R9 390 performance.

[IEC]

However, I did see this [H] comparison though, and the performance gains are negligible...currently.

http://www.hardocp.com/article/2015/06/18/msi_r9_390x_gaming_8g_video_card_review/9#.VaaA5vlVhBc

The problem with [H]'s comparison is they downclocked the 390X to 290X speeds. At least some (maybe all?) of the "clock for clock" (@core) gain appears to be due to selecting different GDDR5 chips with a higher spec and having a higher memory clock out of the gate. Hawaii, contrary to what your average forum poster will claim, can certainly benefit from increased memory clock speeds (especially when paired with a core OC). The 390/X is capable of memory speeds none of my 290/Xs were simply due to having better GDDR5 chips (on average).

 

[Mondozei]

Interesting thread, and thanks for the benchmarking, OP.

Just a note on the 780 guy re: Witcher 3 benchmark. It ought to be mentioned that you have a 780 Lightning Edition card, which is hardly typical for most aftermarket cards. Secondly, Witcher 3 as of right now has a decidedly pro-NV bent, like GTA V.
Whose fault is that? I don't know, maybe AMD needs better drivers to close that gap, maybe it's gameworks messing with things(even with hairworks off).

If you'd take a more neutral game like Crysis 3 you would see the performance delta being much larger. And of course games like Hitman Absolution or CoH 2 are totally pro-AMD in their benchmarking, which similarily is why many serious review sites include at least one of them to counter-balance GW games.

Also, downclocking a 390 for "clock for clock" comparisons is pretty dumb. The 390 overclocks a lot better and how many gamers aren't going to push the envelope to the maximum amount that they can to get all the performance out of it? This is why I like the jayztwocent's method of always overclocking all cards to the max and then comparing them from there, because that is much more realistic of real world behaviour.

EDIT: Here is Sweclocker's 1440p performance index.

Unclocked, yeah, so the rankings of the 390X vs 980 are not very natural, but should still kill all the "rebrandings" arguments precisely because they are stock speeds. The small natural OC for the stock speeds don't justify almost 15% more performance a 390X has over a 290X, despite being supposedly the "same card".

 

[Chaotic0ne]

Just a note on the 780 guy re: Witcher 3 benchmark. It ought to be mentioned that you have a 780 Lightning Edition card, which is hardly typical for most aftermarket cards. Secondly, Witcher 3 as of right now has a decidedly pro-NV bent, like GTA V. ebrandings" arguments precisely because they are stock speeds.

It looks like he's got his 780 water cooled, and if hes got a crazy overclock on that 780 (like +40%), he might be ahead.

 

[x3sphere]

Unclocked, yeah, so the rankings of the 390X vs 980 are not very natural, but should still kill all the "rebrandings" arguments precisely because they are stock speeds. The small natural OC for the stock speeds don't justify almost 15% more performance a 390X has over a 290X, despite being supposedly the "same card".

I assume that's a reference 290X though... it throttles heavily. Non-ref cards like the Tri-X don't have this issue.

 

[amenx]

So far I'm pleased. Witcher 3 performance is very good. I'm using version 1.06, with catalyst 15.7 drivers.

Specs:

I5 2500k @ 4.0ghz
MSI R9 390 gaming 8G - OC'd to 1200 core/1600 memory (not bad)
16GBs DDR3 1600mhz
Windows 7 64 bit.

Witcher 3 - 1080p - full ultra - all post processing enabled.

Getting 45-55fps on average in most areas. If I turn hairworks off, and set foliage to high, I can up to 75-80fps with minimums above 60. I get FPS dips with Vsynch on, unfortunately.

With settings set to low, all post processing disabled I can get up to 135fps with the minimums in the 110fps range.

Compared to the benchmarks I'm seeing here:

http://www.guru3d.com/news-story/the-witcher-3-benchmarks.html

I think the R9 390 does very well. I'm averaging better FPS than the 980 listed and that benchmark had SSAO instead of HBAO+ and hairworks turned off. With both those off I'm in the 60fps range.

My conclusion, the R9 390 is a beast. Awesome buy for $320, and it smokes the GTX 970 in the same price range.

Can't wait to bios flash it to a 390x.


Dragon Age Inquisition on the other hand runs like complete crud. Its hardly even playable. It'll chug along at 60fps then bam, dips into the 9-15fps range; massive stutter, and it does it every now and then like its on a timer. Dunno wtf is up, but AMD needs to get that fixed. It runs so bad right now, that the game ran smoother on my old Radeon 6970. Again, its not the GPU, its Bioware's crap game, and AMD's crap drivers.

Depends which set of benches you look at. Guru3ds own benches:

http://www.guru3d.com/articles_pages/the_witcher_3_graphics_performance_review,5.html

and TPU benches more in line with guru3ds benches.

http://www.techpowerup.com/reviews/Performance_Analysis/The_Witcher_3/3.html

 

[perithimus]

I am actually getting worse performance out of my 390 than I did with my 970. I could run the witcher 3 with everything at ultra except shadows on high, and hairworks off and get 50-60 fps. I dip into the low 40's with the 390 at the same settings. I have to turn things down to get the same fps that my 970 got.

 

[Face2Face]

The problem with [H]'s comparison is they downclocked the 390X to 290X speeds. At least some (maybe all?) of the "clock for clock" (@core) gain appears to be due to selecting different GDDR5 chips with a higher spec and having a higher memory clock out of the gate. Hawaii, contrary to what your average forum poster will claim, can certainly benefit from increased memory clock speeds (especially when paired with a core OC). The 390/X is capable of memory speeds none of my 290/Xs were simply due to having better GDDR5 chips (on average).

I haven't seen any testing on the benefits of increased memory bandwidth for the 390 series, but I have a hard time believing it's the only thing providing a 9% boost in performance @ 1440p.

I think it's a combination of couple of improvements. The 390X has better cooling (No throttling in quiet mode, like the 290X), 50MHz bump in core clock, and has faster memory with tighter timings vs. a reference 290X. All of these combined is enough for it be 9% faster than the 290X @ 1440p.

Just a note on the 780 guy re: Witcher 3 benchmark. It ought to be mentioned that you have a 780 Lightning Edition card, which is hardly typical for most aftermarket cards.

Yes, I have Lighting that clocks around 1300Mhz on the core. believe it or not, but a lot of Classifieds, HOF, and other high-end GTX 780's can do the same.

If you'd take a more neutral game like Crysis 3 you would see the performance delta being much larger.

Looking at AT's recent review, a reference GTX 780 boosting below 1GHz is almost as fast as a R9 290X in uber mode...

The 390 overclocks a lot better

I'm not convinced that it does... I think we need to some more cards floating around the forums before we can truly say this.

It looks like he's got his 780 water cooled, and if hes got a crazy overclock on that 780 (like +40%), he might be ahead.

Correct. Keep in mind the clocks in the video are 1293/1656, which are completely obtainable with the stock cooler @75-80% fan speed. In games like Shadow of Mordor and Metro LL, your card should pull well ahead. Sorry guys if I rustled some feathers, I was just making some observations on my end.

 

[Headfoot]

I am actually getting worse performance out of my 390 than I did with my 970. I could run the witcher 3 with everything at ultra except shadows on high, and hairworks off and get 50-60 fps. I dip into the low 40's with the 390 at the same settings. I have to turn things down to get the same fps that my 970 got.

Total side grade, I don't know why you'd switch between the two if you already had a 970. Was that because of the 3.5gb fiasco?

 

[Chaotic0ne]

I just did a benchmark run through Novigrad.

Settings:

Full ultra except for hairworks which is off, all post processing settings turned on/maxed. Vsynch off, FPS unlimited. No modifications to the Witcher 3 profile in CCC. 1080p resolution.

This bench was done @ 1170/1570 which is a mild overclock because its 1040/1500 stock.

Here is the route I took. I started at glory gate, took a right, and circled around through Hierarch square, and then back to Glory gate. I looked straight forward the entire time while using full sprint till I ran out of stamina, let the bar recover to full, then full sprint, repeat. I never opened the menu once, nor picked up any items, nor talked to any people, nor did I fight any mobs.

Here are my results with Fraps for the above run:


Frames, Time (ms), Min, Max, Avg
6781, 110604, 49, 71, 61.309

And that's running through some of the most demanding spots in the game. Outdoors, away from cities, I get considerably higher FPS.

My CPU did bottleneck in a few spots, so that has to be taken into consideration, particularly with regards to the minimum FPS. But the bottleneck is minimal, and it only occasionally hits 100% usage, while most of the time its hovering in the 80-90% usage range. Its safe to say a 4.5ghz overclock on an I5 2500k would eliminate any bottleneck in Witcher 3. Unfortunately for me, my CPU is a dud and it won't OC past 4.0ghz. DX12 is supposed to be more efficient with regards to CPU usage, so its unlikely I'll have to worry about bottlenecks in DX12 games on the CPU side in the near future. At least, I hope.

I dunno, 61.3fps average running through Novigrad with full ultra except for hairworks is pretty damn good for a $329 GPU. This is a new card, and the drivers are only going to improve with time, so there is lots of potential for the R9 390.

I'd like to include another benchmark I did to the above post.

This time I went with FULL Ultra including hairworks while using an I5 2500k @ 4.3ghz with the same overclock as listed above.

I started the route at 12:30 PM game time.

FPS here:

Frames, Time (ms), Min, Max, Avg
5636, 111837, 44, 55, 50.395

I am actually getting worse performance out of my 390 than I did with my 970. I could run the witcher 3 with everything at ultra except shadows on high, and hairworks off and get 50-60 fps. I dip into the low 40's with the 390 at the same settings. I have to turn things down to get the same fps that my 970 got.

That's way worse than what I'm getting. If I turn hairworks off and shadows to high I'm getting considerably higher performance than what you're claiming. I'll do that benchmark in a few minutes. Dips into the low 40s could be a CPU bottleneck, as explained above. Parts in Novigrad bottleneck an I5 2500k even at 4.3ghz in spots.

**Edit: Here is the benchmark for the same run with an R9 390 at stock clocks for this model or 1040/1500. Using the same settings that perithimus did. I started the run at 1:30pm game time.

Frames, Time (ms), Min, Max, Avg
6736, 110964, 52, 71, 60.704

Sorry bud, but I'm not seeing low 40s, I'm seeing 60fps average through one of the most demanding areas of the game. What CPU do you got?

Shadows appear to be one of the CPU intensive settings, because I got considerably higher minimums than usual by putting shadows on high. I might try medium shadows and see what happens. I honestly can't tell a difference between medium and ultra shadows.

Correct. Keep in mind the clocks in the video are 1293/1656, which are completely obtainable with the stock cooler @75-80% fan speed. In games like Shadow of Mordor and Metro LL, your card should pull well ahead. Sorry guys if I rustled some feathers, I was just making some observations on my end.

D: Wahhh!! My GPU got beat by someone elses on the internet, must rage hard and reply with a massive butthurt rebuttal!!

But naw, I don't care, I'm not a fanboy.

 

[perithimus]

I'd like to include another benchmark I did to the above post.

This time I went with FULL Ultra including hairworks while using an I5 2500k @ 4.3ghz with the same overclock as listed above.

I started the route at 12:30 PM game time.

FPS here:

Frames, Time (ms), Min, Max, Avg
5636, 111837, 44, 55, 50.395



That's way worse than what I'm getting. If I turn hairworks off and shadows to high I'm getting considerably higher performance than what you're claiming. I'll do that benchmark in a few minutes. Dips into the low 40s could be a CPU bottleneck, as explained above. Parts in Novigrad bottleneck an I5 2500k even at 4.3ghz in spots.

**Edit: Here is the benchmark for the same run with an R9 390 at stock clocks for this model or 1040/1500. Using the same settings that perithimus did. I started the run at 1:30pm game time.

Frames, Time (ms), Min, Max, Avg
6736, 110964, 52, 71, 60.704

Sorry bud, but I'm not seeing low 40s, I'm seeing 60fps average through one of the most demanding areas of the game. What CPU do you got?

Shadows appear to be one of the CPU intensive settings, because I got considerably higher minimums than usual by putting shadows on high. I might try medium shadows and see what happens. I honestly can't tell a difference between medium and ultra shadows.



D: Wahhh!! My GPU got beat by someone elses on the internet, must rage hard and reply with a massive butthurt rebuttal!!

But naw, I don't care, I'm not a fanboy.

My cpu is an i5 4690k. I am running it stock right now but the cpu jumps to 3.9 while playing the game. Interestingly enough I set the tessellation settings in the driver control panel to use application settins, and in the game options set the framerate to unlimited. I am now seeing 50-60fps averages again with a steady 60 when wandering around outdoors. I only saw it jump to 48 once but that might have just been a hitch.

I just ran around Novigrad and mainly in the area in the center where they burned the woman I only saw it dip to 54 fps.

 

[yacoub]

Still waiting for 390 and 390X to be added to the 2015 GPU Bench...

 

[3DVagabond]

Still waiting for 390 and 390X to be added to the 2015 GPU Bench...

They're still searching for "quiet mode".