[Review by ELP3] AMD Radeon FURY X

Fassou

Well, we have already separated the posts, and I think what I leave is material for all audiences, although some comments are on the edge of being passionate, but I suppose it can be tolerated.
**
Do we continue with the topic of the big review of ELP3?**, and no more putting in reviews of websites of chiquitistan, to discuss nonsense.

Cheers!

Handrox

Guru3D.com Forums - View Single Post - AMD Fury X Owners' Thread

RE: Firestrike + drivers = LoL
Firestrike-Link (Added new result - Graphical scores are just…)
On left: Fury X release driver for W8.1 with Fury X code path (15.150.0.0) - 15664
In middle W10 driver which does not have Fury X code path, so I used Tonga instead (15.200.1040.0.) - 16320
On right W10 driver which does not have Fury X code path, This time tested Hawaii instead (15.200.1040.0.) - 17296

And Yes, Graphical score went from 15664 to 16320 and then to 17296 which is more than Hilbert's 16081. (4.2% and then 10.4% improvement).
And at this moment I have highest Graphics score from all people who benched with Single Fury X, that's even against those overclocked.

Javisoft

@Handrox:

Guru3D.com Forums - View Single Post - AMD Fury X Owners' Thread

RE: Firestrike + drivers = LoL
Firestrike-Link (Added new result - Graphical scores are just…)
On left: Fury X release driver for W8.1 with Fury X code path (15.150.0.0) - 15664
In middle W10 driver which does not have Fury X code path, so I used Tonga instead (15.200.1040.0.) - 16320
On right W10 driver which does not have Fury X code path, This time tested Hawaii instead (15.200.1040.0.) - 17296

And Yes, Graphical score went from 15664 to 16320 and then to 17296 which is more than Hilbert's 16081. (4.2% and then 10.4% improvement).
And at this moment I have highest Graphics score from all people who benched with Single Fury X, that's even against those overclocked.

Por lo que me comentan los compañeros que tienen instalada la RC de win10 dicen que va mejor en todo y dan algo mas de puntuacion en test y juegos, quizas tenga algo que ver, pero es curioso lo del driver XD.

Handrox

Quote:
Originally Posted by Joschi View Post
has anyone managed to run Fury X on W10 Build 10130? bought mine, installed it, but the drivers won't install… only the basic Microsoft Display Driver. i've tried with 15.15 15.200.1040... same effect, GPU-Z 0.8.4 says Fiji

Yes I am running int on W10 build 10130.
And I am on 15.200.1040 which gives better performance than initial review driver.

How to install?

1st: you have to disable driver signature/conflict enforcement in command line running as dministrator:

bcdedit -set loadoptions DISABLE_INTEGRITY_CHECKS

bcdedit -set TESTSIGNING ON

2nd: modify inf file to include Fiji as device for which driver is:

Go to your unpacked driver: 15.200.1040\\Packages\\Drivers\\Display\\WT6A_INF\\

Edit: C0185174.inf

find line like this: "%AMD67B1.1%" = ati2mtag_Hawaii, PCI\\VEN_1002&DEV_67B1&REV_00
(it has to have ati2mtag_Hawaii there)

replace PCI\\VEN_1002&DEV_67B1&REV_00 with PCI\\VEN_1002&DEV_7300

Take 1st part if string: AMD67B1.1 and search for it

at bottom of file you'll find line like this: AMD67B1.1 = "AMD Radeon R9 200 Series"

change AMD Radeon R9 200 Series to AMD Radeon R9 Fury X
(or whatever you want your card to be named as)

3rd: reboot

4th: installation

Install CCC via standard setup as usual

go to device manager and chose manual installation for driver, navigate to modified inf file, select your card. Confirm that you want to use unsigned driver.

Guru3D.com Forums - View Single Post - AMD Fury X Owners' Thread

ELP3

Yes, but from what I see, the driver is different because the operating system is different. WIN 10.

I'm not going to put WIN 10 yet.

It is therefore not known if the improvement is from the driver, or simply from the OS.

A firestrike is a bench dependent on many cosas.No it is not just a bench of GPU.Es it is a bench of the complete system. Like the use of the CPU.Si WIN 10, it makes better use of this, that may be the gain.

In fact, in SLI for example, the improvement from using WIN7 to WIN 8 in firestrike. Is encrypted in almost 1500 points... that is... and it is simply due to the improvement of the CPU threads.

Handrox

Hey… look at that..

AMD Fury X Reviews - Page 27 - AnandTech Forums

wwwendigo

@Handrox:

Oye… mira eso..

AMD Fury X Reviews - Page 27 - AnandTech Forums

Vaya, AMD also sells its Fury X card as a rebranded, it seems to be a hypervitaminated GTX 970 at its core, hehe.

:osvaisacagar:

This has to be a joke or a program error. Or that or we have AMD in the same dock as NVIDIA.

Handrox

@wwwendigo:

Wow, AMD also sells its Fury X as a rebranded card, it seems to be a hypervitaminized GTX 970 at its core, hehe.

:osvaisacagar:

This has to be a joke or a program error. Or else, we have AMD in the same dock as NVIDIA.

What raises suspicion is the fact that in many reviews, the Vram usage of the FX didn't go beyond 3500MB and AMD already said that they are making a fix to unload things (data) from the Vram to RAM so as not to fill the 4GB…

Javisoft

@Handrox:

What I found suspicious was the fact that in many reviews the use of the FX's Vram did not exceed 3500MB and AMD already said that they are making a fix to unload things (data) from the Vram to RAM and thus not fill the 4GB…

I already commented that the highest use I had seen in the reviews was 3914 mb of vram in shadows of mordor 4K, what was being talked about in the anandtech forums makes no sense at all.

The fact of implementing HBM makes the use of MC considerably reduced and therefore there is more space in the die to put shaders (for example), partitioning the vram makes no sense, besides, that benchmark is quite unreliable …

Ciclito

@wwwendigo:

Wow, AMD also sells its Fury X as a rebranded product, it seems to be a hypervitaminized GTX 970 at its core, hehe.

:osvaisacagar:

This has to be a joke or a program error. Or else, we have AMD in the same dock as NVIDIA.

At minute 1':36'' onwards, look at those scratches…:facepalm:

"Super HMB FuryX power":ugly:… I get around 50 and some fps in GTA V with those settings, smooth as silk. :fumeta:

ELP3

Interesting. Very interesting.

I experienced the same thing in my tests.

I didn't comment on anything because I didn't want any controversy, nor was I very sure how the afterburner measured this new graphics card.

But certainly, and being aware of the limitations of this program, and more so with AMD as I mentioned before. Starting from 3600mb consumed, the games or benchmarks, had significant scratches, like being short of Vram. Curiously it was from a similar figure to that of the often mentioned GTX 970.

That's why I commented that their playable experience was not satisfactory.

I don't know exactly where the problem was, but someone hay.Ni knows what the solution is.. I suppose drivers? as always..?

Ciclito

It seems that the new drivers are not the solution… It performs just as poorly.

Ciclito

Since I can't edit it, I'm putting it here that I've posted a video.

wwwendigo

One is bad and the other is good regarding the issue of the pump noise, first the bad one:

Retail Fury X coolers still whine, don't include fix - The Tech Report

Retail AMD Fury X Sound Testing - Pump Whine Investigation | PC Perspective

Why do I put two links for the "bad" part? Because both show different things about the same issue, in Techreport they are interested in this:

AMD received feedback that during open bench testing some cards emit a mild "whining" noise. This is normal for most high speed liquid cooling pumps; Usually the end user cannot hear the noise as the pumps are installed in the chassis, and the radiator fan is louder than the pump. Since the AMD Radeon R9 FuryX radiator fan is near silent, this pump noise is more noticeable.

The issue is limited to a very small batch of initial production samples and we have worked with the manufacturer to improve the acoustic profile of the pump. This problem has been resolved and a fix added to production parts and is not an issue.

AMD reported that the pump noise issue had been fixed and only appeared in a few production units of the first batch, before fixing it. It is important to point this out because the reality was different, in basically all reviews and units sold to users, at least Spanish ones, have found said buzzing, that is, it is not limited only to a few production units, as will be clearer when talking about the "good" part.

PCPer adds tests with 2 units bought from newegg where they have found the same or worse pump buzzing, you can listen to recordings, etc. So Techreport conveniently points out that AMD said something that was not true, and PCPer demonstrates it.

Now the good part:

AMD fixes R9 Fury X Whining Noises

AMD fixes the error by clearly changing the pump issue without small patches, although this movement is so clear and almost 2 weeks after the model's presentation it is a bit contradictory with their words that assured that the problem was fixed in the units that were going to stores.

The bad thing is that you can't distinguish at first sight the good units from the bad ones, since you need to remove the front cover of the card (something that does NOT invalidate the warranty, for those who want to check, know that there is no problem).

The error has been definitely recognized and corrected, although it would have been better to have said clearly that it would be corrected in the following days after the launch, not that it was already "done".

Handrox

jopet.. what a mess is that…

Handrox

Someone who has a FuryX and is good at doing some benchmarks with the GPU Clock at 700MHz or 800MHz would be appreciated ;D

fjavi

@ELP3:

Interesting. Very interesting.

I experienced the same thing in my tests.

I didn't comment on anything because I didn't want any controversy, nor was I very sure how the afterburner measured this new graphics card.

But certainly, and being aware of the limitations of this program, and more so with AMD as I mentioned before. Starting from 3600mb consumed, the games or benchmarks, threw up significant scratches, like being out of Vram. Curiously, it was from a similar figure to that of the often mentioned GTX 970.

That's why I commented that their playable experience was not satisfactory.

I don't know exactly where the problem was, but someone hay.Ni knows what the solution is.. I suppose drivers? as always..?

In that GTA video, you can see that when the frame rate drops, it releases vram, in some cases it goes from 3.6 to 2.6 gb, I don't know if the operating system doesn't treat that memory well but the video shows that it releases vram during the frame.

If it took advantage of the 4GB, something similar would happen, because you can see that it loads and sometimes reaches 3.7 or close to it and that's when it releases vram, but that card is the one that would need 8gb and not the 390x.
Although with one card it won't be able to set those parameters to go to 30 fps or less and you would have to lower some options to go to 60 fps, it would lower consumption, but with two cards or more it will definitely limit that memory.

If they keep releasing games that are so voracious of vram, this card is doomed and I thought that even before they released it, that memory seems very green, they need to release with more density and that it's not too expensive.
Although if it works well in W10, adding memory, the same thing if it serves them, but I don't think everyone will switch quickly to w10.

It could also be a matter of the games releasing the vram, the Titan X seems to be able to locate 8Gb, the 980Ti stores up to 6Gb, and this one releases aa at 3.8gb, the more memory the graphics card has, the more it marks as used, it could be that the games accumulate vram and don't release it until they need it and give the frame.

regards