13 minutes ago, Ray Proudfoot said:

Matt,

Your post started off fine but then you felt the need to have a dig at me when you could have just kept it objective.

I learned to test that way from Pete Dowson. It removes as much of the subjectivity as possible from the test.

But if your methos works for you then that's fine.

Just returning the favor Ray.  Life is a two way street.

Now wouldn’t GPU testing be more accurate using usage % over FPS?  I would think it would be more accurate testing with a stable frame rate.

2 minutes ago, mpw8679 said:

Just returning the favor Ray.  Life is a two way street.

Now wouldn’t GPU testing be more accurate using usage % over FPS?  I would think it would be more accurate testing with a stable frame rate.

Not for me, no. What's good enough for Pete Dowson is good enough for me. I find my method removes as much of the subjectivity as possible from the test. If you keep changing values I don't see how it's possible to determine how much faster one card is over another.

Each to their own.

8 minutes ago, Ray Proudfoot said:

Not for me, no. What's good enough for Pete Dowson is good enough for me. I find my method removes as much of the subjectivity as possible from the test. If you keep changing values I don't see how it's possible to determine how much faster one card is over another.

Each to their own.

What is the relation between GPU usage % and frame rates.  I fail to understand as it seems GPU performance stays relatively consistent till the card is fully saturated.  

17 minutes ago, mpw8679 said:

What is the relation between GPU usage % and frame rates.  I fail to understand as it seems GPU performance stays relatively consistent till the card is fully saturated.  

I don't know. If I want to compare card x against card y I have to apply exactly the same parameters to the test. That's done by using the same scenario. The only thing that has changed is the card.

I notice in your test you're changing shadows, cloud resolutions etc. You say it isn't scientific but more of a feeling for how it's performing.

I believe the test I do takes out all the subjectivity as no settings are changed. It's purely down to how one piece of hardware compares to another all other things being unchanged.

5 minutes ago, Ray Proudfoot said:

I don't know. If I want to compare card x against card y I have to apply exactly the same parameters to the test. That's done by using the same scenario. The only thing that has changed is the card.

I notice in your test you're changing shadows, cloud resolutions etc. You say it isn't scientific but more of a feeling for how it's performing.

I believe the test I do takes out all the subjectivity as no settings are changed. It's purely down to how one piece of hardware compares to another all other things being unchanged.

Fair enough Ray.  

13 hours ago, Ray Proudfoot said:

Bob, if the 2080Ti is 10-15% faster than two 1080Ti cards in SLI mode how fast is it against one 1080Ti? Few people have two expensive cards to run SLI. More typical is a single card user. Testing like with like would be more helpful.

That 10-15% number isn't mine.  The testing Rob Ainscough and others did some time ago concluded that in P3D there was something around a 30% performance increase from a single 1080Ti to two in SLI based on frame rates. Problem is, I run with frame rate locked at 30, and instead of watching for a delta in frame rates, I watch for GPU load and the transition point into the stuttering regime during a controlled test scenario as the measure of how well the card is performing.  The 2080Ti was able to match the 1080Ti SLI setup by keeping the test scenario smooth, and with a lower GPU load, and when increasing the GPU load (e.g. with resolution, DL, SSAA/SGSS AA) it resists dropping below the VSync lock at 30fps and/or reaching the stutter threshold better than the 1080Ti SLI config did, and far better than a single 1080Ti did. 

So given generally-accepted controlled tests that conclude 1080Ti -> 2x1080Ti is a ~30% improvement, and my subjective observations that the 2080Ti bests the SLI setup, I think transitive logic supports my conclusion that the 2080Ti produces a performance increase over a single 1080Ti somewhere out beyond the 30% mark, anyway.

I understand that some folks want hard metrics--I just don't find it worth my while to spend the significant amount of time needed to do the controlled testing needed to determine those numbers.  If you were to jump into a VW bug and drive it a few times around the block, then jump into a BMW and drive it a few times around the block, you may not be able to quantify exactly how much better the BMW accelerates or takes a corner, but you still can correctly observe that the BMW is considerably faster and more nimble.  That's all I'm trying to do...let folks know that I observed that the 2080Ti easily outperforms my old 1080Tis, and even a pair of them--and that the difference is non-trivial.  But if you want hard numbers, I'm not your guy.

Regards

 

That's fine Bob, each to their own. I'm more than happy with the performance of my 1080Ti as I get around 20fps inbound to EGLL from the west with a couple of hundred Ai. I have no need for the extra power of the 2080Ti.

Maybe when Nvidia bring out their next gen of cards they'll be more reasonably priced. Or the UKP will be stronger against the dollar making it better value for money.

9 hours ago, Ray Proudfoot said:

Maybe when Nvidia bring out their next gen of cards they'll be more reasonably priced.

If you recall the 1080ti prices were very high also due to the crypto mining demand. It wasn't until just before the release of the 2080 series that the prices of the 1080tis dropped to reasonable. As soon as the high pricing on the 2080s were announced the 1080 prices shot back up. I consider myself very fortunate to have purchased a 1080ti during that brief dip in prices. Nvidia cards are the top dogs now and Nvidia knows it, so they set prices to what the market will bear. It is going to take a competitive video card from AMD to bring prices down.

Regarding your test methodology. My 1080ti replaced a 780GTX. The 1080ti did not increase my FPS but it did allow me to maintain my FPS during the tough graphics conditions like stormy overcast weather. I noticed that when flying into bad weather the GPU usage would peg at 100% and then the FPS would drop significantly. With the 1080ti this no longer occurs. The 1080ti has been well worth the purchase just for this. I also have been able to turn up other graphics settings without any impact on the FPS. With the 1080ti I have never seen a GPU usage over 50% but I believe that my CPU is now the bottleneck on my system. When my 8086 system is built this may no longer be the case and I may then be in the market for a 2080ti. As Nick Needham says, I'll be back on the upgrade hamster wheel.  

Ted

 

 

Hi Ted,

I knew about the prices of the 1080Ti rocketing when the crypto guys started buying them in droves. But I wasn’t in the market back then so didn’t check the prices.

But around Sept I decided if I wanted a good price for my old i7-4770K computer with a 1080 in it then that was the time to move. I got a good price and used that money towards my new system that I’m delighted with.

When I can taxi around 90% of my 3rd party airports with frames in the high 30s and then see the frames locked at 60 when airborne that is nivarna.

I had the 780 on my system before I replaced it with the 1080 I bought from a friend. But back then I was running FSX and it made no difference to frames. That’s because FSX was holding things back. Went to P3D and the difference was brilliant.

Once the 3080Ti cones out that might be the time to get a 20 series card. They used to say you paid through the nose for the fastest CPUs. Seems it’s now graphics cards.

I can’t imagine what sort of scenarios require the power of a 2080Ti but maybe flying low and slow in built-up areas is more demanding than high level IFR stuff.

The UKP/USD exchange rate is pretty low these days which doesn’t help us Brits much. Prices are probably a lot lower in the US. That makes a big difference.

If your 1080ti is not maxed out with your 8086 that is good news to me as I will have the same system soon.

Ted

It’s around 70% running a UHD display at 60Hz and 35 at 30Hz.

I don’t like 30 as the mouse movement is quite jerky. I hope you’re going to run UHD as the card can do it with ease and panels look superb.

Ah, I see from your sig you already are.👍

Edited January 17, 20197 yr by Ray Proudfoot

3 hours ago, Ray Proudfoot said:

I can’t imagine what sort of scenarios require the power of a 2080Ti but maybe flying low and slow in built-up areas is more demanding than high level IFR stuff.

You are correct Ray.  My 1080ti struggled with overcast conditions flying GA low and slow.  Add some night lighting and weather effects it was game over.  Hence my upgrade path to the 2080ti.

6 hours ago, mpw8679 said:

You are correct Ray.  My 1080ti struggled with overcast conditions flying GA low and slow.  Add some night lighting and weather effects it was game over.  Hence my upgrade path to the 2080ti.

W hat's your processor Mark? I would have thought the 1080Ti could cope with most things. Maybe your sliders are further to the right than mine.

5 hours ago, Ray Proudfoot said:

W hat's your processor Mark? I would have thought the 1080Ti could cope with most things. Maybe your sliders are further to the right than mine.

Currently a 9700k at 5.1ghz.  

57 minutes ago, mpw8679 said:

Currently a 9700k at 5.1ghz.  

Okay, thanks.