Since I don't know what EXACT EVGA 3080Ti you have, is the thickness of the EVGA thicker or thinner than the FE version? If it is thicker, then when you swapped then with each other, the thinner FE has more 'breathing' room, hence it will get a bit cooler.
I believe it is nearly the same. Both are thinner than the high-end rtx 3080 ti you have. And the FE has a fan on top and on the bottom instead of all on the bottom.
The gpus are in the 7th (next to the CPU) and 4th? slot so they are not "cheek and jowl". The 6th slot is right next to the 7th slot.
I have some more Nvidia gpus coming in that I will "try" installing in the "cheek and jowl" positions to see what happens to the GPU temperatures. I will be astonished if temperatures don't go up :) But really pleased if they don't.
Tom M
A Proud member of the O.F.A. (Old Farts Association). Be well, do good work, and keep in touch.® (Garrison Keillor)
This is a fascinating review of GPU's that completely out of most hobbyist's price range.
It also answers the question of what could replace a rtx 4090...
Tom M
We are going to try and get our hands on these sometime next school year, but no promises. Of course, the common question would be- "why not just get 4090s?". It's pretty simple- if we are getting another system from an approved vendor, then a 4090 is not an option in their workstations. Also, power delivery to multiple 4090s is more complicated, whereas the RTX 6000 Ada is pretty easy, since they are 300 watts (the power supply just has to have the new plugs). Lenovo, HP and now Dell all have just recently announced workstations that have room/power for 3 or 4 of the RTX 6000 Ada GPUs and will support them (soon).
You are comparing apples to oranges. Nothing can replace the RTX 4090 at the top of the consumer end of the market.
@BOCA RATON COMMUNITY HS
Taking into account recent events, and past trends of "gun violence", it would be a wise decision to get your hands on ballistic protection gear instead of GPUs, this school year, as soon as possible.
I've been an AMD GPU guy for the last few generations. Back then, here at Einstein the AMD cards offered better einstein computation per purchase dollar and higher productivity per watt than the Nvidia cards, so I switched fleets after a number of Nvidia generations. I understand this switched back a while ago, and am assuming I should choose Nvidia for my next purchase.
I may choose to buy a new moderate performance card quite soon as I'm in the throes of trying to restore my main daily driver to health, replacing one suspect part at a time. This post is not about the debugging process for my machine, but seeking advice and counsel narrowly on GPU selection.
My existing card is a Radeon RX 6800, bought just before the great GPU price runup had really got going crazy.
My fleet power consumption (for just three machines) is higher than I would like, which I restrain by underclocking, undervolting, and using BOINC to not run Einstein in hours where the heat addition to a room is least welcome.
In buying a new card, I'd want to shoot well under the RX6800 in purchase price, but get as much Einstein compute capability per watt as I can.
Just to give an idea where I'm thinking, on looking into things just a little I find myself thinking about an RTX 3060.
If that cost neighborhood has better options (considering new purchase--not aftermarket) regarding purchase price per Einstein compute, and Einstein compute per watt, I'm eager to hear your suggestions.
If that cost neighborhood has better options (considering new purchase--not aftermarket) regarding purchase price per Einstein compute, and Einstein compute per watt, I'm eager to hear your suggestions.
Looking at price point, this is probably the best option. Any RTX 4xxx that is similar in price will "probably" have lower performance in Einstein compute (not necessarily other things like games). Has anyone seen other RTX 4xxx GPUs here and their ACTUAL (versus theoretical) performance?
Just to give an idea where I'm thinking, on looking into things just a little I find myself thinking about an RTX 3060.
...
I started with NV (550ti/560ti/650/580) cards in 2012, migrated to AMD (7970/280x/380/480/580/6800) and am now exclusively 30x0 NV.
I'd pass on the RTX 3060(ti). It offers almost the same performance as a RTX 3070 and both are limited by the GDDR6 memory speed. From a performance perspective on FGRP, your RX6800 (152W card oob) lands in between the RTX 3070 (240W card oob) and RTX 3070ti (320W card oob).
Attached is a recent performance/operating cost summary of my hosts - Kubuntu/3070/3070ti/3080ti with one Win10/3060. All my NV cards run power limited to reduce card temperature and wall power. With the change back to NV cards from AMD my RAC is up by a factor of 3 while power consumption has been reduced by 30%. The 3070ti is my pick for best all around power consumption/performance.
GWGeorge007 wrote:Since I
)
I believe it is nearly the same. Both are thinner than the high-end rtx 3080 ti you have. And the FE has a fan on top and on the bottom instead of all on the bottom.
The gpus are in the 7th (next to the CPU) and 4th? slot so they are not "cheek and jowl". The 6th slot is right next to the 7th slot.
I have some more Nvidia gpus coming in that I will "try" installing in the "cheek and jowl" positions to see what happens to the GPU temperatures. I will be astonished if temperatures don't go up :) But really pleased if they don't.
Tom M
A Proud member of the O.F.A. (Old Farts Association). Be well, do good work, and keep in touch.® (Garrison Keillor)
https://www.pugetsystems.com/
)
https://www.pugetsystems.com/labs/articles/nvidia-rtx-6000-ada-vs-rtx-a6000-for-content-creation/
This is a fascinating review of GPU's that completely out of most hobbyist's price range.
It also answers the question of what could replace a rtx 4090...
Tom M
A Proud member of the O.F.A. (Old Farts Association). Be well, do good work, and keep in touch.® (Garrison Keillor)
Tom M
)
We are going to try and get our hands on these sometime next school year, but no promises. Of course, the common question would be- "why not just get 4090s?". It's pretty simple- if we are getting another system from an approved vendor, then a 4090 is not an option in their workstations. Also, power delivery to multiple 4090s is more complicated, whereas the RTX 6000 Ada is pretty easy, since they are 300 watts (the power supply just has to have the new plugs). Lenovo, HP and now Dell all have just recently announced workstations that have room/power for 3 or 4 of the RTX 6000 Ada GPUs and will support them (soon).
@TOM M You are comparing
)
@TOM M
You are comparing apples to oranges. Nothing can replace the RTX 4090 at the top of the consumer end of the market.
@BOCA RATON COMMUNITY HS
Taking into account recent events, and past trends of "gun violence", it would be a wise decision to get your hands on ballistic protection gear instead of GPUs, this school year, as soon as possible.
I've been an AMD GPU guy for
)
I've been an AMD GPU guy for the last few generations. Back then, here at Einstein the AMD cards offered better einstein computation per purchase dollar and higher productivity per watt than the Nvidia cards, so I switched fleets after a number of Nvidia generations. I understand this switched back a while ago, and am assuming I should choose Nvidia for my next purchase.
I may choose to buy a new moderate performance card quite soon as I'm in the throes of trying to restore my main daily driver to health, replacing one suspect part at a time. This post is not about the debugging process for my machine, but seeking advice and counsel narrowly on GPU selection.
My existing card is a Radeon RX 6800, bought just before the great GPU price runup had really got going crazy.
My fleet power consumption (for just three machines) is higher than I would like, which I restrain by underclocking, undervolting, and using BOINC to not run Einstein in hours where the heat addition to a room is least welcome.
In buying a new card, I'd want to shoot well under the RX6800 in purchase price, but get as much Einstein compute capability per watt as I can.
Just to give an idea where I'm thinking, on looking into things just a little I find myself thinking about an RTX 3060.
If that cost neighborhood has better options (considering new purchase--not aftermarket) regarding purchase price per Einstein compute, and Einstein compute per watt, I'm eager to hear your suggestions.
archae86 wrote:I've been an
)
..
archae86 wrote: If that cost
)
Looking at price point, this is probably the best option. Any RTX 4xxx that is similar in price will "probably" have lower performance in Einstein compute (not necessarily other things like games). Has anyone seen other RTX 4xxx GPUs here and their ACTUAL (versus theoretical) performance?
I'd shop for one of the 12GB
)
I'd shop for one of the 12GB 3060 versions. With that much VRAM, you can run 3X for a lot of projects like GPUGrid Python.
archae86 wrote:...Just to
)
I started with NV (550ti/560ti/650/580) cards in 2012, migrated to AMD (7970/280x/380/480/580/6800) and am now exclusively 30x0 NV.
I'd pass on the RTX 3060(ti). It offers almost the same performance as a RTX 3070 and both are limited by the GDDR6 memory speed. From a performance perspective on FGRP, your RX6800 (152W card oob) lands in between the RTX 3070 (240W card oob) and RTX 3070ti (320W card oob).
Attached is a recent performance/operating cost summary of my hosts - Kubuntu/3070/3070ti/3080ti with one Win10/3060. All my NV cards run power limited to reduce card temperature and wall power. With the change back to NV cards from AMD my RAC is up by a factor of 3 while power consumption has been reduced by 30%. The 3070ti is my pick for best all around power consumption/performance.
Thanks for tips and
)
Thanks for tips and responses.
For better or worse, I ordered an Asus DUAL-RTX3060-O12G-V2, so I do get 12 Gbytes.