ASUS Radeon R9 380X Strix 4GB

[evilacg]

Two games have huge performance issues on Windows 10 on 2 GB NVIDIA cards.
GTA V and Watch Dogs.
This seems to be an NVIDIA driver issue, on Windows 7 everything was fine, using the same settings.
AMD 2 GB cards are unaffected too.

@W1zzard
Please make this issues clearer on your review, some sites said TPU fake the results of 770 2GB & 960 2GB in Watch Dogs & GTAV.

 

[rtwjunkie]

@W1zzard
Please make this issues clearer on your review, some sites said TPU fake the results of 770 2GB & 960 2GB in Watch Dogs & GTAV.

Site names?

 

[evilacg]

Site names?

Chinese website like Chiphell & 百度ati吧 & 百度显卡吧
http://www.chiphell.com/thread-1424467-1-1.html
http://www.chiphell.com/thread-1424220-1-1.html

http://tieba.baidu.com/p/4169946121

This one I can't find the specific link, maybe deleted
http://tieba.baidu.com/f?kw=显卡

 

[W1zzard]

@W1zzard
Please make this issues clearer on your review, some sites said TPU fake the results of 770 2GB & 960 2GB in Watch Dogs & GTAV.

Why would I fake results to make them look so bad, if obviously I know pretty much exactly how I should fake them without anyone ever knowing?

 

[evilacg]

Why would I fake results to make them look so bad, if obviously I know pretty much exactly how I should fake them without anyone ever knowing?

Hmmm, Guru3d's GTAV with Windows 10 & Forceware 358.91 doesn't seem to have this issues, driver 358.50 causing this problem?
http://www.guru3d.com/articles_pages/amd_radeon_r9_380x_4gb_review,9.html

 

[W1zzard]

Hmmm, Guru3d's GTAV with Windows 10 & Forceware 358.91 doesn't seem to have this issues, driver 358.50 causing this problem?
http://www.guru3d.com/articles_pages/amd_radeon_r9_380x_4gb_review,9.html

What are their settings?

 

[medi01]

• 4 GB VRAM provides little benefit at 1080p/1440p

As a downside, and lack of HDMI 2.0 as the next line... Seriously, guys, how could that be reasonable "cons"?

And to power consumptions, how could results be so different? You have 70w more CARD consumptions, while others show +29w total system (380x vs 960):

http://www.hardwareheaven.com/2015/11/xfx-amd-radeon-r9-380x-review/

 

[minorisprit]

What are their settings?

“Our settings are as follows with very high quality, 16xAF, 2xMSAA and FXAA enabled.”

 

[evilacg]

What are their settings?

Seems like these settings and 960 managed to get average 37fps on 2560x1440 in their test:

Our settings are as follows with very high quality, 16xAF, 2xMSAA and FXAA enabled.

 

[W1zzard]

Seems like these settings and 960 managed to get average 37fps on 2560x1440 in their test:

I did some testing. Looks like "High Resolution Shadows" are causing the slowdowns because they increase GPU memory usage by a few hundred MB. Turning them off will just mask the problem for the 2 GB NVIDIA cards until another game is released that requires significantly more than 2 GB.

The red highlighted setting

GTX 960 1600x900 High Res Shadows On: 12.1 FPS
GTX 960 1600x900 High Res Shadows Off: 56.6 FPS

So what do you think? Should we use lower than highest settings to make sure some GPUs run better in our benchmarks?

 

[rtwjunkie]

So what do you think? Should we use lower than highest settings to make sure some GPUs run better in our benchmarks?

I don't believe so. By using the same highest settings for all cards, it's easy to see what different cards can do and what others are not capable of.

Common sense would dictate to a potential buyer that if the card they want only does 22 fps for example on ultra, then if they lowered srttings to high, with shadows at medium, they will have a perfectly playable experience.

 

[evilacg]

I did some testing. Looks like "High Resolution Shadows" are causing the slowdowns because they increase GPU memory usage by a few hundred MB. Turning them off will just mask the problem for the 2 GB NVIDIA cards until another game is released that requires significantly more than 2 GB.

The red highlighted setting

GTX 960 1600x900 High Res Shadows On: 12.1 FPS
GTX 960 1600x900 High Res Shadows Off: 56.6 FPS

So what do you think? Should we use lower than highest settings to make sure some GPUs run better in our benchmarks?

It's your call, my opinion would be pointing the facts that NV's card 2GB VRAM maybe a problem on max settings in future game.
Why AMD's 2GB card doesn't have this issues? Is their driver's VRAM management avoided this issues?

 

[EarthDog]

So what do you think? Should we use lower than highest settings to make sure some GPUs run better in our benchmarks?

I love your sarcasm.

I wonder though, do you get significant slowdowns with a 4GB+ card when you flip that setting on?

 

[W1zzard]

I wonder though, do you get significant slowdowns with a 4GB+ card when you flip that setting on?

Haven't noticed any. All my current benchmark data (including this review) is with High Res Shadows on

 

[EarthDog]

I would be curious to see the difference. Though I certainly don't expect a 4GB to show the same differences, I am wondering if its more than ram and even 4GB+ cards take a significant hit.

 

[Steevo]

Its interesting to note that the 285, a 2GB card from AMD does not suffer this same effect, perhaps due strictly to the way the driver allocates resources, or perhaps there is 970esque problem with memory since its essentially half of a 980 core on the 960. Any chance of getting a bandwidth checkup on the 960?

 

[geon2k2]

I did some testing. Looks like "High Resolution Shadows" are causing the slowdowns because they increase GPU memory usage by a few hundred MB. Turning them off will just mask the problem for the 2 GB NVIDIA cards until another game is released that requires significantly more than 2 GB.

....

The red highlighted setting

GTX 960 1600x900 High Res Shadows On: 12.1 FPS
GTX 960 1600x900 High Res Shadows Off: 56.6 FPS

So what do you think? Should we use lower than highest settings to make sure some GPUs run better in our benchmarks?

Bravo, I really appreciate that you came forward and defended the results. Obviously you should compare apples with apples and pears with pears and all the settings should be the same.

I do have a suggestion though, which will help with such cases in which one setting is messing up the results and which you might or might not consider.
You could add testing on a common resolution like1080p on medium settings. This will also be useful to individuals with less powerful hardware, which don't even dream to play at maximum settings.

 

[xorbe]

I do have a suggestion though, which will help with such cases in which one setting is messing up the results and which you might or might not consider.
You could add testing on a common resolution like1080p on medium settings. This will also be useful to individuals with less powerful hardware, which don't even dream to play at maximum settings.

I have been pointing this out forever at various sites. Good luck with that ... pm me if you want my personal thoughts on the matter.

 

[mastershake575]

The 7970ghz was selling for $280 exactly two years ago from today so this is pretty sad (we need a die shrink badly, two years have pasted and we have a new card offering slightly more performance at only $30 cheaper)

 

[ASOT]

A 280x same specs..so just a rebranding in my opinion!

New features are good,like FreeSync,Delta compresion,pwm..other than that NOTHING

AMD forgot to put HDMI 2.0/ Happy i still have 280x old boy

Better take a 290/970/390/X

 

[NC37]

A 280x same specs..so just a rebranding in my opinion!

New features are good,like FreeSync,Delta compresion,pwm..other than that NOTHING

AMD forgot to put HDMI 2.0/ Happy i still have 280x old boy

Better take a 290/970/390/X

Its a rebranding of the 285. Not the 280X. A 280X would actually beat this as the 285, even with GCN1.2, was very underwhelming in performance. Part of the reason why Fiji has been somewhat underwhelming too. It's based on the same bad designs that were implemented in the 285. Granted it's a large step up. But still, 285 was a weak GPU.

Either way, the 380 series is disappointing. The performance of the 380X does not fill the gap between 380 and 390. If they had released an even more stripped down Grenada and called it a 380X it would have been better. Watch nVidia releases the 192bit 204 on the 960Ti and it'll totally blow the 380 series out of the water.

 

[ASOT]

@NC37 Yes,a rebrand of 285 with specs almost of 280x