AMD Fury X Postmortem: What Went Wrong?

[Hitman928]

I saw that, but it seems the drivers didn't make much difference according to other discussions.

Reddit discussions? I did look at the review in question and it paints the Fury in a much better light performance wise. I don't know the review site though and I'm very skeptical as to the purported cause being the drivers, just curious if someone actually looked into it.

 

[Fallen Kell]

Sure reviewers can test any game, but if they justify it on "popular games" then for sure Witcher 3 is fair (but turning on NV features like HairWorks for AMD isn't). Dying Light & Project Cars are not popular at all.

Look for yourself:
http://steamcharts.com/top

But the point is in reviews that test a lot of games (including GW titles), Fury X matches 980Ti at 1440p and 4K.

In regards to tossing out games:
Tech Report for example, they tossed out Dirt Showdown because one feature: Global Illumination (which they could disable) ran badly on Kepler. They said its a clearly AMD favored game so they can't include it in their tests. Why now do they include Project Cars? There's no features on/off that help AMD, the game just runs crap and even more biased towards NV.

I don't know what reviews you are looking at that shows Fury X matching the 980Ti at 1440p. It is almost universally crushed by 10-20% at that resolution in almost every game (Battlefield 4 20% slower, Crysis 3 14% slower, Shadow of Mordor effectively tied (pretty much one of the few games, but min frame rates are about 10% lower on Fury X), Civ: Beyond Earth 15% slower, Dragon Age: Inquisition 22% slower, Grid Autosport 73% slower!!!, GTA:V 17% slower....

I could go on and on. It is even worse at 1080p resolution as well. The fact of the matter is Fury X really only competes at 4k resolution, yet the vast majority of the gamers out there are not and will not be using that as their resolution, and instead are gaming at 1440p and 1080p, where Fury X gets spanked by the 980 Ti in most games, and is tied in the others, with very, very few wins (I think the only win might be Far Cry 4).

 

[ocre]

It is depending on the review you check. If some reviews use more GPU demanding games and more MSAA at 1440P, 980 gets destroyed.

23% faster at Sweclockers

19% at normal quality and 20% at high quality at Computerbase
http://www.computerbase.de/2015-06/amd-radeon-r9-fury-x-test/6/

While I really enjoy TPU's reviews, you gotta account for majorly CPU limited benches they have like WoW or gaming benchmarks that make 0 sense against all other sites like their Wolfenstein benchmark which is flat out wrong it seems.



.

What? Your basically saying that it is 20% faster than a 980.

How is that your response?

This card should have been priced at $549 but considering we heard the most insane non-sense like Fury X = R9 290X with HBM or Fury X = Dual Tonga XTs or Fury X is only 20% faster than a 980 at high resolutions

The chart in your very own post shows that fury x is only 16% ahead of 980 at 4k.

BTW Nice edit on your post. When I first clicked reply you had said

He was being sarcastic. I mean, 16% you posted, the 19%, or the 20%.....you were back peddling.

Anyway....
we all got screwed this round. Being stuck at 28nm, i guess we all dreamed too big

 

[Silverforce11]

I don't know what reviews you are looking at that shows Fury X matching the 980Ti at 1440p. It is almost universally crushed by 10-20% at that resolution in almost every game

I don't cherry pick a few games here or there that skews the result, I look at sites that test a lot of games and offer a summary chart:

TPU:

Computerbase:

HWC:

Sweclockers:

3-10% slower at 1440p. Tie at 4K.

If you want to cherry pick, go to [H] and Fury X is 25-50% slower.

 

[Subyman]

Reddit discussions? I did look at the review in question and it paints the Fury in a much better light performance wise. I don't know the review site though and I'm very skeptical as to the purported cause being the drivers, just curious if someone actually looked into it.

I believe it was on OCN. IIRC, there were some reviews that used the correct drivers and they were compared to the other reviews. There wasn't much of a difference. I'll try to dig it up.

 

[Subyman]

I'd be curious to see if a new driver is released that bolsters performance. Wouldn't AMD, if this is true, be asking reviewers to redo tests?

I think there's likely quite a bit of performance to be gained through better drivers, especially on Gameworks titles, but until I see that become a reality I really don't have a great incentive to upgrade over my current setup. Barely better than a 390X in some games isn't exactly making me reach for my wallet.

That could of course change with the release of aftermarket Fury cards (possibly better value) and better drivers (happened with 7950). I'm content to wait and see before I upgrade in the fall. If Fury X reaches parity with 980 Ti it would be a better deal. If not, I go back to nVidia for a generation

I'd like to clarify that I meant the review drivers, not future releases.

 

[Silverforce11]

The review in question, that used a more recent driver dated 18th June (others used the older driver for the 300 series launch):

Edit:
https://translate.google.com/transl...u=http://www.ixbt.com/video3/fiji-part3.shtml

 

[.vodka]

Hm. Interesting. Those numbers are more appropriate for the specs on paper. They make MUCH more sense. It shouldn't lose as bad as it does now on lower resolutions with these drivers, too. The 300 series driver is for Hawaii anyway. This could explain a lot of weird things in all the other reviews using the other driver.


If it turns out they screwed up on what driver they sent out for reviews... well, I don't have words for that :\ I don't recall this happening on other launches before...

 

[iiiankiii]

Wow. The alleged new driver looks to be much more competitive. If AMD sent out the wrong drivers, honestly, they deserve to go bankrupt. That's just piss poor management. Fury X is probably the most important launch in the history of AMD. They can't screw up this bad, can they?

Those numbers make much more sense. It would actually justify the pricing. All of a sudden, the Fury X doesn't look 'bad' at all.

 

[geoxile]

Silver, could you please link the review.

 

[Zstream]

Hm. Interesting. Those numbers are more appropriate for the specs on paper. They make MUCH more sense. It shouldn't lose as bad as it does now on lower resolutions with these drivers, too. The 300 series driver is for Hawaii anyway. This could explain a lot of weird things in all the other reviews using the other driver.


If it turns out they screwed up on what driver they sent out for reviews... well, I don't have words for that :\ I don't recall this happening on other launches before...

It's a bad product manager, if that's truly the case. Well, if they're using an agile methodology this is a bad situation and shows that there are new people working for AMD.



I've been a product owner for the past five years and that's inexcusable. Just my 2 cents.

 

[Subyman]

Sorry guys:

For reference on what driver strings mean by the way:
15.15-150611a-185358E (This is the real driver we provided)
15.15 - is the branch
150611 - is the date of the build, YY/MM/DD
185358 - is the build request from our system to create this driver based off the information above

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.

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

 

[Silverforce11]

It would still look crap at 1080p where the majority of the market is at. It would still lose to OC Custom 980Ti, which is the problem. Not reference 980Ti.

I don't think they can fix driver overhead issues until Windows 10 become popular.

ps. AMDMatt made a fool of himself re the HDMI 2 issue. I don't get why they have such incompetent people in marketing.

 

[YBS1]

You would think if this was the case though AMD's PR department would have instantly went into hyperactive mode, not to mention making sure this never happened to begin with. It would be fantastic if it turned out Fury really did have a lot more on tap and was crippled inadvertently. Still, at this point it would almost be par for the course with AMD.

 

[gamervivek]

ixbt have been around for quite some time, and they're related to xbitlabs iirc.

 

[Hitman928]

Sorry guys:



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

I saw that, but then Brent reported this for his driver version (includes an E at the end which apparently isn't suppose to be there according to AMDMatt). The site did say they got a later driver version, maybe just a typo in the review? I don't know, like I said, I'm much more likely to think the site just tested in a way that was favorable to AMD, but I was just curious if anyone actually tested a difference. The testing I saw elsewhere wasn't making a comparison to the driver used by the new review site.

 

[Hitman928]

You would think if this was the case though AMD's PR department would have instantly went into hyperactive mode, not to mention making sure this never happened to begin with. It would be fantastic if it turned out Fury really did have a lot more on tap and was crippled inadvertently. Still, at this point it would almost be par for the course with AMD.

I agree with this, lol.

 

[Zstream]

It would still look crap at 1080p where the majority of the market is at. It would still lose to OC Custom 980Ti, which is the problem. Not reference 980Ti.

I don't think they can fix driver overhead issues until Windows 10 become popular.

ps. AMDMatt made a fool of himself re the HDMI 2 issue. I don't get why they have such incompetent people in marketing.

You'd be surprised how some people are terrible at marketing and yet still fill the position.

 

[Azix]

Man if that is true it would be really funny and sad at the same time. I was wondering if AMD really messed up and forgot something but thought that was unlikely.

Imagine how silly this would be if true. Launch a new product and reviewers using the wrong drivers...

 

[DDH]

https://translate.google.com/translate?hl=en&sl=ru&tl=en&u=http%3A%2F%2Fwww.ixbt.com%2Fvideo3%2Ffiji-part3.shtml

The website that those benches came from


At the end they claim to have overclocked the core to 1120 and memory to 600.

Anyway, Id love a fury X to run my own tests. But I expect that the fury will be sold out here in Aus for a while yet

 

[boozzer]

It would still look crap at 1080p where the majority of the market is at. It would still lose to OC Custom 980Ti, which is the problem. Not reference 980Ti.

I don't think they can fix driver overhead issues until Windows 10 become popular.

ps. AMDMatt made a fool of himself re the HDMI 2 issue. I don't get why they have such incompetent people in marketing.

losing to oc 980 ti is fine, as long as fury x can also oc. I want to see some oc numbers.

edit: this is my reasoning. for the same price, if oc vs oc, fury x is within 5% of 980 ti, that is 100% A ok in my book. because I get a nice and quiet clc cooler out of it. that is a 100$ value right off the bat. the best part is the noise reduction. that alone could seal the deal for me. I am not jumping the gun, still waiting on more info. this is one hell of a rocky launch.

we all need oc numbers! performance has to be within 5% of the best factory oc 980 ti.

 

[x3sphere]

losing to oc 980 ti is fine, as long as fury x can also oc. I want to see some oc numbers.

I've only seen a 10% overclock at best on the Fury.

Basically all reviews show Maxwell hitting around 25-30%+ on stock volts. I've seen higher too but I'll be conservative with that since I doubt all GPUs are getting over 1400 MHz without a voltage bump.

The overclock headroom is the only real disappointment for me with Fury. Supposedly the voltage will be unlocked later, but it's baffling why this wasn't ready for launch reviews considering AMD made a big deal about it at the E3 event.

 

[ocre]

The review in question, that used a more recent driver dated 18th June (others used the older driver for the 300 series launch):

So you couldnt post a link to those claimed magic drivers?

https://translate.google.com/translate?hl=en&sl=ru&tl=en&u=http%3A%2F%2Fwww.ixbt.com%2Fvideo3%2Ffiji-part3.shtml

The website that those benches came from


At the end they claim to have overclocked the core to 1120 and memory to 600.

Anyway, Id love a fury X to run my own tests. But I expect that the fury will be sold out here in Aus for a while yet

Oh. now we get them

I thought the memory was locked and couldnt be overclocked? Is this even real?
Why were the charts being posted without a source?

EDIT--->
guess that is that

Sorry guys:



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

but i bet this wont be the last time i see these faked charts

 

[Silverforce11]

I've only seen a 10% overclock at best on the Fury.

Basically all reviews show Maxwell hitting around 25-30%+ on stock volts. I've seen higher too but I'll be conservative with that since I doubt all GPUs are getting over 1400 MHz without a voltage bump.

It's a myth, Maxwell like Kepler, ups the vcore dynamically in response to the target clock speed & power limit.

GCN needs manual vcore control. It's useful in that you can also undervolt it and save a lot of power.

AMDMatt also said vcore modding is coming soon on that same thread. Hope he knows what he is talking about because if its vcore LOCKed, would be major fail. lol

From that Russian review:
https://translate.google.com/transl...u=http://www.ixbt.com/video3/fiji-part3.shtml
"drivers AMD version 15.15-180612a-18565BE (obtained from AMD 18 June)"

 

[geoxile]

It's a myth, Maxwell like Kepler, ups the vcore dynamically in response to the target clock speed & power limit.

GCN needs manual vcore control. It's useful in that you can also undervolt it and save a lot of power.

AMDMatt also said vcore modding is coming soon on that same thread. Hope he knows what he is talking about because if its vcore LOCKed, would be major fail. lol

From that Russian wing of Xbitlabs:
https://translate.google.com/translate?hl=en&sl=ru&tl=en&u=http%3A%2F%2Fwww.ixbt.com%2Fvideo3%2Ffiji-part3.shtml
"drivers AMD version 15.15-180612a-18565BE (obtained from AMD 18 June)"

Only a small sample of games and Fury generally seems to perform better with AA turned up. I somehow doubt drivers have much to do with it so much as different settings. AF is ezpz for modern GPU so it shouldn't even be much of a factor.