AMD Radeon R9 Fury X Reviews

[Rys]

@swaaye said:

It was stunning....stunningly loud. But just about every blower is similarly loud if it ramps up. 5800's blower has very limited speed management though. I think it only has 2Dspeed and 3Dspeed. And apparently it actually moves very little air for all that noise. Which leads me to believe it was mostly an attempt to appeal to aural/visual hardcoreness instead of any practicality.

There is a GeForce4 Ti from Abit with a blower branded OTES that is similarly stupid and loud. Possible FlowFX inspiration.

 

[Rys]

@Gandahar said:

Indeed, seems like Ati just need an Nv35 around the corner for this beast.

Polish the diamond.

You can't rely on drivers though, ATi are poor at those, it needs hardware polishing like the nv35 did.

Nvidia pulled themselves up with their bootstraps after the crap nv30, did an ok job with nv35 and better with nv40.

Peaked with the G80. Now peaked again with Maxwell. The G80 peaked on performance, Maxwell peaked on that per watt.

ATi need to do same, or what is left of ATI under AMD banner. The whole company seems on the ropes to be honest.

 

[Rys]

@swaaye said:

ATi need to do same, or what is left of ATI under AMD banner. The whole company seems on the ropes to be honest.

It's hard to say. Clearly there isn't a wealth of resources to go around but I don't think Fiji is all that disappointing. It is a lot faster than Hawaii. Yeah it's not quite as efficient as Maxwell, but for the most part it looks like 28nm is just at the limit. They simply can't make the GPUs bigger.

Maybe the drivers are to blame for some showings, but that's something that is hazy at best...

 

[Rys]

@A1xLLcqAgt0qc2RyMz0y said:

Overclock 3D

The Tech Report

Hardware.fr

Guru 3D

PCPer

Tom's Hardware

TechPowerUp

PCGH

HardOCP

Hardware Canucks

Forbes

HardwareBG

Bit-Tech

Hexus

TweakTown

Legit Reviews

HardwareLUXX

ComputerBase

iXBT

Hardware.info

Hot Hardware

Hardware Heaven

SweClockers

Is AnandTech on AMD's *HIT List?

 

[Rys]

@entity279 said:

Nope. Ryan is sick or so they said

 

[Rys]

@Lightman said:

Is AnandTech on AMD's *HIT List?

I was going to ask similar question.

Where is Fury X review from AT?

Ryan, are you sleeping?

 

[Rys]

@Lightman said:

Nope. Ryan is sick or so they said

I see! Wish him well then! Hope he recovers quickly!

 

[Rys]

@Ryan Smith said:

Ryan, are you sleeping?

A little writing, a lot of sleeping. I'm not dying, but this is wholly unpleasant and it kills my productivity. (Insert conspiracy theory here about AMD getting me sick)

And thank you for the kind words.

 

[Rys]

@Rys said:

Isn't Fury writing more than its theoretical 67 gigapixels per second in this test?

Is there a bug in the test?

It's not purely measuring RBE export bandwidth. It's mostly measuring texture ingress bandwidth. All the test does is sample from a bunch of very large textures into a very large off-screen surface.

 

[Rys]

@Jawed said:

It's not purely measuring RBE export bandwidth. It's mostly measuring texture ingress bandwidth. All the test does is sample from a bunch of very large textures into a very large off-screen surface.

Thanks.

The implication here seems to be: looking for a percentage difference between compressible and incompressible. The actual numbers are nearly meaningless, and so the percentage difference is the number to look at.

 

[Rys]

@Rys said:

In a manner of speaking. The shader consumes all samples and all of the off-screen surface pixels are referenced, so none of the I/O can be thrown away, so in that sense the bandwidth measurement is hopefully pretty correct (for the random texture case) in terms of the reported GB/sec number. How you interpret a higher number for compressible surfaces is another matter.

 

[Rys]

@fellix said:

Fury X possibly reviewed with incorrect drivers

it was confirmed recently by an AMD rep that the 180612a driver is fake and was probably meant to start trouble. this however does not explain why there are multiple 15.15 drivers, june 15 and the 20th (in one of my previous edits). Anything you read below may no longer relavent and new comments should be taken with a grain of salt. If anyone here has a fury could you please bench for us with old and new drivers so we can put this to rest.
"Just putting this out there. Here's a review that seems quite different from the other ones:https://translate.google.com/transl...u=http://www.ixbt.com/video3/fiji-part3.shtml This one actually shows that Fury beats 980 ti in a lot of tests and even titan x in some tests. Of course in lower resolutions it seems that NVidia still wins a lot of times, but it's not as bad as it is with some other reviewers. Now here's a possible reason why. Driver version used in this review is 15.15-180612a-18565BE which the reviewer was sent by AMD on June 18th. The press driver on AMD's FTP server is 15.15-150611a-185358E. I think this is probably the reason of this inconsistency."

edit: you can download the "20th" driver here it should end with june20.exe

edit: guru3d with correct drivers

 

[Rys]

@CarstenS said:

Wow, awesome headline. Should read "Did AMD sent incorrect driver to reviewers?"

FWIW, in the Fury X reviewers guide AMD states:
"Display Driver Version AMD Graphics Driver 15.15 Beta Performance Driver"
We checked consistency with 3DMark FSU, where we got a slightly higher score than in their RG.

 

[Rys]

@entity279 said:

Thouse reviewers, they can't get anything right!

 

[Rys]

@lanek said:

Wow... i see trouble coming

 

[Rys]

@pharma said:

Ouch! Now AMD knows how their customers feel with lack of driver support.

Never heard of ixbt.com anyway. Did AMD state whether this review site was in the officialpoolofFuryXreviewsites? Or is this an isolated external review site that had no access to the press driver (AMD ftp server) and used the latest available to them?

AMDMatt:

In this case the review is suggesting a driver dated from June 12th 2018 and the build request that has a letter instead of a number for its last digit so it’s either a lot of typos or someone being misleading on purpose.

 

[Rys]

@Ryan Smith said:

Aww come on, not here too. Beyond3D is supposed to be better than this... :|

Anyhow, anyone claiming there's a 15.15-180612a driver is either making typos or full of hooey. If you look at AMD's driver strings, they're in YYMMDD format. For example, Catalyst 15.3.1 from March is 14.502.1014-150313a-181517E.

I don't claim to speak for AMD, but there is no other driver besides the original website driver for the Fury/300 series (15.15-150611a-185375E), the press driver (15.15-150611a-185358E), and the current Batman driver (15.15.1004-150619a-185674E).

Also: http://forums.overclockers.co.uk/showthread.php?p=28230087#post28230087

 

[Rys]

@Silent_Buddha said:

Never heard of ixbt.com anyway.

They've been reviewing video cards for quite a number of years now and was one of the major ones mentioned here whenever a new video card was released in the past. I believe they are a Russian site. I remember reading video card reviews there from the R600 and even before.

IIRC they were also one of the first review sites to start recording board power for graphics cards versus total system power.

Regards,
SB

 

[Rys]

@BRiT said:

Yup, ixbt is an old site and has been around for quite some time. I used to refer to them when i was shopping for video cards many years ago because their reviews always seemed to be more than just the suggested video benchmarks in the product reviewer guides distributed by the vendors.

 

[Rys]

@Alexko said:

Aren't they related to Xbitlabs?