Cat's little mini-Adventure with a GTX1080(Fin)! Now with MOAR benchmarks!

[CAT-THE-FIFTH]

Maximum Geralt! MOAR The Witcher 3 testing!

Even though Geralt was still grumpy from his last encounters with The Wild Hunt I decided I needed to wheel him and his super animated locks out for another set of benchmarking.

Now lets start at qHD.



The GTX1080 is just over the 60FPS mark on average and the RX470 is around 32.83 FPS or almost half the speed - The Witcher 3 is not a VRAM heavy title,so it is simply the case that the RX470 is running out of grunt here.

Lowering the resolution to 1080p does help the RX470 a tad more though.



It is 80.44 FPS against 45.17 FPS,this time but if you do want to have the game at max settings and get 60FPS at 1080p I think you need at least a GTX1070 or a Fury X.

Power consumption for the RX470 varied at between 224W to 238W at qHD as opposed to the GTX1080 which around 245W to just under 280W at the wall.

[CAT-THE-FIFTH]

Dragon Age:Inquisition

I thought Miss sole survivor Inquisitor lady probably need to go for another stroll round the ramparts,after all being stuck in that War Room can't be that great for her health,can it??

Now let's start with the qHD benchmarks again.



Instead of a PCMR 65.85 FPS experience with the GTX1080 you get a cinematic 33.71 FPS experience with the RX470 instead,together with more HD fan noise to boot. Power consumption varied between 235W to 243W instead of the 243W to 276W at the wall for the GTX1080.

Now,some of you Eagle eyed folk might ask,but CAT doesn't the game support Mantle?? Yes,you would be right,but since none of the software I had could be used to record FPS and since the AMD software package didn't seem to have any obvious ways to do it,I decided to test the card in DX11. Mantle might help but I don't have the eye sight of a Pigeon.

Considering qHD was a bit too much for the RX470,how about 1080p you might ask??



The GTX1080 runs through the game at 83.16 FPS,and the RX470 almost gets there at 51.06 FPS.

Considering this is the closest the RX470 gets to 60FPS in the titles I tested so far,lets see if some overclocking might help.

AMD,has radically changed its menu system,so I had to consult the very useful HardOCP review of the RX470 Devil:

http://www.hardocp.com/article/2016/...cking_review/3



So I settled on +115MHZ for the core and +100MHZ for the memory.

Now lets see if it makes much difference.



The FPS went from 51.06 FPS to 55.93 FPS,which is almost 10% extra but not quite enough to get to 60FPS either. The power consumption went up by around 30W to between 265W to 276W,which is what the GTX1080 was consuming at qHD whilst pushing more FPS.



As you can see here,at 90% fan speed,which was very noisy,the core was only at 65C,and GPU utilisation was at 100% and the core had no problem maintaining its overclock.

[CAT-THE-FIFTH]

LARA doing even more traipsing around!

I must ask her what brand of walking boots she wears....!

Lets look at the internal benchmark before looking at the time runs.



At qHD the RX470 produces around 40FPS with the same caveats as seen with the GTX1080 testing. The GTX1080 has double the score.



At 1080p things do improve and the RX470 is just shy of the magical 60 FPS mark. The GTX1080 is just under 100 FPS in the same benchmark.

However,the RX470 does seem to have more of the very quick framerate drops when compared to the GTX1080 even at 1080p.

Now,I left the best for last - Rise of the Tomb Raider seemed very taxing in the Village of the.....Chickens,so lets see how the RX470 stacks up. Will it be clucked into oblivion??

Baby steps first,lets see how DX11 performance is for the RX470.

First the qHD results.



The RX470 produces a sub-cinematic 19.84 FPS against the 51.32 FPS of the GTX1080. The RX470 probably ran out of VRAM. Power consumption varied at between 134W to 238W which kind of supports this theory.

Lets look at the less VRAM taxing resolution of 1080p.



The RX470 does a bit better,producing a framerate of 27.84 FPS against the 54.51 FPS of the GTX1080 - VRAM usage for the RX470 peaked at just under 3.9GB of VRAM.

Sorry for the slightly inconsistent time-runs,framerates got so low for the RX470 it was hard to control the mouse.

Now for some DX12 testing - will it save the RX470??



At qHD there seems to be no improvement.



At 1080p it is the same - remember the software I use for DX12 benchmarking has a worse performance hit than FRAPS. However the GTX1080 went from 95 FPS to 113 FPS,whereas the RX470 got slightly worse.

Now I started the Village benchmark run in DX12 on the RX470,and the FPS was hovering around the 20FPS mark,so it wasn't really doing much.

Hence,I quickly went to 1080p instead.



The RX470 did see a decent improvement from 27.84 FPS to 36.67 FPS which was a 32% improvement,and even though the deviations were worse on the GTX1080,you can still see some variability in results,ie,sometimes framerates would drop for no reason.

Now,the GTX1080 improved from 54.51 FPS to 66.81 FPS overall which was a 22.6% improvement,but if you took the outliers out the average increased to 70.23 FPS,which was a 28.8% improvement.

However,as a percentage of the average framerate,the RX470 did seem to suffer much worse drops though down to 10 FPS to 20 FPS. The GTX1080 in the worse case scenario was just under 40 FPS at its lowest point.

Summary Re-dux!

Having tested out the Powercolor RX470 Devil 4GB,its not a bad card for its segment,but when you look at things like the power consumption of the GTX1080 and the acoustic profile of the GTX1080,it really does seem a step down especially when you consider the fact,even in its weakest moments,it can still give an overclocked RX470 4GB a good thrashing.

Its one thing AMD probably needs to learn from Nvidia - a half decent blower cooler on its cards,since it really is useful for SFF PC users like me.

[kalniel]

Thanks for the review Cat. Can I ask some questions about the green teams image quality and drivers? How do you rate them, primarily? Are the AA modes and texture filtering options good? And do they have anything like AMD's negative LoD bias when using adaptive AA to keep textures sharp?

[CAT-THE-FIFTH]

Thanks for the review Cat. Can I ask some questions about the green teams image quality and drivers? How do you rate them, primarily? Are the AA modes and texture filtering options good? And do they have anything like AMD's negative LoD bias when using adaptive AA to keep textures sharp?

Well this will be the first AMD card I have tried out since an HD7850,since I had a GTX660 and a GTX960 after that. Is there a good way to test this,so I can give a definite answer??

[CAT-THE-FIFTH]

Possible explanation for DX12 inconsistencies!


Right,I started mucking around with some settings - I think I can explain some of the weirdness I saw under DX12. It looks like if I turn down texture settings when I hit the runs which are a bit crap things get much better. I have a feeling there is some intermittent issue between the very high texture setting and DX12,which isn't apparent under DX11 in the area I am testing. However,looking at VRAM usage under DX12,it is a bit higher than DX11,and under DX11 at qHD at 1080p it is between 7GB and 7.5GB,and DX12 mode seems to be a bit higher than that. It would also explain why at 1080p the issue happened less often,as VRAM usage should be a bit less overall.

[bsodmike]

Wow, now that's an amazing bit of work CAT. My stock Zotac founders 1080 also hits over 2GHz with a small OC (similar to yours) and power draw from the wall is in the ~310W range as well.

[kalniel]

Well this will be the first AMD card I have tried out since an HD7850,since I had a GTX660 and a GTX960 after that. Is there a good way to test this,so I can give a definite answer??

Hmm I guess try some of the different AA and texture modes and see what happens Novigrad in TW3 might be a good test scene as plenty of objects with jaggies yet also high res textures to try and keep sharp. Skyrim and trees/vegetation is another one. You may encounter some issues with various AA modes not making any difference where a game uses non-forward rendering - sometimes drivers are clever and manage to inject some form of AA in there, so it's a good test of the driver teams ingenuity, but where that fails the only remaining options are post process AA methods like FXAA which blur textures, and super sampling/downscaling.

[CAT-THE-FIFTH]

Hmm I guess try some of the different AA and texture modes and see what happens Novigrad in TW3 might be a good test scene as plenty of objects with jaggies yet also high res textures to try and keep sharp. Skyrim and trees/vegetation is another one. You may encounter some issues with various AA modes not making any difference where a game uses non-forward rendering - sometimes drivers are clever and manage to inject some form of AA in there, so it's a good test of the driver teams ingenuity, but where that fails the only remaining options are post process AA methods like FXAA which blur textures, and super sampling/downscaling.

I will have a look,but I need to get to Novigrad first!!

Edit!!

It might take a few days though,due to stuff.

[CAT-THE-FIFTH]

Wow, now that's an amazing bit of work CAT. My stock Zotac founders 1080 also hits over 2GHz with a small OC (similar to yours) and power draw from the wall is in the ~310W range as well.

Thanks - good to see that my numbers are kind of confirmable regarding clockspeed and power consumption.

[outwar6010]

https://youtu.be/4aM8CMSPfWQ?t=438

Seems like this thread has gone mainstream