Anyone running a 1660 Ti? Want to see how it does here and what your temps are like.


Thanks,

Zalster

---

Anyone running a 1660 Ti? Want to see how it does here and what your temps are like.


Thanks,

Zalster

Pretty sure it's turing architecture so it won't run until the 2000 series runs on GPUGrid

Yes, it's Turing. You'll get immediate computation errors, no support for those yet unfortunately.

For those interested in the 1660's compute performance, I posted some runtimes from mine over at PrimeGrid. FWIW, I remember temps were <70C most of the time.

Thank you, Azmodes!

another 1660Ti user eagerly awaiting Turing support on GPUGrid :-)

They will skip Turing gen.
https://www.tweaktown.com/news/65178/nvidia-tease-next-gen-ampere-gpu-7nm-gtc-19/index.html
Ampere is near heere

I just went from Maxwell to Pascal, by the use of Link connector from VGA to Displayport, so I upgraded from a GTX 750 to a GTX 1050 Ti. The speed increase in GPUGRID was notable, now I am testing Einstein.
Tullio

---

I guess this is the error because GTX 1660 Ti can't run this project. If it is when can an update be expected. It can be run on a number of projects like SETI, Primegrid, Amicable Numbers and Moo. As these new cards are becoming more popular can an update be seen soon:

Stderr output
<core_client_version>7.14.2</core_client_version>
<![CDATA[
<message>
(unknown error) - exit code -59 (0xffffffc5)</message>
<stderr_txt>
# GPU [GeForce GTX 1660 Ti] Platform [Windows] Rev [3212] VERSION [80]
# SWAN Device 0 :
# Name : GeForce GTX 1660 Ti
# ECC : Disabled
# Global mem : 6144MB
# Capability : 7.5
# PCI ID : 0000:01:00.0
# Device clock : 1830MHz
# Memory clock : 6001MHz
# Memory width : 192bit
# Driver version : r419_50 : 42531
#SWAN: FATAL: cannot find image for module [.nonbonded.cu.] for device version 750

</stderr_txt>
]]>

---

Proud Founder and member of



Have a look at my WebCam

I guess this is the error because GTX 1660 Ti can't run this project. If it is when can an update be expected. It can be run on a number of projects like SETI, Primegrid, Amicable Numbers and Moo. As these new cards are becoming more popular can an update be seen soon:

There has been a large number of unsent "longs" for some time. I am putting an extra GTX 1060 and 1070 on it. There is usually not enough work for them. At some point, they will need to get another CUDA expert, or whatever it takes.

Pretty sure the next GPU architecture will be out before they support turing

I guess this is the error because GTX 1660 Ti can't run this project. If it is when can an update be expected. It can be run on a number of projects like SETI, Primegrid, Amicable Numbers and Moo. As these new cards are becoming more popular can an update be seen soon:

There has been a large number of unsent "longs" for some time. I am putting an extra GTX 1060 and 1070 on it. There is usually not enough work for them. At some point, they will need to get another CUDA expert, or whatever it takes.

I'm no expert but I think likely all it takes is to update this
[.nonbonded.cu.] module for device version 750 API's.

I needed to update a device list in a C module for my temp sensor driver because it only had models listed for the older generation of AMD processors. I just added the latest second generation Threadripper models to the list and it compiled fine with the ability to recognize the new processors.

Same thing I think is going on here. Probably just need a revised list of known devices in that .nonbonded.cu file.

Would be great if someone from the GPUGRID people could tell us what their plans are.
Turing support coming up soon? Or skipping Turing and heading for Ampere support?

I guess for some crunchers here (including myself) such information would be valuable in connection with replacement plans of older cards.

As far as I know that's the status: link

MrS

---

Scanning for our furry friends since Jan 2002

As far as I know that's the status: link

thanks for the link.
Although time schedule and feasibility seem rather vague :-(

Since 4 months have passed since that statement from end of Dezember 2018, perhaps GDF could give us an update ?

I read that post as supporting the new RTX functions vs supporting the same CUDA functions on newer cards as the new features would require coding. I would think a Turing card would still support the same compute features with the same code if it was recognized.

Yes, supporting the current app on Turing should be rather low effort. Making use of exclusive Turing features or newer CUDA levels, that requires more effort.

I'd really like to upgrade my GTX1060 to a GTX2060, but knowing it's unsupported by GPUGrid, I'm not going to do it. Pity for the project itself, it could tap into so much more compute power by supporting Turing. :-(