GeForce RTX 3080 FE GPGPU Compute Workstation Performance

Brent_Justice

Administrator
Staff member
Joined
Apr 23, 2019
Messages
949
Points
93
banner.jpg




Introduction



It has happened, the NVIDIA GeForce RTX 3080 Founders Edition video card has been launched.  In our launch review, we looked deeply at gaming performance.  We evaluated nine games, both with and without Ray Tracing and DLSS at both 1440p and 4K resolutions.  This provided us a very good look at performance at both resolutions.  It also gave us a comparison to see how Ray Tracing and DLSS affect performance in games.  We also tested without either feature to see how good ole Rasterization performs and if the GeForce RTX 3080 FE truly provides an upgrade in performance at the $699 price segment.  Please check out our full gaming review if you have not already.



Today’s review is going to...
Click to expand...

Continue reading...
 
I appreciate you putting in the 2080Ti for comparison purposes, even though the true matching comparison is to the 2080 Super.

I'd love to really know why Handbrake didn't show much distinction across all the cards listed... Mind you, I don't use Handbrake, but I'd love to know why there is only a 4.2 second difference between all cards.
 
Space_Ranger said:
I appreciate you putting in the 2080Ti for comparison purposes, even though the true matching comparison is to the 2080 Super.

I'd love to really know why Handbrake didn't show much distinction across all the cards listed... Mind you, I don't use Handbrake, but I'd love to know why there is only a 4.2 second difference between all cards.
Click to expand...

My theory is application optimization, the NVENC encoder is the same between all the GPUs. Other video encoding/transcoding apps could behave differently. It's hard to find good free ones, most are pay, and as far as video editors go, the good ones that have hardware GPU acceleration are also pay and hard to constantly install and reinstall on different machines. I used Handbrake cause it is the most popular free one. So testing is a bit limited in this area. If anyone knows of any good free editing programs or transcoding programs that have hardware GPU encoding via NVENC and AMD VCE, let me know. But usually, you get what you pay for. I'll take suggestions on testing video encoding better.

My theory though is that we will need application patches for the Ampere architecture and new video encoder on board, which honestly may not be that different from Turing.
 
Both 3080 reviews are really good reviews -- With this workstation review, while I realize the 3080 is not intended to be a Titan RTX replacement, it would be pretty interesting seeing the Titan RTX on those charts with the 3080 -- $2,499.00 vs $699.00.
 
maybe I missed it, as I was interrupted several times when reading the review, but which benchmark lines up best with folding@home and gpu grid performance?
 
Incorrect. This 3080 supersedes the 2080 base model, not the 2080 Super or Ti. Even still, it is minimum 20% faster than the Ti across the board in all benchmarks, and often times even faster than that. You'll be seeing $200-$300 brand new 2080's out there very soon.
 
Nice review. Good info.

I wouldn't mind, one of these days, getting a 3080 for my TR box. Looks like it'd boost Blender and a few other things a bit without sacrificing children to unknown gods as payment.
 
Thanks @Brent_Justice for all your hard work on this and it also explains the data I came across on the weekend comparing leaked 3090 Blender scores. I compared them to 3080 FE, 2080 FE, and 2080 Ti with version 2.90 and was blown away the gains.

On a side note. Over the years I've had a number of paid licenses for video encoders. I usually use them for ripping various things to my kodi box. NVENC can be all over the map with it comes to driver support. I'll have it work great only to have an MS update come through and kick something out that NV was using and essentially break, or an NV update do the same. I also have upgraded cards in the same rigs only to lose my hardware accelleration that I had on the previous, inferior card. When it works it can be phenonmenal but even paid licenses won't help you if it takes the developer months, if at all, to update for it. It's always a kind of buyer beware when it comes to paid encoders even when they say they support NVENC.
 
This is what I was expecting on gaming performance, more so with RTX and DLSS.
 
"... an OpenCL benchmark took based on LuxCoreRender. "

typo to fix: change took to tool
 
Brent_Justice said:
My theory is application optimization, the NVENC encoder is the same between all the GPUs. Other video encoding/transcoding apps could behave differently. It's hard to find good free ones, most are pay, and as far as video editors go, the good ones that have hardware GPU acceleration are also pay and hard to constantly install and reinstall on different machines. I used Handbrake cause it is the most popular free one. So testing is a bit limited in this area. If anyone knows of any good free editing programs or transcoding programs that have hardware GPU encoding via NVENC and AMD VCE, let me know. But usually, you get what you pay for. I'll take suggestions on testing video encoding better.

My theory though is that we will need application patches for the Ampere architecture and new video encoder on board, which honestly may not be that different from Turing.
Click to expand...

StaxRip is a GUI fronted which can do encoding and I believe it does hardware encoding for nVidia and AMD cards. I've seen it recommended before for CPU video encoding because it handles threads much better than Handbrake, especially with x265.

That said, it can be confusing to figure out and use and I have no real experience with it. It is free, though. I know most of you don't have much spare time but it might be worth checking out at some point.
 
You must log in or register to reply here.
Become a Patron!
Back
Top