Work Desktop for Research Advice

[BeefyWorkstationAdvice]

Budget (including currency): $5,000 USD (We get a very significant discount through Dell, so my prices are very flexible)

Country: US

Games, programs or workloads that it will be used for: Python and Anaconda using CUDA

Other details (existing parts lists, whether any peripherals are needed, what you're upgrading from, when you're going to buy, what resolution and refresh rate you want to play at, etc): 

 

This is a build request I received for a beefy faculty research machine. I've really gotten into LMG lately and have enjoyed all the videos. I've got plenty of experience with gaming, and creating standard workstations for my organization as well as some specialty machines for SAS research. This one however is different as I have not dealt with CUDA. This will be a customized Dell machine as we have contracts with them and do save significant amounts of money with them. The last SAS machine I speced out on their website went from ~$7,500 to $4,000 after our discount.

 

So, what details I could get are that they will be using Python/Anaconda with the CUDA toolkit. Where I'm sitting now in my build is below:

 

Precision 7920 Tower

Single Intel Xeon Gold 5215

128 GB 8x16GB DDR4 2933MHz RDIMM ECC

1TB M.2 NVME

Extra SATA Storage

 

Basically I'm unsure where to go with the GPU(S). I can tweak a lot and work with the Dell rep to get it in budget. I am considering Nvidia Quadro RTX4000 to RTX8000. There isn't an option for RTX A series, I'm not sure if the Dell rep can put that in or not.

 

Benefits for CUDA with dual gpus? Does this place the end user in needing to know how to write their code for multiple GPUs, or after initial setup is that handled but the CUDA toolkit? Cost/Benefit for a single more expensive card vs a cheaper dual card setup?

 

CUDA is new to me and I won't be using it myself after initial setup and deployment. I'd like something very powerful that will last a while and the inner workings of the gpu handled as simply by the end user if at all.

 

Thank you all

 

 

[jaslion]

Any chance for a threadripper/epyc workstation? They just smash intel in every which way. Their second gen stuff is still better than intels current xeons so amds third gen stuff is insanely strong. Would really be a waste to get intel here. I know dell should have epyc/threadripper workstations.

 

I don't know what application you will use. I'd say contact the people that make it about it and if it supports multiple cards for acceleration here. See what they say.

[Downkey]

4 hours ago, BeefyWorkstationAdvice said:

Budget (including currency): $5,000 USD (We get a very significant discount through Dell, so my prices are very flexible)

Country: US

Games, programs or workloads that it will be used for: Python and Anaconda using CUDA

Other details (existing parts lists, whether any peripherals are needed, what you're upgrading from, when you're going to buy, what resolution and refresh rate you want to play at, etc): 

 

This is a build request I received for a beefy faculty research machine. I've really gotten into LMG lately and have enjoyed all the videos. I've got plenty of experience with gaming, and creating standard workstations for my organization as well as some specialty machines for SAS research. This one however is different as I have not dealt with CUDA. This will be a customized Dell machine as we have contracts with them and do save significant amounts of money with them. The last SAS machine I speced out on their website went from ~$7,500 to $4,000 after our discount.

 

So, what details I could get are that they will be using Python/Anaconda with the CUDA toolkit. Where I'm sitting now in my build is below:

 

Precision 7920 Tower

Single Intel Xeon Gold 5215

128 GB 8x16GB DDR4 2933MHz RDIMM ECC

1TB M.2 NVME

Extra SATA Storage

 

Basically I'm unsure where to go with the GPU(S). I can tweak a lot and work with the Dell rep to get it in budget. I am considering Nvidia Quadro RTX4000 to RTX8000. There isn't an option for RTX A series, I'm not sure if the Dell rep can put that in or not.

 

Benefits for CUDA with dual gpus? Does this place the end user in needing to know how to write their code for multiple GPUs, or after initial setup is that handled but the CUDA toolkit? Cost/Benefit for a single more expensive card vs a cheaper dual card setup?

 

CUDA is new to me and I won't be using it myself after initial setup and deployment. I'd like something very powerful that will last a while and the inner workings of the gpu handled as simply by the end user if at all.

 

Thank you all

 

 

What epic/TR workstations available? All Xeons get destroyed by modern TR and Epyc CPUS.