1080 TI vs 1080 TI SLI vs 2080 TI Personal Review

[noko]

Well when the 2080 Ti hit, regardless of the hardware failure horrors, reported, lack of games using new features, games that did was not impressive -> The performance numbers of many of the games were way below my 1080Ti SLI performance as well, . Why bother was my final conclusion, the price was utterly ridiculous for what was being offered. Now that RTX has time to mature some, that tech is perking my interests so Ampere makes much more sense as an upgrade if it is good and should give a significant performance bump when I upgrade to a higher resolution, high refresh rate, HDR . . . monitor. Vulcan78 critiqued the initial review in which the review was done in good faith but had significant issues which we missed. If one does not set up SLI right, proper cooling, consistent clock speeds between the cards (very important for eliminating/minimizing micro-stutter), consistent PCIe lanes etc. yeah you will get less than a stellar experience. SLI can give you performance/Quality settings unmatched by a single card, so half empty or half full, for the games SLI works well you will not be able to match it with a single card - that has always been the case. The games that don't play well with SLI or MGPU, turn off SLI and enjoy, those that do? Be ready to be blown away.

 

[vulcan78]

I was big into SLI for a good while. My last SLI setup was a pair of GTX 970. I went to a single GTX Titan X. In pure FPS the Titan was a little slower but the experience was a lot better. With the declining SLI support I have not even considered doing it again since.

That's not how PCI-E lanes on Intel work. The video cards get 16 lanes dedicated to them from the CPU. The M.2 slot get 4 lanes from the PCH. Running 2 video cards and an M.2 NVME drive gives x8/x8 to the video cards and x4 to the SSD.

Yes, you are correct, I stand corrected, apparently M.2 takes PCI-E lanes from chipset and you sacrifice a SATA port but it doesn't affect the 16 lanes dedicated to CPU. You do drop down to 8x/8x though with SLI on Z370.

Well, given that there is only a 1% difference in performance between 8x and 16x I don't know how else to account for the massive discrepancy in performance OP is reporting from what others are reporting, such as this:

OP

Results:

Gigabyte 1080 TI Turbo
Min: 56
Average: 95.25
Max: 137

Gigabyte 1080 TI Turbo SLI
Min: 89
Average: 109.6
Max: 151

Nvidia 2080 TI FE
Min: 102
Average: 120.16
Max: 181

And this:

DudeRandom84 BF5 direct comparison (also Z370 chipset):



Same goes for the rest of the titles, including Apex Legends. Maybe there was a driver update between this post and now as it's been a year. Not sure.

 

[IdiotInCharge]

don't know how else to account for the massive discrepancy in performance OP is reporting from what others are reporting, such as this:

This is why benchmarking is hard. Professional reviewers struggle with repeatedly of their own results regularly enough, let alone replicating other's results, that making hard claims is always perilous.

This is graduate-level research at a minimum once you start to apply the scientific method; anything less opens your work up to broad, valid criticism surrounding potential inconsistencies and holes in your argument you may have missed or simply forgot to address.

 

[vulcan78]

Well when the 2080 Ti hit, regardless of the hardware failure horrors, reported, lack of games using new features, games that did was not impressive -> The performance numbers of many of the games were way below my 1080Ti SLI performance as well, . Why bother was my final conclusion, the price was utterly ridiculous for what was being offered. Now that RTX has time to mature some, that tech is perking my interests so Ampere makes much more sense as an upgrade if it is good and should give a significant performance bump when I upgrade to a higher resolution, high refresh rate, HDR . . . monitor. Vulcan78 critiqued the initial review in which the review was done in good faith but had significant issues which we missed. If one does not set up SLI right, proper cooling, consistent clock speeds between the cards (very important for eliminating/minimizing micro-stutter), consistent PCIe lanes etc. yeah you will get less than a stellar experience. SLI can give you performance/Quality settings unmatched by a single card, so half empty or half full, for the games SLI works well you will not be able to match it with a single card - that has always been the case. The games that don't play well with SLI or MGPU, turn off SLI and enjoy, those that do? Be ready to be blown away.

Yes but my critique was flawed, as per my last comment above, M.2 doesn't actually take lanes away from CPU but the chipset on Z370. 16 lanes are still available, I just don't understand how OP has such poor performance with 1080 Ti SLI. Take BF5, where 1080 Ti SLI is easily 30% faster than single 2080 Ti, but OP has single 2080 Ti some 20% faster than 1080 Ti SLI (as though SLI is not even working at all considering single 2080 Ti is only some ~30% on avg faster than single 1080 Ti).

 

[CruisD64]

Yes but my critique was flawed, as per my last comment above, M.2 doesn't actually take lanes away from CPU but the chipset on Z370. 16 lanes are still available, I just don't understand how OP has such poor performance with 1080 Ti SLI. Take BF5, where 1080 Ti SLI is easily 30% faster than single 2080 Ti, but OP has single 2080 Ti some 20% faster than 1080 Ti SLI (as though SLI is not even working at all considering single 2080 Ti is only some ~30% on avg faster than single 1080 Ti).

At least for me, the last time I used SLI was with a pair of GTX 980's. Prior to that GTX 680's and prior to that GTX 470's. I remember playing the witcher 3 and it just never felt right...stuttering was real. FPS isn't everything. I decided to buy a single 980ti back then and never looked back. The idea of 1080ti's in SLI sound great, especially since you can pick up a 1080ti in the $400 range used now but after my previous experience's I wouldn't even consider it. Faster single card all the way. I dunno. Just personal preference. ¯\_(ツ)_/¯

 

[IdiotInCharge]

At least for me, the last time I used SLI was with a pair of GTX 980's. Prior to that GTX 680's

Had GTX970s (3.5GB master race!) and GTX670s before that -- and HD6950s before those.

The Nvidia cards were infinitely better than pre-frame-pacing Crossfire, but there was some stutter in some games. Depending on the game and settings, though, the increase in performance and fluidity was still there, even over the fastest GPU of the same generation.

Now, had I known that a 980Ti was coming, I would have skipped the pair of 970s. When the 1080Ti came around and one became available at MSRP during the mining craze, I didn't hesitate, and yeah, I prefer single card wherever possible.

I don't think we're going to be going back to multi-GPU until they get VR figured for both eyes with one GPU per eye, something that's been in the work for years, and / or when they break up the functions into chiplets transparent to the API / software and then use that understanding to expand to multiple cards.

 

[JMCB]

Yes, you are correct, I stand corrected, apparently M.2 takes PCI-E lanes from chipset and you sacrifice a SATA port but it doesn't affect the 16 lanes dedicated to CPU. You do drop down to 8x/8x though with SLI on Z370.

Well, given that there is only a 1% difference in performance between 8x and 16x I don't know how else to account for the massive discrepancy in performance OP is reporting from what others are reporting, such as this:

OP

Results:

Gigabyte 1080 TI Turbo
Min: 56
Average: 95.25
Max: 137

Gigabyte 1080 TI Turbo SLI
Min: 89
Average: 109.6
Max: 151

Nvidia 2080 TI FE
Min: 102
Average: 120.16
Max: 181

And this:

DudeRandom84 BF5 direct comparison (also Z370 chipset):



Same goes for the rest of the titles, including Apex Legends. Maybe there was a driver update between this post and now as it's been a year. Not sure.

Well I did the review over a year ago now, and there is a simple explanation- I’m not a professional reviewer, but also driver versions were probably different as well. Also all your arguments about PCIE express lanes made me laugh, as the performance doesn’t drop to x4, it will go to 8/8x (Z370 m.2 will use PCIE lanes from CPU, not board).

Overall, this was a review I did for fun on my personal rig. Congrats on necromancing a topic from a year ago, and attacking my non-professionally reviewed numbers with your new account that you created just for that purpose. If you like, I can add 2080 Ti NVLink numbers as well.

 

[Wooshoo]

Can you run the new checkerboard rendering through the 1080ti I haven't seen much of it since a video that Linus Tech Tips had of the performance on 2x2080tis

 

[vulcan78]

Well I did the review over a year ago now, and there is a simple explanation- I’m not a professional reviewer, but also driver versions were probably different as well. Also all your arguments about PCIE express lanes made me laugh, as the performance doesn’t drop to x4, it will go to 8/8x (Z370 m.2 will use PCIE lanes from CPU, not board).

Overall, this was a review I did for fun on my personal rig. Congrats on necromancing a topic from a year ago, and attacking my non-professionally reviewed numbers with your new account that you created just for that purpose. If you like, I can add 2080 Ti NVLink numbers as well.

Yes my apologies, my critique was flawed, but the performance figures you included are most certainly out of date, presumably due to a driver update. No idea how you got such low performance in say BF5 when "DudeRandom84" is showing 1080 Ti SLI 30% faster than single 2080 Ti, or the other video where 1080 Ti SLI at stock clocks is easily 30% faster on avg in a large number of games vs 2080 Ti running at 2200 MHz.

There may have been a few games that motivated you to ditch 1080 Ti SLI for single 2080 Ti but my experience has been, just disable one GPU and turn settings down. We are talking about a 30% performance difference here on avg.

Anyone who can afford 2080 TI SLI obviously doesn't care about bang for your buck though, unfortunately, members of your socio-economic demographic are providing a lot of reinforcement to NGreedia's pricing at the detriment of a good 75% of us waiting for a compelling upgrade from 1080 Ti that doesn't require donating a spare organ.

https://hardforum.com/threads/former-or-current-1080-ti-titan-pascal-owners.1990973/
But honestly, I would have just thrown both of those 1080 Ti's under water-blocks and skipped over 2080 Ti altogether. That's just me, someone for whom $2600 after taxes for GPU's is an insane amount of money for marginal performance.

It's bizarre that your motivation to go 2080 Ti was SLI, then you needlessly add a second 2080 Ti. I thought the whole point of this post was to show how obsolete SLI was.

 

[5150Joker]

Apex Legends
Apex Legends is a free-to-play Battle Royale game where legendary competitors battle for glory, fame, and fortune on the fringes of the Frontier.

Results:

Gigabyte 1080 TI Turbo
Min: 76
Average: 113.82
Max: 145

Gigabyte 1080 TI Turbo SLI
Min: 37
Average: 89.6
Max: 143

Nvidia 2080 TI FE
Min: 82
Average: 118
Max: 146

Between the single card and SLI 1080 TI setups, we can see that SLI scaling is broken, despite NVIDIA saying it received support for SLI in this title with the latest driver version. While playing it in SLI, it seemed like a hot mess – stutters, inconsistent frame drops. Switching to one card greatly improved the experience. In addition, we weren’t getting a whole lot better performance from a 1080 Ti vs a 2080 Ti. I’m guessing this is also due to newer drivers that aren’t optimized for this title at all. I’m sure future updates will bring some improvements.

Something is very wrong with your Apex Legends results. My 2080 Ti vs my old Titan X Pascal sees at least a 35% uplift in framerate if not more. I play at 1440p with everything low except the texture streaming budget on high for reference. I have my fps capped at 165 and my 2080 Ti stays pegged there 90% of the time and the only time it drops down to around 110 fps is when we're in the drop ship overlooking the city. Even the most intense fights the fps never drops below 140 fps.

 

[thesmokingman]

I would just add that you should be glad you have these first world problems.

 

[RazorWind]

For what it's worth, I think "because it's cool" is still a perfectly good reason to build an SLI setup, if the cost isn't an issue for you.

You can't use a bazillion horsepower car on the road either, but that doesn't stop people from buying or building them.

 

[Armenius]

Can you run the new checkerboard rendering through the 1080ti I haven't seen much of it since a video that Linus Tech Tips had of the performance on 2x2080tis

Checkerboard rendering only works with Turing, as far as I know. I wouldn't explicitly enable it for games in Inspector that don't have it enabled in the drivers already.

 

[Furious_Styles]

Something is very wrong with your Apex Legends results. My 2080 Ti vs my old Titan X Pascal sees at least a 35% uplift in framerate if not more. I play at 1440p with everything low except the texture streaming budget on high for reference. I have my fps capped at 165 and my 2080 Ti stays pegged there 90% of the time and the only time it drops down to around 110 fps is when we're in the drop ship overlooking the city. Even the most intense fights the fps never drops below 140 fps.

You can increase lots of the graphics settings with very little performance impact. I used to run it all on low but I get virtually the same FPS with a bunch of settings turned up. I'd love to turn on AA but I don't like the blur effect of TAA. Wish that game engine had other options because the shimmer sucks too.

 

[JMCB]

Something is very wrong with your Apex Legends results. My 2080 Ti vs my old Titan X Pascal sees at least a 35% uplift in framerate if not more. I play at 1440p with everything low except the texture streaming budget on high for reference. I have my fps capped at 165 and my 2080 Ti stays pegged there 90% of the time and the only time it drops down to around 110 fps is when we're in the drop ship overlooking the city. Even the most intense fights the fps never drops below 140 fps.

When I ran this test, the game was just released, and nothing was optimized. I’m sure it’s a lot better now.