Mirror's Edge Catalyst 'Hyper' Settings Made Possible by 8GB VRAM

[Bacon1]

Some people will say yes, others a big no.

Depends on your e-peen IMHO.

You can always figure out a way to use more. My guess is AMD will probably keep steady at 4gb for the next few years, while nvidia will try to push new options which will require 8gb for very slight improvements.

That being said, if the 8gb version of the RX480 is indeed only $40 or $50 more... my answer would be... why not?

Well really, $50 is 25% of the cost of the card, $40 is 20%.

The only reason would be to use it for 1440p, or dual card setups. Most games you wouldn't have the power to run them with all the eye candy that the 4gb+ will require. I mean look at the 980, it is limited to 4GB but doesn't seem to restrain it. 480 will have similar power.

970 did have jacked up ram so has had issues.

 

[mohit9206]

Yeah the first Mirrors Edge was not to my liking. Piss poor story, repetitive platforming and terrible combat segments. So I won't be playing this one as its more of the same.
Also as saying goes better eat Ramen for a week if you have to but don't settle for 4gb if 8gb model is available.

 

[Thala]

That being said, if the 8gb version of the RX480 is indeed only $40 or $50 more... my answer would be... why not?

Because chance is, that this could be wasted $40.

 

[Yakk]

Well really, $50 is 25% of the cost of the card, $40 is 20%.

The only reason would be to use it for 1440p, or dual card setups. Most games you wouldn't have the power to run them with all the eye candy that the 4gb+ will require. I mean look at the 980, it is limited to 4GB but doesn't seem to restrain it. 480 will have similar power.

970 did have jacked up ram so has had issues.

I don't disagree which is why I mentioned e-peen. 4gb is fine in most cases for 1080, I've run 3gb in CF for years now, even at 1440 without issue (with some options turned down, yes). Though some people can pay thousands of dollars to try and satisfy their e-peen, and yet it always wants more...

 

[3DVagabond]

You guys think 4gb is enough on the 480 for 1080p gaming?

As long as the game doesn't include settings specifically aimed at making 1080p users to upgrade their cards. Since most people play @ 1080p how else are they to get enough people to buy their shiny new 8gig cards? BS settings to intimidate them into it.

Don't worry 4 gig @ 1080p is fine.

 

[imaheadcase]

I was kind of surprised we did not see 16gig cards as a option with new pascal cards. Maybe they are reserving that for Ti or refresh.

 

[Sweepr]

GameGPU results for the 'retail' version of the game out. Unfortunately they didn't test at Hyper settings. At Ultra you only go above 4GB VRAM at 4K.

http://gamegpu.com/action-/-fps-/-tps/mirror-s-edge-catalyst-test-gpu

Both Radeon 290X and Geforce GTX 980 suffer more in comparison to Geforce GTX 980 Ti at 4K (70-80% Geforce GTX 980 Ti performance at 1440P vs 60-65% at 4K).

 

[moonbogg]

Wow, that 980ti had a short life not to mention all the other cards. An Nvidia flagship card can only max games for 12 months it seems. 980ti barely a year old and already can't cut the mustard in this game on max. LOL. What a joke. 7.5gb required Vram is ridiculous.
Being a leet PC gamer is too expensive these days. I think I'll just become an audiophile instead. Buy headphones once and done.

 

[Bacon1]

Wonder why Gamer's Nexus Nvidia cards are all much, much, faster than gamegpu while the AMD cards are much much slower...

970: 44 => 64 = 45% faster

980 Ti: 65 => 92 = 42% faster

290x (390x missing ram and lower speeds): 40 min => 23 min = over 40% perf decrease. Hard to tell upper limit since obviously different clocks and VRAM, but higher clocks and more ram should never decrease performance.

Can we agree that the GamersNexus numbers look completely wrong?

 

[IllogicalGlory]

-snip-

 

[Silverforce11]

You can tell GameGPU is accurate here, based on the 280X vs 780 scenario and 290 vs 780Ti. That's the current trend of major cross-platform games.

Also very lame that AMD does not have a CF profile for this game. Really it's Frostbite, how hard is it to support it out of the box already?

 

[kraatus77]

You can tell GameGPU is accurate here, based on the 280X vs 780 scenario and 290 vs 780Ti. That's the current trend of major cross-platform games.

Also very lame that AMD does not have a CF profile for this game. Really it's Frostbite, how hard is it to support it out of the box already?

IF it's a gw game, they won't be able to do anything untill games releases. and later you will see better performance.

 

[Bacon1]

Have any reviews covered the IQ difference between Hyper and Ultra yet?

 

[Sweepr]

Have any reviews covered the IQ difference between Hyper and Ultra yet?

PCGamesHardware has some pictures: www.pcgameshardware.de/Mirrors-Edge-Catalyst-Spiel-55642/News/Hyper-Grafik-Vergleich-Ultra-1197795

 

[Bacon1]

http://www.computerbase.de/2016-06/mirrors-edge-catalyst-benchmark/

That shows it, and also shows results closer to gamegpu, which is the normal/expected ordering and not gamersnexus's.

Fury X is actually above 980 TI @ 1440p/4k, even passing 1070 @ 4k.

390 / 970 are neck and neck @ 1080p, with the 390 pulling ahead @ 1440p.

Something is wrong with GamersNexus benchmarks.

Please update the OP.

 

[Sweepr]

Please update the OP.

Yes, it will be updated with this quote:

The Radeon R9 390 easy to lose 25 percent higher and the Radeon R9 Fury X 27 percent. The latter has the problem that the four gigabyte memory is no longer sufficient. The FPS-loss thereby is indeed low, but the game not running smoothly.

 

[BlitzWulf]

http://www.overclock3d.net/reviews/gpu_displays/mirror_s_edge_catalyst_pc_performance_review/4

 

[linkgoron]

http://www.computerbase.de/2016-06/mirrors-edge-catalyst-benchmark/

That shows it, and also shows results closer to gamegpu, which is the normal/expected ordering and not gamersnexus's.

Fury X is actually above 980 TI @ 1440p/4k, even passing 1070 @ 4k.

390 / 970 are neck and neck @ 1080p, with the 390 pulling ahead @ 1440p.

Something is wrong with GamersNexus benchmarks.

Please update the OP.

This is a Sweepr thread. You should expect the worst AMD results possible in the OP, even when other sites show quite different results.







Personal attacks are not allowed.


esquared
Anandtech Forum Director

 

[Sweepr]

This is a Sweepr thread. You should expect the worst AMD results possible in the OP, even when other sites show quite different results.

ComputerBase reduced the details not to hit VRAM limitations, the thread is about how current VGAs manage to run the game at Hyper setting:

The graphic settings in the benchmark
1920 × 1080 Hyper , texture Details Ultra
2,560 × 1,440 Ultra
3,840 × 2,160 High

And their impressions about Hyper settings are:

Who wants hyper textures, requires a minimum of 6 GB graphics card memory

It requires at the "Hyper-textures" in Full HD at least a graphics card with six gigabytes and from 2,560 × 1,440 then must it be eight gigabytes.

... Who uses Ultra, on the other hand comes with four gigabytes in 1,920 × 1,080 (two gigabytes is not enough), the same applies to 2,560 × 1,440. But for Ultra HD, it must also be more, otherwise show up from time to time unpleasant stuttering.

...The Radeon R9 390 easy to lose 25 percent higher and the Radeon R9 Fury X 27 percent. The latter has the problem that the four gigabyte memory is no longer sufficient. The FPS-loss thereby is indeed low, but the game not running smoothly.

Now please carry on your personal crusade.

 

[RussianSensation]

This is a Sweepr thread. You should expect the worst AMD results possible in the OP, even when other sites show quite different results.

Sure seems odd the poster is so attached to using GamersNexus (historically NV-biased, which is why almost no one posted anything from them on here in 5+ years) and PCLabs (the most NV-biased European GPU review site ever! On there, NV practically wins in 99% of cases). In light of GameGPU's and Overclock3D's score, GamersNexus benches make no sense which shouldn't be surprising to anyone who has followed that site's reviews for years -- they were never reliable. GamersNexus has 970 beating Fury X at 1440p on Ultra, not even on Hyper settings but Overclock3D shows Fury X beating 980Ti. But then I see 64 fps average for 970 and only 70.5 fps average for 980Ti at Overclock3D at 1440p. Who to believe....

BlitzWulf beat me to it but I'll post the scores.

Drivers

When testing Mirror's Edge: Catalyst we will be using the newest GPU drivers that are currently available, which is Nvidia's Game Ready Geforce 368.39 driver and AMD's 16.6.1 driver. Both of these drivers feature game specific optimizations for Mirror's Edge: Catalyst

"At higher resolutions and graphical settings AMD's Fury X offers over 10% more performance than Nvidia's GTX 980Ti, though both GPUs are perfectly playable at even Hyper settings."

It's only once we get to 4K that 4GB becomes an issue but the 980Ti is slow anyway.

 

[Bacon1]

ComputerBase reduced the details not to hit VRAM limitations, the thread is about how current VGAs manage to run the game at

Ok, if you aren't going to keep a valid OP I've created a new thread for actual benchmark discussion and not just being about Hyper settings.

Please move non-hyper setting discussion to the new thread: http://forums.anandtech.com/showthread.php?p=38280745

 

[Sweepr]

Ok, if you aren't going to keep a valid OP I've created a new thread for actual benchmark discussion and not just being about Hyper settings.

Please move non-hyper setting discussion to the new thread: http://forums.anandtech.com/showthread.php?p=38280745

Fine. Nice showing by the Radeon(s) in ComputerBase, as long as you use the right settings.

 

[swilli89]

Now please carry on your personal crusade.

I've actually found I enjoy these forums when I just let both sides continue to try to dog eachother, but in ALL fairness, looking at your post history, you absolutely jump at the chance to post negative news articles about AMD while leaving out the positive ones.

I remember your threads from 2012-2013 and I would always expect a good CPU news article without thinking of any bias.

In your latest streak of posting dating back to March of this year, when you post about AMD its exclusively bad news items and when its been about nVidia/Intel its been exclusively good. Maybe that's just a giant coincidence, and I'm willing to give you the benefit of the doubt that you are just posting news as it becomes available. Just seems that your selections are pretty far off the regression curve if I'm being totally honest.

 

[Sweepr]

To clear up some confusion. Both Gamers Nexus and ComputerBase managed to run the game at Hyper @ 1080P/1440P with 4GB VGAs (same as Overclock3D), the latter even got 32% better average FPS with Fury X compared to an 8GB Radeon R9 390. The problem according to them is instability and not running smoothly, despite what the framerate suggests.

The Fury X and GTX 970 refused to run with any stability – we think that's a VRAM limitation – and so the chart only shows the 390X, 1080, 980 Ti, and 1070. All of these cards have 8GB of VRAM, except the 980 Ti and its 6GB.

The Radeon R9 390 easy to lose 25 percent higher and the Radeon R9 Fury X 27 percent. The latter has the problem that the four gigabyte memory is no longer sufficient. The FPS-loss thereby is indeed low, but the game not running smoothly.

GameGPU goes one step further, claiming even Geforce GTX 980 Ti has problems:

Maximum setting Hyper in the test have not been used, since their activation RAM and memory consumption increases many times that even in failure for the GTX 980 Ti with its 6 gigabytes

Also, ComputerBase found out something interesting:

By default, a protective function in the option menu is active, which turns off some details automatically when memory shortage. The function is switched off for all tests.

 

[Silverforce11]

I have to question the point of Hyper settings here, on my 4K monitor with those screenshots, the only major difference is the roads in the distance are more reflective. However at such a distance, the reflections are mere blobs of the car's headlamps.

I would much prefer had they gone for 4K textures for all the game assets on that setting, it would actually massively benefit image quality and has a excuse to use more vram.