Message boards : Graphics cards (GPUs) : RTX 3080 Ti vs RTX 2080 Ti
| Author | Message |
|---|---|
 Retvari Zoltan Send message Joined: 20 Jan 09 Posts: 2343 Credit: 16,201,255,749 RAC: 851 Level  Scientific publications   | |
|
Workunit batch: ADRIA_BanditGPCR_APJ_b1 | -------------- RTX 3080 Ti ------------- | ----- RTX 2080 Ti ------ |
power | boost | | speed | power | GPU | boost | | GPU |
limit | clock | runtime | incr. | incr. | temp | clock | runtime | temp |
350W | 1815MHz | 17.347s | 32.0% | 36%* | 71°C | | | |
300W | 1710MHz | 17.961s | 27.5% | 20% | 67°C | 1860MHz | not yet | 71°C |
250W | 1575MHz | 19.176s | 19.4% | 0% | 63°C | 1800MHz | 22.900s | 68°C |
* the actual power draw was "only" 340W, so the increase is only 36%, not 40% to be fairThe base for the speed increase calcualtion is the RTX 2080 Ti @ 250W. The RTX 3080 Ti founders edition has remarkably good cooling, provided that you will use only one in a single PC. I will to put the GTX 1080 Ti into the comparison, but I decided to postpone it after this expired certificate issue is solved. I have a feeling that the improvement of the RTX 3080 Ti over the RTX 2080 Ti is much less than the improvement of the RTX 2080 Ti over the GTX 1080 Ti, or is it just the bonus system of folding@home playing with my mind? | |
ID: 57931 | Rating: 0 | rate:
 /   Reply
Quote
| |
|
Ian&Steve C.  Send message Joined: 21 Feb 20 Posts: 1035 Credit: 36,941,282,483 RAC: 38,995,534 Level Scientific publications  | |
|
I leave my 3080Ti set to 300W. It's a nice balance between power and performance in my opinion. | |
| ID: 57958 | Rating: 0 | rate:
/ Reply
Quote
| |
|
Aurum  Send message Joined: 12 Jul 17 Posts: 399 Credit: 13,031,999,382 RAC: 846,850 Level Scientific publications  | |
|
No comparison, I'd take the 2080 Ti over the 3080 Ti any day. The 3080 Ti is way overpriced. You should also look at the comparisons that have been done between the 3080 and the 3080 Ti. Very small increase in performance compared to the 2080 Ti vs 2080 or the 1080 Ti vs 1080. It's the stupid FP32 + (FP32 OR INT) architecture and some other design mistakes they made. And the 3080 Ti comes with LHR Malware factory installed. The last 2080 Ti I bought was like new, 3 fans, for $495 on eBay, ahh those were the days :-) | |
| ID: 58375 | Rating: 0 | rate:
/ Reply
Quote
| |
|
Ian&Steve C. Send message Joined: 21 Feb 20 Posts: 1035 Credit: 36,941,282,483 RAC: 38,995,534 Level Scientific publications | |
LHR Malware factory installed. lol @ "malware" depends on your use-case and/or goals. if you ONLY contribute to GPUGRID and/or mine Ethereum, then yes the 3080Ti isn't worth the price. I'm perfectly happy with the four 3080Ti's I have. they are the fastest GPUs (mine specifically) in the world on Einstein@home right now, even at 300W limited, and the 3080Ti is about 60% faster and 17% more power efficient than the 2080Ti (for Einstein). Since I contribute a lot to Einstein, it's very worth it for me for the power savings. (GDDR6X is where the gains are for Einstein since it relies so heavily on memory performance). for GPUGRID, the Ampere cards have about the same power efficiency as Turing. They are faster, but use more power at about the same rate. and I don't mine, so I don't need to participate in the rat race for FHR cards. I sold off my 8x RTX 2070 cards to some miner for a nice profit, and replaced them all with 5x RTX 3070Ti cards. faster for the same total power use while freeing up CPU resources. Sold off my 7x RTX 2080 cards, and used the proceeds to upgrade several of my 2080Tis to 3080Tis. I now have 3 systems performing the same work with less power use than my previous setup with 4 systems. regarding drivers, there were lots of reports of the 500+ branch being buggy in many different ways, and there was no compelling reason to upgrade (no performance gain), so I've just stayed on the 470 branch. ____________  | |
| ID: 58376 | Rating: 0 | rate:
/ Reply
Quote
| |
|
Retvari Zoltan Send message Joined: 20 Jan 09 Posts: 2343 Credit: 16,201,255,749 RAC: 851 Level Scientific publications | |
The performance of the GPUGRid app on 3080Ti is *not* limited by the LHR.the 3080 Ti comes with LHR Malware factory installed. I sold off my 8x RTX 2070 cards to some miner for a nice profit, and replaced them all with 5x RTX 3070Ti cards. faster for the same total power use while freeing up CPU resources. Sold off my 7x RTX 2080 cards, and used the proceeds to upgrade several of my 2080Tis to 3080Tis. I now have 3 systems performing the same work with less power use than my previous setup with 4 systems.This is true for Einstein@home, right? | |
| ID: 58377 | Rating: 0 | rate:
/ Reply
Quote
| |
|
Aurum Send message Joined: 12 Jul 17 Posts: 399 Credit: 13,031,999,382 RAC: 846,850 Level Scientific publications | |
The performance of the GPUGRid app on 3080Ti is *not* limited by the LHR.the 3080 Ti comes with LHR Malware factory installed. I never said it was. But I bet it is reduced by the stupid FP32 + (FP32 OR INT). I sold off my 8x RTX 2070 cards to some miner for a nice profit, and replaced them all with 5x RTX 3070Ti cards. faster for the same total power use while freeing up CPU resources. Sold off my 7x RTX 2080 cards, and used the proceeds to upgrade several of my 2080Tis to 3080Tis. I now have 3 systems performing the same work with less power use than my previous setup with 4 systems.This is true for Einstein@home, right? | |
| ID: 58378 | Rating: 0 | rate:
/ Reply
Quote
| |
|
Ian&Steve C. Send message Joined: 21 Feb 20 Posts: 1035 Credit: 36,941,282,483 RAC: 38,995,534 Level Scientific publications | |
The performance of the GPUGRid app on 3080Ti is *not* limited by the LHR.the 3080 Ti comes with LHR Malware factory installed. I also never said it was. I said if you're doing either of those things ONLY, then a 3080Ti upgrade probably isnt worth the cost, but for different reasons. for GPUGRID, the cost increase of the 3080Ti over a 2080Ti isn't justified since even though you gain some more speed, it scales with power use and you really dont gain much overall efficiency. you'll pay a lot more money about 2x, and get a card that's only a little faster with no efficiency boost (~30% or so, but also using 30% more power), so costing more money in electric. and of course not worth it for ETH mining since it's LHR and the hashrates are limited artificially by Nvidia's BIOS. (though from what I've read there are some tricks to work around this to unlock at least some more ETH mining performance) That's why I said it depends on your individual use case. there are many people that contribute to a variety of projects, and some projects will absolutely see a big gain with Ampere. I listed Einstein as just one example. I sold off my 8x RTX 2070 cards to some miner for a nice profit, and replaced them all with 5x RTX 3070Ti cards. faster for the same total power use while freeing up CPU resources. Sold off my 7x RTX 2080 cards, and used the proceeds to upgrade several of my 2080Tis to 3080Tis. I now have 3 systems performing the same work with less power use than my previous setup with 4 systems.This is true for Einstein@home, right? I'm not sure what you mean, but yes, this is true for Einstein with my specific situation. after switching a lot of my cards from Turing to Ampere (GDDR6X models), I have reduced the number of systems that I am running, reduced the overall power consumption, while keeping about the same overall Einstein production. As I mentioned, Einstein is very sensitive to GPU memory performance so the GDDR6X models are the most power efficient. only small efficiency gains from Turing/GDDR6 to Ampere/GDDR6. for example, under the same ~1200W GPU power budget, 5x 3070Ti perform about 20% better than 8x 2070. the extra power of the 3080Tis lets them push a little harder and they are even a little better than the 3070Tis, but they also come with a bit higher cost. ____________ | |
| ID: 58379 | Rating: 0 | rate:
/ Reply
Quote
| |
Message boards : Graphics cards (GPUs) : RTX 3080 Ti vs RTX 2080 Ti
