.
That's just the way Nvidia GPU apps work compared to AMD. Nvidia uses CPU support 100% of the time the task is running. It has something to do how the driver is made, I believe.
I'm thinking of buying another GPU. Does anyone know the running times for the gravity wave WUs on a rtx 2060 and rtx 3050. Other models are possible but I like the relatively low power (heat) and used price for those two.
Reportedly the Nvidia 8GB 4050 will be announced in the next day or so for a brand new price of $399 and in July the 16GB 4050Ti will be announce for a brand new price of $499. The 4050 will use a 12pin power plug but I don't know what the 4050Ti will use, hopefully the same one. That doesn't really fit with your "low power" thing but it might for the price.
Seems there is virtually no uplift with a 4070Ti which is odd/
The Gamma ray (and BRP) app is fairly reliant on memory bandwidth. Nvidia nerfed the memory bandwidth on all the 40-series cards. making the increase in clock speeds and IPC of the 40-series almost a moot point since they're bottlenecked by the smaller memory bus. Einstein is the most reliant on this part of the GPU. most other projects don't have much of an issue with it and see at least generational improvements.
the high invalid % is another blow to their overall production.
You have 1663 Valid tasks on that same pc, 48 is relative. Yes it's higher than most people would like but alot of people would take that 40 series card of your hand for the price of shipping and live with the 48 invalids!! They would probably do it for upto around $500 as well. IOW you have a great gpu Einstein just needs to figure out how to lower the number of Invalids this series of cards seems to be producing. Someone sending Bernd one of them might help.
I'm thinking of buying
)
I'm thinking of buying another GPU. Does anyone know the running times for the gravity wave WUs on a rtx 2060 and rtx 3050. Other models are possible but I like the relatively low power (heat) and used price for those two.
Any ideas why my my CPU time
)
Any ideas why my my CPU time on 2 seperate computers, one running a 3070Ti & the other a 4070Ti is horrendous?
4070Ti with a 3950X = CPU times range from 195 - 540sec.
The cpu time pretty much matches the gpu time. No idea why.
Running 2 & 3 tasks to test.
That's just the way Nvidia
)
That's just the way Nvidia GPU apps work compared to AMD. Nvidia uses CPU support 100% of the time the task is running. It has something to do how the driver is made, I believe.
Ben Scott wrote: I'm
)
Reportedly the Nvidia 8GB 4050 will be announced in the next day or so for a brand new price of $399 and in July the 16GB 4050Ti will be announce for a brand new price of $499. The 4050 will use a 12pin power plug but I don't know what the 4050Ti will use, hopefully the same one. That doesn't really fit with your "low power" thing but it might for the price.
DOA! NGEEDIA continuing to
)
DOA!
NGEEDIA continuing to rip off consumers.
Thx Harri. I did not know that.
Seems there is virtually no uplift with a 4070Ti which is odd/
Chooka wrote: DOA! NGEEDIA
)
I think this is the first 4070Ti I have seen here. Very curious to watch the valid/invalid rates on this.
Chooka wrote: Seems there is
)
The Gamma ray (and BRP) app is fairly reliant on memory bandwidth. Nvidia nerfed the memory bandwidth on all the 40-series cards. making the increase in clock speeds and IPC of the 40-series almost a moot point since they're bottlenecked by the smaller memory bus. Einstein is the most reliant on this part of the GPU. most other projects don't have much of an issue with it and see at least generational improvements.
the high invalid % is another blow to their overall production.
_________________________________________________________________________
Ah! Thank you very much
)
Ah! Thank you very much Ian&Steve!
Hmm well that sucks a bit. I have been leaning towards Primegrid a bit lately. That with power efficiency is drawing me to NVIDIA.
Boca - I'm running the duel slot INNO X3D OC (but I should have just bought the base model imo)
Currently 48 invalids. That's
)
Currently 48 invalids. That's not so good either.
Chooka wrote: Currently 48
)
You have 1663 Valid tasks on that same pc, 48 is relative. Yes it's higher than most people would like but alot of people would take that 40 series card of your hand for the price of shipping and live with the 48 invalids!! They would probably do it for upto around $500 as well. IOW you have a great gpu Einstein just needs to figure out how to lower the number of Invalids this series of cards seems to be producing. Someone sending Bernd one of them might help.