RTX 2080

Message boards : Number crunching : RTX 2080
Message board moderation

To post messages, you must log in.

Previous · 1 · 2 · 3 · 4 · 5 · Next

AuthorMessage
Ian&Steve C.
Avatar

Send message
Joined: 28 Sep 99
Posts: 4267
Credit: 1,282,604,591
RAC: 6,640
United States
Message 1985321 - Posted: 15 Mar 2019, 16:21:54 UTC - in response to Message 1985318.  

How does the run time of a 2070 or 2080 compare to a 1080 ti with special app? Does it just scale with the streaming multiprocessor number?


2070 is about equal to a 1080ti

2080 is a little faster. I want to say like 25% faster? I can’t remeber the exact amount. Maybe Keith can give a better number since he has both cards.
Seti@Home classic workunits: 29,492 CPU time: 134,419 hours

ID: 1985321 · Report as offensive
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13161
Credit: 1,160,866,277
RAC: 1,873
United States
Message 1985333 - Posted: 15 Mar 2019, 17:16:11 UTC - in response to Message 1985318.  

How does the run time of a 2070 or 2080 compare to a 1080 ti with special app? Does it just scale with the streaming multiprocessor number?

I would say that they are about equal. The 1080 Ti has more SM's than the 2080 but the 2080's SM's are more efficient. Depends on the workload. Some tasks have the 2080 about 20% faster than the 1080 Ti. Some tasks have the two exactly equal. Rarely have I seen the 2080 lose to the 1080 Ti in compute times on a similar type of task. Would be definitive if the same workunit was crunched on the same host with both cards.

I haven't run the two cards through a benchmark yet to prove one way or another. Just by looking at reported tasks on this host which houses both cards.

https://setiathome.berkeley.edu/show_host_detail.php?hostid=8030022
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 1985333 · Report as offensive
Profile Freewill Project Donor
Avatar

Send message
Joined: 19 May 99
Posts: 766
Credit: 354,398,348
RAC: 11,693
United States
Message 1985336 - Posted: 15 Mar 2019, 17:22:59 UTC - in response to Message 1985333.  

How does the run time of a 2070 or 2080 compare to a 1080 ti with special app? Does it just scale with the streaming multiprocessor number?

I would say that they are about equal. The 1080 Ti has more SM's than the 2080 but the 2080's SM's are more efficient. Depends on the workload. Some tasks have the 2080 about 20% faster than the 1080 Ti. Some tasks have the two exactly equal. Rarely have I seen the 2080 lose to the 1080 Ti in compute times on a similar type of task. Would be definitive if the same workunit was crunched on the same host with both cards.

I haven't run the two cards through a benchmark yet to prove one way or another. Just by looking at reported tasks on this host which houses both cards.

https://setiathome.berkeley.edu/show_host_detail.php?hostid=8030022


Thanks, Keith. That's where I'm getting confused. The wiki for GeForce 10 series shows 28 SM for the 1080ti and the GeForce 20 series wiki shows 46 for the 2080. Am I looking at the wrong column?
ID: 1985336 · Report as offensive
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13161
Credit: 1,160,866,277
RAC: 1,873
United States
Message 1985340 - Posted: 15 Mar 2019, 17:38:17 UTC - in response to Message 1985336.  

How does the run time of a 2070 or 2080 compare to a 1080 ti with special app? Does it just scale with the streaming multiprocessor number?

I would say that they are about equal. The 1080 Ti has more SM's than the 2080 but the 2080's SM's are more efficient. Depends on the workload. Some tasks have the 2080 about 20% faster than the 1080 Ti. Some tasks have the two exactly equal. Rarely have I seen the 2080 lose to the 1080 Ti in compute times on a similar type of task. Would be definitive if the same workunit was crunched on the same host with both cards.

I haven't run the two cards through a benchmark yet to prove one way or another. Just by looking at reported tasks on this host which houses both cards.

https://setiathome.berkeley.edu/show_host_detail.php?hostid=8030022


Thanks, Keith. That's where I'm getting confused. The wiki for GeForce 10 series shows 28 SM for the 1080ti and the GeForce 20 series wiki shows 46 for the 2080. Am I looking at the wrong column?

Ahh, but that is where Nvidia snookered you. Their definition of a SM changed with the Turing cards. The Pascal SM comprises 128 CUDA cores. The Turing SM comprises 64 CUDA cores. So to be on equal footing, you need to halve the Turing SM count to make it equivocal to a Pascal. So the 2080 only has "23" Pascal SM's compared to the 28 of the 1080 Ti.
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 1985340 · Report as offensive
Profile Freewill Project Donor
Avatar

Send message
Joined: 19 May 99
Posts: 766
Credit: 354,398,348
RAC: 11,693
United States
Message 1985343 - Posted: 15 Mar 2019, 17:52:22 UTC - in response to Message 1985340.  
Last modified: 15 Mar 2019, 17:52:41 UTC

How does the run time of a 2070 or 2080 compare to a 1080 ti with special app? Does it just scale with the streaming multiprocessor number?

I would say that they are about equal. The 1080 Ti has more SM's than the 2080 but the 2080's SM's are more efficient. Depends on the workload. Some tasks have the 2080 about 20% faster than the 1080 Ti. Some tasks have the two exactly equal. Rarely have I seen the 2080 lose to the 1080 Ti in compute times on a similar type of task. Would be definitive if the same workunit was crunched on the same host with both cards.

I haven't run the two cards through a benchmark yet to prove one way or another. Just by looking at reported tasks on this host which houses both cards.

https://setiathome.berkeley.edu/show_host_detail.php?hostid=8030022


Thanks, Keith. That's where I'm getting confused. The wiki for GeForce 10 series shows 28 SM for the 1080ti and the GeForce 20 series wiki shows 46 for the 2080. Am I looking at the wrong column?

Ahh, but that is where Nvidia snookered you. Their definition of a SM changed with the Turing cards. The Pascal SM comprises 128 CUDA cores. The Turing SM comprises 64 CUDA cores. So to be on equal footing, you need to halve the Turing SM count to make it equivocal to a Pascal. So the 2080 only has "23" Pascal SM's compared to the 28 of the 1080 Ti.


*Almost snookered* I have smart friends that saved me before I opened my wallet. Thanks!
ID: 1985343 · Report as offensive
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13161
Credit: 1,160,866,277
RAC: 1,873
United States
Message 1985386 - Posted: 15 Mar 2019, 22:07:39 UTC
Last modified: 15 Mar 2019, 22:21:57 UTC

I would say the special app scales with the CUDA core count. Since is it a parallel processing app, the more CUDA cores you can bring into use, the faster it can crunch through a task. Compare the CUDA core count between the various cards of Pascal, Volta and Turing.

Pascal

GTX 1060 6GB- 1280 CUDA cores
GTX 1070 - 1920 cores
GTX 1070 Ti - 2432 cores
GTX 1080 - 2560 cores
GTX 1080 Ti - 3584 cores
Titan Xp- 3840 cores

Turing

GTX 1660 - 1408 CUDA cores
GTX 1660 Ti - 1536 cores
RTX 2060 - 1920 cores
RTX 2070 - 2304 cores
RTX 2080 - 2944 cores
RTX 2080 Ti - 4532 cores
Titan RTX - 4608 cores

Volta

Titan V - 5120 CUDA cores

So pay attention to the CUDA core count. A good comparison is between the Pascal GTX 1070 and the Turing RTX 2060. They have the same exact CUDA core count. So a comparison between those two cards will show the efficiency of the Turing CUDA core architecture plus the benefits of the change in GDDR RAM architecture with both having the same 8GB quantity and just the differences between GDDR5 and GDDR6.
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 1985386 · Report as offensive
Grant (SSSF)
Volunteer tester

Send message
Joined: 19 Aug 99
Posts: 13720
Credit: 208,696,464
RAC: 304
Australia
Message 1985392 - Posted: 15 Mar 2019, 22:41:44 UTC - in response to Message 1985386.  

A good comparison is between the Pascal GTX 1070 and the Turing RTX 2060. They have the same exact CUDA core count. So a comparison between those two cards will show the efficiency of the Turing CUDA core architecture plus the benefits of the change in GDDR RAM architecture with both having the same 8GB quantity and just the differences between GDDR5 and GDDR6.

And they have fairly similar boost clocks (the base is lower) for the reference/Founders Edition cards, yet Shaggie's graphs show the RTX 2060 to be well ahead of the GTX 1070. It's actually on par with (or slightly ahead of) the GTX 1080Ti.
Grant
Darwin NT
ID: 1985392 · Report as offensive
Profile Freewill Project Donor
Avatar

Send message
Joined: 19 May 99
Posts: 766
Credit: 354,398,348
RAC: 11,693
United States
Message 1989635 - Posted: 11 Apr 2019, 12:23:45 UTC - in response to Message 1985392.  

A good comparison is between the Pascal GTX 1070 and the Turing RTX 2060. They have the same exact CUDA core count. So a comparison between those two cards will show the efficiency of the Turing CUDA core architecture plus the benefits of the change in GDDR RAM architecture with both having the same 8GB quantity and just the differences between GDDR5 and GDDR6.

And they have fairly similar boost clocks (the base is lower) for the reference/Founders Edition cards, yet Shaggie's graphs show the RTX 2060 to be well ahead of the GTX 1070. It's actually on par with (or slightly ahead of) the GTX 1080Ti.


Can anyone compare the RTX 2060 vs GTX 1080 ti on the special Linux app? The 2060 is quite a bit cheaper new than the 1080 ti used currently.
ID: 1989635 · Report as offensive
Ian&Steve C.
Avatar

Send message
Joined: 28 Sep 99
Posts: 4267
Credit: 1,282,604,591
RAC: 6,640
United States
Message 1989637 - Posted: 11 Apr 2019, 12:27:19 UTC - in response to Message 1989636.  

2060 is about on par with the 1080 (for SETI) . I don’t think it’s as fast as the 1080ti.
Seti@Home classic workunits: 29,492 CPU time: 134,419 hours

ID: 1989637 · Report as offensive
juan BFP Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 16 Mar 07
Posts: 9786
Credit: 572,710,851
RAC: 3,799
Panama
Message 1989638 - Posted: 11 Apr 2019, 12:52:13 UTC - in response to Message 1989637.  
Last modified: 11 Apr 2019, 13:02:39 UTC

2060 is about on par with the 1080 (for SETI) . I don’t think it’s as fast as the 1080ti.

Any clues about the real performance of the 1660Ti vs 2060 in SETI?

By the latest available graphs http://setiathome.berkeley.edu/forum_thread.php?id=81962&postid=1988037 we cant be sure of that. The sample of 1660 is to small to be representative.
ID: 1989638 · Report as offensive
Profile -= Vyper =-
Volunteer tester
Avatar

Send message
Joined: 5 Sep 99
Posts: 1652
Credit: 1,065,191,981
RAC: 2,537
Sweden
Message 1989640 - Posted: 11 Apr 2019, 13:19:56 UTC - in response to Message 1989638.  
Last modified: 11 Apr 2019, 13:20:28 UTC

You can look here.

https://setiathome.berkeley.edu/results.php?hostid=8664852&offset=0&show_names=0&state=4&appid=

But i would say it depends on the WU itself. 12-30% slower than 2060. Most bang for the buck is still 2060 i would say.

_________________________________________________________________________
Addicted to SETI crunching!
Founder of GPU Users Group
ID: 1989640 · Report as offensive
Ian&Steve C.
Avatar

Send message
Joined: 28 Sep 99
Posts: 4267
Credit: 1,282,604,591
RAC: 6,640
United States
Message 1989652 - Posted: 11 Apr 2019, 14:31:32 UTC - in response to Message 1989638.  

2060 is about on par with the 1080 (for SETI) . I don’t think it’s as fast as the 1080ti.

Any clues about the real performance of the 1660Ti vs 2060 in SETI?

By the latest available graphs http://setiathome.berkeley.edu/forum_thread.php?id=81962&postid=1988037 we cant be sure of that. The sample of 1660 is to small to be representative.


i don't trust those charts. there's too much variance. it only samples the stock apps (i'm not sure if he's done any filtering to only include SoG or includes all apps) which leads to scenarios showing a properly configured 2060 beating an improperly configured 1080ti. I like manually comparing cards by looking at individual WU performance on the special app, which has less chance of configuration error and a lot less variance in the data.

best bang for buck is a tossup between 2060 and 2070 in my opinion. with the 2070 being faster, but costs more too. I've been slowly swapping my 1080tis for 2070s, and i can reliably sell the 1080tis for $525-550, and reliably buy the 2070s for $420-450 on ebay. it's been working out great. I have no worries buying used EVGA cards because of the excellent warranty from EVGA.
Seti@Home classic workunits: 29,492 CPU time: 134,419 hours

ID: 1989652 · Report as offensive
Profile Freewill Project Donor
Avatar

Send message
Joined: 19 May 99
Posts: 766
Credit: 354,398,348
RAC: 11,693
United States
Message 1989654 - Posted: 11 Apr 2019, 14:43:53 UTC - in response to Message 1989652.  


best bang for buck is a tossup between 2060 and 2070 in my opinion. with the 2070 being faster, but costs more too. I've been slowly swapping my 1080tis for 2070s, and i can reliably sell the 1080tis for $525-550, and reliably buy the 2070s for $420-450 on ebay. it's been working out great. I have no worries buying used EVGA cards because of the excellent warranty from EVGA.


Do the 20x0 series cards work on any version of the special app or is there a required or preferred (higher performing) version? I've got some older cards (980, 1060) that I'm looking to upgrade and want at least 1080 ti performance. Sounds like 2070 may be a sweet spot and perhaps more power efficient.
ID: 1989654 · Report as offensive
juan BFP Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 16 Mar 07
Posts: 9786
Credit: 572,710,851
RAC: 3,799
Panama
Message 1989656 - Posted: 11 Apr 2019, 14:47:54 UTC - in response to Message 1989652.  
Last modified: 11 Apr 2019, 14:51:53 UTC

best bang for buck is a tossup between 2060 and 2070 in my opinion. with the 2070 being faster, but costs more too. I've been slowly swapping my 1080tis for 2070s, and i can reliably sell the 1080tis for $525-550, and reliably buy the 2070s for $420-450 on ebay. it's been working out great. I have no worries buying used EVGA cards because of the excellent warranty from EVGA.

IMHO the best for me is the 2070, but EVGA don't sell hybrids of them, i know i could buy the kit and mount but i'm to old to do that, too many small parts to loose. LOL
That's is why i'm looking for a air one who works better in my hot environment. Unfortunately 4 x 2070 running on air cooling is not an option here. My 4x1070 hybrids are running at 52-62C even with the fan control at 60-100%. Now imagine the temperature if they run on air.......
My case is a TT cube X9 so i have room if i need to add some more fans to the mix.
From the test's i see the 2060 uses about 100-130W so i imagine they could work in my case even if they running on air.
About EVGA, i totally agree with you, i'm an EVGA fan for years, all my GPU's buyed in the last 5 years or more are EVGA, their quality is very good and their service makes all the difference in the case of a problem.
ID: 1989656 · Report as offensive
Ian&Steve C.
Avatar

Send message
Joined: 28 Sep 99
Posts: 4267
Credit: 1,282,604,591
RAC: 6,640
United States
Message 1989659 - Posted: 11 Apr 2019, 15:01:48 UTC - in response to Message 1989654.  
Last modified: 11 Apr 2019, 15:05:05 UTC


best bang for buck is a tossup between 2060 and 2070 in my opinion. with the 2070 being faster, but costs more too. I've been slowly swapping my 1080tis for 2070s, and i can reliably sell the 1080tis for $525-550, and reliably buy the 2070s for $420-450 on ebay. it's been working out great. I have no worries buying used EVGA cards because of the excellent warranty from EVGA.


Do the 20x0 series cards work on any version of the special app or is there a required or preferred (higher performing) version? I've got some older cards (980, 1060) that I'm looking to upgrade and want at least 1080 ti performance. Sounds like 2070 may be a sweet spot and perhaps more power efficient.


yes the 20 series cards work well on the special app. you can use them on any of the recent releases, but I've found the 20-series to work best on the CUDA 10 version from my testing.

as far as performance, 2070 = 1080ti, but uses about 15% less power.
Seti@Home classic workunits: 29,492 CPU time: 134,419 hours

ID: 1989659 · Report as offensive
Profile petri33
Volunteer tester

Send message
Joined: 6 Jun 02
Posts: 1668
Credit: 623,086,772
RAC: 156
Finland
Message 1989702 - Posted: 11 Apr 2019, 22:02:08 UTC - in response to Message 1989659.  


best bang for buck is a tossup between 2060 and 2070 in my opinion. with the 2070 being faster, but costs more too. I've been slowly swapping my 1080tis for 2070s, and i can reliably sell the 1080tis for $525-550, and reliably buy the 2070s for $420-450 on ebay. it's been working out great. I have no worries buying used EVGA cards because of the excellent warranty from EVGA.


Do the 20x0 series cards work on any version of the special app or is there a required or preferred (higher performing) version? I've got some older cards (980, 1060) that I'm looking to upgrade and want at least 1080 ti performance. Sounds like 2070 may be a sweet spot and perhaps more power efficient.


yes the 20 series cards work well on the special app. you can use them on any of the recent releases, but I've found the 20-series to work best on the CUDA 10 version from my testing.

as far as performance, 2070 = 1080ti, but uses about 15% less power.


Take a close look of Ian&Steve's tasks stderr.txt during the weekend and if you see V0.98 then take a look at the run-times. On Monday everything should look mundane again. It is a test with V0.98 on 20x0 and I have asked Ian&Steve not to give the test executable to anyone else.

Petri
To overcome Heisenbergs:
"You can't always get what you want / but if you try sometimes you just might find / you get what you need." -- Rolling Stones
ID: 1989702 · Report as offensive
Ian&Steve C.
Avatar

Send message
Joined: 28 Sep 99
Posts: 4267
Credit: 1,282,604,591
RAC: 6,640
United States
Message 1989738 - Posted: 12 Apr 2019, 3:52:33 UTC - in response to Message 1989702.  

v0.98 is nice so far!

here is a visual summary for 75 (60 2070, 15 1080ti) WU samples before and after (v0.97 CUDA10 vs. v0.98b1 CUDA10).



v0.98b1 is about 20% faster on these BLC32 vlar tasks on this system.

I'll try to run a benchmark suite tomorrow to compare the two with the same WUs for a better comparison with a more varied WU set. but i could see these changes immediately.
Seti@Home classic workunits: 29,492 CPU time: 134,419 hours

ID: 1989738 · Report as offensive
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13161
Credit: 1,160,866,277
RAC: 1,873
United States
Message 1989743 - Posted: 12 Apr 2019, 4:50:41 UTC - in response to Message 1989738.  

Hi Ian. Yes, it would be nice to see the benchmark results along with the similarity comparisons. Are you going to use Rick's benchMT tool or the older Lunatics tools? Looks good so far with your limited sample size.
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 1989743 · Report as offensive
Ian&Steve C.
Avatar

Send message
Joined: 28 Sep 99
Posts: 4267
Credit: 1,282,604,591
RAC: 6,640
United States
Message 1989744 - Posted: 12 Apr 2019, 5:02:11 UTC - in response to Message 1989743.  

i have no experience using his tool, so i'll probably just use the old tool.
Seti@Home classic workunits: 29,492 CPU time: 134,419 hours

ID: 1989744 · Report as offensive
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13161
Credit: 1,160,866,277
RAC: 1,873
United States
Message 1989746 - Posted: 12 Apr 2019, 5:23:38 UTC - in response to Message 1989744.  

OK, whatever is easiest and you are used to. I think his tool is pretty easy to use, but I have been testing for him for a long time, back to old beta days. Looking forward for general release of the 0.98 special app.
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 1989746 · Report as offensive
Previous · 1 · 2 · 3 · 4 · 5 · Next

Message boards : Number crunching : RTX 2080


 
©2024 University of California
 
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.