Review: MSI Radeon R9 390X Gaming 8G OC

Techpowerup says at 1080p its only slightly more powerful than the GTX 970 by 2% .WTF

techpowerup.com/reviews/MSI/R9_390X_Gaming/30.html[/url]


"The MSI R9 390X Gaming roughly matches GTX 980 performance in 4K and 1440p when averaged over all our benchmarks. At lower resolutions like 1080p though, it is only as fast as the GTX 970"

The GTX 970 got as low at $300 at a time

 

To be fair AMD has made it very clear that they have "created" this card for 4K gaming (we are in all fairness moved further into the future than 290x), and this is especially true when you combine two cards which you probably will need anyway as an enthusiast too get the desired frame rates. I wish for their own sake they would have priced lower though, but prices will likely fall soon to levels people will be more comfortable with.

If they do this these cards will be fairly future proof if you buy two and the issues surrounding multi-GPU might be far less existing with the moving over to D3D12.

 

All you have to do is look at the Project CARS or Wolfenstein: NO results and conclude AMD have some serious performance issues here and there that skew the average result.

 

hmm curious to why there is so much difference in power consumption on different review sites.

Some show the promised improved power consumption in typical gaming while others show 50W more!

 

Could be rig whole power consumption or GPU only.

 

Yeah, but if you do play those games (and I have both) then you're in for a world of pain. Certain engines will run terribly on AMD, I'm afraid. IdTech is one of them. pCARS is a nvidia advert, they used nvidia physx that can't be done on AMD gpu so you'll always see AMD struggle there, nothing you can do.

 

AMD has always had abysmal OpenGL support. This is going back like a decade now. I remember struggling to run Nexuiz on a 4850. My previous 8600GT got better frames in that game at higher settings.

I'm not sure why it's never been a priority for them to fix it, but I guess they just said "eh, most games aren't OpenGL anyway" and never got around to it. Same thing with Linux support.

 

This is normal for all cards due to silicon quality, binning, which affects the voltage applied (similar to CPUs).

 

Just did your homework, next time do it yourself before posting.

You linked percentage. They have it wrong as I went through all their benches and here are only those where it is at 970 level.
CoD:AW - 137fps vs 135fps for gtx970
GTA5 - 63fps
Project CARS - 64 fps vs 89fps for gtx970 (this is what affects that average result)
Witcher 3 - 58fps vs 58fps for gtx970 (considering friendliness of gameworks it is good result)
Wolfenstein: NO - 65 vs 97fps for gtx970 (this is what affects that average result)
WoW: 135fps vs 174fps - (this is what affects that average result)

Conclusion:
Test has 22 games, in most of them it performs around 980 level, in some it is in middle of 970/980
Then you have few games where it performs at gtx970 level, but those run at around 140fps+ anyway.
And then entire result gets skewed by 3 games from which 2 are borked by nVidia, and one simply runs flawlessly on both r9-390x/gtx970 and stronger (WoW).

So unless you play only Project CARS and Wolf: NO, then r9-390x will be close match to gtx980. (but 980 may prove to be cheaper card in few years of use)

 

Yes, OGL is badly supported, not to mention Linux drivers (fglrx), but that is not priority (obviously), it works, it is better, but...

Oh, and look at those 1080p performance, ofc it is irrelevant when GPU does 150 FPS, but it's not like magically nvidia cards become faster at that point, it is in fact "CPU limitation", because of driver code is too heavy, and it hinders performance on lower resolution where GPU can push much more.

 

One thing AMD does fairly well is dropping prices quickly so in a few months you will probably be able to get a 390X for $350 or less and even though I really need a new GPU right now I can wait a bit longer. Haven't had a Nvidia card in a desktop sine my 570's but no problem going back if AMD doesn't do a good price slash.

 

I must say that I'm a bit surprised about the power consumption test. In every other MSI R9 390X the power usage has been much larger than reference 290X. Both Techpowerup and Hardocp got way higher numbers. What would explain the large difference?







By the way Hilbert, if you could save a lot of bandwidth by using PNGOut to optimise those png charts.

 

were those TPU and Hardocp power draw numbers with an overclocked cards? 290X? Those cards' power draw can skyrocket when overclocked.

 

Same cards. I don't know what modes (silent/gaming/oc) they used though.

 

The cards may not have been cooled properly (low fan setting, or Auto), or it may have been tested inside a case, or with warmer ambient temps.

 

HardOCP is showing total system power draw, not 390x power draw. TPU power draw seems off though, almost 100w more than a 290x, doesn't seem right. Maybe it's a typo, maybe they made a mistake measuring it or they just got a really power hungry sample and overclocked it and upped the voltage and that's what they're showing?

 

Spoiler

344w wow c:
I was going to buy a R9 390 but i change my plans, better getting R9 Nano due to low power 175w. i had a 970 GPU & it ran cool & it use low power & it ran on 500w PSU .

 

What is more worrisome is the multi-monitor usage and the Blu-ray one... Using this for a multi screen setup will actually raise your power bills by ALOT even if you let it idle alot. I wonder if there is a BIOS/driver fix planned. It shouldn't be consuming this much for simply being hooked up to another monitor at idle (vs the 290X too).

 

They noticed that the stock voltage was higher than on 290X. It could be that Hilbert got really good sample. If I remember correctly Techreport (edit: was another site; Hardocp and Techpowerup at least) noticed same high power usage like Hardocp and Techpowerup. Hardocp did the power testing at least twice because they couldn't believe the results.

 

Or they got sample of non-tweaked Hawaii which someone soldered on new PCB.