Create an account on the HP Community to personalize your profile and ask a question
03-31-2021 03:52 PM
The useful GPU's for AI / machine learning will be the ones having Tensor cores: the RTX 4000 and the RTX 3060.
HP z620_2 (2017) (R7) > Xeon E5-1680 v2 (8C@ 4.3GHz) / z420 Liquid Cooling / 64GB (HP/Samsung 8X 8GB DDR3-1866 ECC registered) / Quadro P2000 5GB _ GTX 1070 Ti 8GB / HP Z Turbo Drive M.2 256GB AHCI + Samsung 970 EVO M.2 NVMe 500GB + HGST 7K6000 4TB + HP/HGST Enterprise 6TB / Focusrite Scarlett 2i4 sound interface + 2X Mackie MR824 / 825W PSU / Windows 7 Prof.’l 64-bit (HP OEM) > 2X Dell Ultrasharp U2715H (2560 X 1440)
[ Passmark Rating = 6280 / CPU rating = 17178 / 2D = 819 / 3D= 12629 / Mem = 3002 / Disk = 13751 / Single Thread Mark = 2368 [10.23.18]
HP z420_3: (2015) (R11) Xeon E5-1650 v2 (6C@ 4.3GHz) / z420 Liquid cooling / 64GB (HP/Samsung 8X 8GB DDR3-1866 ECC registered) / NVIDIA GeForce GTX 1060 6GB/ HP/LSI 9212-4i > Samsung 860 EVO 500GB + HGST 4TB / ASUS Essence STX + Logitech z2300 2.1 / 600W PSU > Windows 7 Professional 64-bit (HP OEM ) > Samsung 40" 4K
[Passmark System Rating: = 5644 / CPU = 15293 / 2D = 847 / 3D = 10953 / Mem = 2997 Disk = 4858 /Single Thread Mark = 2384 [6.27.19]
04-01-2021 10:50 AM
Thank you for responding Bambi,
So you think mixing quadro and geforce will produce better results than an older quadro with a newer one?
I am not finding much information regarding settup for either scenario. Most recommend two identical cards working together (SLI) which is not in the budget at this time:)
04-01-2021 11:03 AM
SLI on HP workstations is not supported except for several SPECIFIC model quadro cards that HP has approved no other consumer or quadro cards will work in SLI as HP only paid nvidia a license for these specific card models
note that the approved quadro cards changes depending on the workstation model, and as far as i know you can not use a z840 approved card on a z820 if the z820 does not list it as approved
next,.........the current RTX line of video card for AI usage only has support in newer software packages, or packages where the software vender has added this support, so unless you can upgrade your existing software or buy a new software package with RTX support a RTX card may or may not give you the results you expect
with the above said, i still agree with bambi, that a RTX based card if within your budget is the way to go even if your current software fails to support the RTX line
04-01-2021 11:07 AM
The key feature are the Tensor cores, which the M4000 does not have.
I'm not sure of the configuration to use them together, but you're correct that it won't be SLI which requires two identical GPU's, a special bridge, and a motherboard that supports SLI.
Rather than general advice on multi-GPU setup, consider directly contacting the firms' producing the software and ask if they can advise on setup, or perhaps they have a forum.
1. > installing a single card, test, add the second, test, and see if they are both recognized and the performance changes or:
2. > starting with the GPU having more tensor cores to concentrate on the work rather than complex setup, and
3. > eventually selling all three current GPU's to fund a single RTX 3000 series GPU which are increasingly optimized for machine learning.
04-01-2021 12:10 PM
Thank you again Bambi,
I did that to some extent:
RTX 3060 with RTX4000 benchmarked very similarly to RTX4000 with M4000 in my system and, while benchmark doesn't really seem to examine all options, it seemed like a starting point.
The RTX 3060 doesn't fit with the cooling cover in my workstation so I am playing with the RTX4000 and M4000 for a bit varying the settings within the Nvidia driver.
Will probably switch back to RTX 3060 with the RTX4000 and do the same.
Didn't find what you were looking for? Ask the community