I'd like an update on 1080 Ti in 4.1 or 5.1

[AndreeOnline]

So... I had a 1080 Ti in my cMP a few months back. It was in Sierra just after Nvidia started to make web drivers again.

It worked well in apps with dedicated CUDA support, but there was some overall flakiness (but no show stoppers) on an OS level. And some benchmarks crashed which never feels good, since it implies not everything is quite ironed out. I returned my card after roughly a week and decided to check out Vega.

If there is anyone out there with a 1080 Ti in a cMP, please report here and let's talk. While I have faith that Vega 64 will be fully up and running soon, and that was my plan when I returned my 1080 Ti, I need to face facts that almost all plugins I have are targeted towards CUDA. =(

Anyway, I'd appreciate a status report now that we have High Sierra and a few versions of drivers behind us.

Thanks

 

[h9826790]

I am now with a PNY 1080Ti. I am not a professional in computer, but if there is any test I can do for you, please feel free to voice out (with detail steps, or link required, etc).

I installed this card for about a month now. It work flawlessly, the latest OS security update also didn't give me any trouble.

 

[AndreeOnline]

Thanks for your offer!

Actually, the pro apps worked pretty well back then too, so I'm not that worried about that.

Have you tried the normal benchmarks like Geekbench (Metal, CUDA and OpenCL) and Luxmark's 3 scenes?

I don't mean to put this burden on you, but maybe you've already done them?

I remember having some—not all—of Luxmark's scenes crash on me back then.

But it sounds good that it works well for you. Can you check the level of Metal support in 'about this Mac'?

------

EDIT: I followed your link and saw you already did some benchmarks there. Thanks. Have you tried all modes in GeekBench and LuxMark?

 

[h9826790]

No worries, since web driver need to be update after every OS update. So, I can run all Benchmarks once again with the latest OS build (and web driver).

GeekBench 4 Metal

GeekBench OpenCL

Luxmark 3 Luxball

Luxmark 3 Mic

Luxmark 3 Hotel

All benchmarks finished without crash. However, for Luxmark 3 Hotel scenes. If leave -cl-fast-relaxed-math ON, the result will become extremely poor. And image validation fail.

For METAL support, System Report shows the following

And OpenGL Extension Viewer shows in different format.

Anyway, hasn't hit a single bug / crash / error so far. I even tried reproduce one of the "known bug", but still nothing happen.

The only issue so far is CUDA-Z heavy test can shutdown the Mac, but I consider that's very exceptional, since the card works fine under both Furmark and OCCT stress test. And obviously, it's my own problem because I didn't go though the Pixlas mod (once again, thanks for creating that post) and let the card only run with dual mini 6pin supply.

As a side topic. I studied your Pixlas mod post again few days ago. It seems really very easy to perform (thanks for the detailed but simple explanation).

1) Tap the 12V pin to the 12V wire (inner thick wires)
2) Tap the GND pint to the GND wire (outer thick wires)

I only have one question. Is the sense pin connect to one of the GND wire?

I still feel that I don't need the mod at this moment (I won't even touch CUDA-Z in general), but still want to be fully ready if I need it (or want it) suddenly.

 

[AndreeOnline]

So, I can run all Benchmarks once again with the latest OS build (and web driver).

Like a boss.

Very nice, thanks.

 

[Synchro3]

I modified my Asus GTX 1080 Ti FE with a water cooling solution.

In Windows I can increase the power to 300 Watt (+20%), giving about 20% more performance, beating Titan Xp with no problem. Even in macOS without OC the card is noticeable faster.



With the normal Founders edition cooler the card reaches within a minute 84°C under load and throttles down (no OC):

 

[AndreeOnline]

Has anyone of you guys tried to game in MacOS?

I don't really game on my Mac Pro, except very occasionally, but I remember F1 2016 crashing with my 1080 Ti, but it didn't with the RX 480. Also, the RX 480 was actually faster which is another indication that the 1080 Ti couldn't quite stretch its legs.

 

[h9826790]

I modified my Asus GTX 1080 Ti FE with a water cooling solution.

In Windows I can increase the power to 300 Watt (+20%), giving about 15% more performance, beating Titan Xp with no problem. Even in macOS without OC the card is noticeable faster.

View attachment 740196

With the normal Founders edition cooler the card reaches within a minute 84°C under load and throttles down (no OC):

Yeah, that's why I picked a card that with reference PCB, but after market cooler. I want reference PCB to maximise the compatibility, but also want a after market cooler to stay away from throttling.

Has anyone of you guys tried to game in MacOS?

I don't really game on my Mac Pro, except very occasionally, but I remember F1 2016 crashing with my 1080 Ti, but it didn't with the RX 480. Also, the RX 480 was actually faster which is another indication that the 1080 Ti couldn't quite stretch its legs.

No crash, but sure not performing well if compare to Windows.

The game's auto (recommended) setting.

In games graphics (not too bad actually)

Benchmark result (not perfectly smooth, but absolutely playable, 60FPS max regardless settings)

Game power draw (roughly the same power draw for any setting)

It actually OK to run at max setting

in Game graphic

Benchmark result (it's just playable, not good)

 

[h9826790]

Old OpenGL games are definitely doing better.

 

[fendersrule]

I hardly ever game in MacOS anymore. With the GTX 1070 in MacOS, it quite spanked the RX580's Heaven 4.0 score. I don't recall any steam game crashing in MacOS, regardless if I was using the 7950, RX580, or the GTX 1070. It almost always played slower in MacOS, and that was the only difference.

I typically reserve Indie games (BroForce, Oh Sir, tc) for in MacOS, or classic titles--these are fine playing with MacOS's handicapped GPU optimizations and even if they run slower, I cannot sense it because they are simple games.

I played stuff like Killing Floor, Chroma Squad, Tim Shaffer stuff, and TF2 all with ease in MacOS at one time.

 

[kennyman]

@h9826790

Thank you for taking the time to benchmark the PNY 1080 Ti. Barefeats` tests are also showing more or less the same numbers. I had an EVGA 1080 Ti FE. I sold it, was having some glitches on FCPX. I am now using a 7950 (normal PC card).

I am looking for a new video card, either a Vega 64 or Pascal 1080 Ti will do. The Gigabyte 1080 Ti is cheaper than the Vega 64 and looks like the nvidia driver is good now, less buggy than before.

Have you tried FCPX? Any glitches while previewing/rendering video?

 

[pierrox]

@h9826790
Have you tried FCPX? Any glitches while previewing/rendering video?

If FCPX is your main "power hungry" application, then you should NOT go for a Nvidia. Get an AMD card for sure.

 

[AndreeOnline]

I sold it, was having some glitches on FCPX.

Was this in High Sierra with a recent driver, or a while back?

What issues did you have?

 

[h9826790]

@h9826790

Thank you for taking the time to benchmark the PNY 1080 Ti. Barefeats` tests are also showing more or less the same numbers. I had an EVGA 1080 Ti FE. I sold it, was having some glitches on FCPX. I am now using a 7950 (normal PC card).

I am looking for a new video card, either a Vega 64 or Pascal 1080 Ti will do. The Gigabyte 1080 Ti is cheaper than the Vega 64 and looks like the nvidia driver is good now, less buggy than before.

Have you tried FCPX? Any glitches while previewing/rendering video?

I just did a simple FCPX project few days few days ago. No glitches. It's very smooth indeed, but it's just a 720P 60 project, nothing too demanding. However, if your usage is FCPX, I think it's better to go for RX580. Vega 64 may be more powerful, but IMO the driver is not mature enough. Also, you pay a lot more for little to no extra performance (depends on usage).

1080Ti is good in general, but not particular doing well in FCPX. It's even slower (rendering) than the 7950 (but preview is better).

If you want to, I can try BruceX again with the latest web driver. My last test shows 39s with the 1080Ti. I won't expect there is any significant improvement. The card utilisation rate is very very low, just around 25%. But if you use Resolve, than 1080Ti is a very strong card. I tried to do a similar project on Resolve. Obviously the card perform much better. However, I personally still prefer FCPX, I am so used to the FCPX UI and functions. I just can't quite get what I want in Resolve.

 

[StoneJack]

Has anyone of you guys tried to game in MacOS?

I don't really game on my Mac Pro, except very occasionally, but I remember F1 2016 crashing with my 1080 Ti, but it didn't with the RX 480. Also, the RX 480 was actually faster which is another indication that the 1080 Ti couldn't quite stretch its legs.

Installed RX 560. Exceeds GTX 750 Ti and generally just works unlike Nvidia under High Sierra.

 

[h9826790]

@h9826790

Thank you for taking the time to benchmark the PNY 1080 Ti. Barefeats` tests are also showing more or less the same numbers. I had an EVGA 1080 Ti FE. I sold it, was having some glitches on FCPX. I am now using a 7950 (normal PC card).

I am looking for a new video card, either a Vega 64 or Pascal 1080 Ti will do. The Gigabyte 1080 Ti is cheaper than the Vega 64 and looks like the nvidia driver is good now, less buggy than before.

Have you tried FCPX? Any glitches while previewing/rendering video?

I just ran the BruceX with the latest OS / web driver, 35s. The RX580 is almost 2x faster. So, if your purpose is mainly FCPX, definitely go for the RX580. Even though during the BruceX test, the VRAM usage go up to 90% (out of 11GB), but I believe 8GB VRAM should be enough in general.

It's really no point to go for a non OOTB card that is expensive and way under perform in FCPX.