AMD Radeon R9 Fury reviews

But let me reiterate my main point: it's stunning that a 400mm2 chip with half the BW manages to perform close to a 600mm2.

Except that no matter how much green fairy dust you try to pour on it, your fully enabled GM204 at 400mm^2 does not perform close to a fully enabled 600m^2 Fiji XT.
Gameworks titles aside (where it might get within some 10% on the most extreme cases), the 980 does not perform even remotely close to a Fury X.

That comparison is stupid.

Perhaps you'll try to argue that the GTX 970 (a lower binned and cut-down GM204) comes close to a Fury (lower binned and cut-down Fiji)?
Then there you go:

8DrOLtx.gif
DRY1o1R.gif
QSgVkmq.gif
JCzskmY.gif
LbJskjE.gif



What's that? A bunch of benchmarks showing the lower-binned variant of the 50% bigger chip has 50% more performance than the lower-binned variant of the smaller chip?
And the same thing is happening between the fully-enabled versions of the same chips?

The shock!
 
What's that? A bunch of benchmarks showing the lower-binned variant of the 50% bigger chip has 50% more performance than the lower-binned variant of the smaller chip?
And the same thing is happening between the fully-enabled versions of the same chips?

The shock!
Not sure what graphs you're looking at, but those you posted show a difference of 22.3-43.8%.
 
Except that no matter how much green fairy dust you try to pour on it, your fully enabled GM204 at 400mm^2 does not perform close to a fully enabled 600m^2 Fiji XT.
Am I supposed to be impressed that you've just discovered this?

That comparison is stupid.
That's opinion, and you're entitled to it.

What's not opinion is that AMD has positioned a 600mm2 die with very expensive memory technology against a 400mm2 with dirt cheap memory technology. It matters very little that one is a recovery and the other a perfect: 28nm is about as mature as process can be. 550mm2 perfect dies in the form of GTX 780 Ti have been on the market in high volume over almost 2 years now. So have 420mm2 prefects in the form of Hawaii. Or 360mm2 perfects for 3.5 years in the form of Tahiti. The yields on a perfect 400mm2 gm204 may well be higher than the yield of a recovered Fiji.

Perhaps you'll try to argue that the GTX 970 (a lower binned and cut-down GM204) comes close to a Fury (lower binned and cut-down Fiji)?
Then there you go:
Congratulations! You came up with a trivial premise that nobody would contest in the first place and then proved said premise to be true. Everybody needs to learn to walk before they can run, so a "good boy!" is in order?
 
Everyone reads the test setup section to find out what cards and drivers are used.

No they don't.

Fury is barely beating GTX 980 in 1440p and it have huge issues with incosistencies with frames where GTX 980 is much more fluid.
Get a OCed Gigabyte or MSI and you will beat a Fury by a good margin and enjoy consistent and fluent gaming.

:LOL:

http://www.overclock.net/t/1564303/various-amd-r9-fury-reviews/340#post_24155695

Or maybe they do, which would make it even worse. Nevermind that TechReport should label it as such in their graphs.

Next time you see a 980 performing on par or better than the Fury just realize the review site's GPU comparison is based on price. Fury's price is just too high for what you get. :D

It isn't if 980 is at $500. Nevermind that even OCed cards will find it hard to get a 20% or higher boost.

Really! I would never have imagined from the sound of your earlier posts praising how well they performed in some reviews;-). But lets hope you are right ....

Where? The first thing I came across while reading about Fury reviews was how the new drivers didn't do anything to improve performance which was being celebrated by nvidia fans, not realizing that it does the opposite for their case. :LOL:

I did comment on the first leaker of the Fury X getting impressive results in Far Cry 4, but I expressed my skepticism with them as well. Though not on this forum.

https://forum.beyond3d.com/threads/catalyst-15-7-released-win10.57127/page-2#post-1860147

http://www.overclock.net/t/1561860/various-amd-radeon-r9-fury-x-reviews/2930#post_24148118
 
Am I supposed to be impressed that you've just discovered this?

No, you're supposed to understand that no one is going to buy this "Fiji is so very bad" crap because you decided to compare performance/area between a chip that is fully enabled and another that salvaged by disabling execution units, and are somehow trying to pass this as a legitimate comparison.

I myself have presented my fair share of criticism on Fiji, but this is completely bonkers.
How come the 390X gets about 29% more performance than a GTX 970 in Shadows of Mordor 4K with only a 10% larger die? Wow, GM204 is such a bad chip!!111one



What's not opinion is that AMD has positioned a 600mm2 die with very expensive memory technology against a 400mm2 with dirt cheap memory technology.
You mean those huge memory price differences that you so far have completely made up in your head and insist bringing up in every. single. post. about Fiji?
Yes, I'll totally keep in mind the silent_guy's chart of made-up things the next time I'm comparing GPUs and graphics cards, and I'm pretty sure everyone else here will too.





Yeah, gotta love them 1080p results from a 2 year-old game for >$500 cards.
 
Last edited by a moderator:
Fluctuations in prices do happen and cards are overpriced at release with less of stock.

Still waiting on that ixbt review, but that GTA V bench looks quite cpu bottlenecked, with 980Ti less than 10% faster.
 
Fluctuations in prices do happen and cards are overpriced at release with less of stock.

10 July 2015
http://hexus.net/tech/features/systems/84593-qotw-which-enthusiast-gpu-choose/
The launch of AMD's Fury products - in particular the R9 Fury at around £450 - appears to have had an immediate impact on GeForce pricing. Indeed, a quick scout around UK retailers reveals that GTX 980 Ti has dipped below the £500 mark, while GTX 980 has conveniently tumbled to £360.
 
Last edited:
That's good to know, but haven't seen anything official. And 980Ti dipping in price while the perception is that it's a better card than Fury X sounds strange.
 
Yep, I also love the higher resolutions with newer games ... but I guess you've already seen this.
gta5_2560_2_oc.png


Hard to miss, since you posted that very same graph not one, not two but three times in this thread, with the last two being within the span of ~30min.
Perhaps if you post it another 3 times we'll start to collectively believe that GTA V at 1440p is the only game we'll ever play, and we'll only play it in Paleto Boulevard.
(I'm kidding, please don't do that..)


Still waiting on that ixbt review, but that GTA V bench looks quite cpu bottlenecked, with 980Ti less than 10% faster.

ixbt already have GTA V with the Fury X in their latest 3D Digest.
It seems the game scales like ass when Antialiasing is turned on, but the multi-GPU solutions show that it's not CPU-limited.
Without that, the game scales quite well:

1Dsp3Pd.jpg
Mvb0Bbh.jpg
5pSQKy8.jpg
89vIvto.jpg
 
It's a game of "chicken" ... all about pricing strategy and who has the lower profit point.

Well, the strange part is nvidia doing it for a card that is supposedly better. Very unlike them hence my wait to see if it's really official. Maybe they got traumatised by the ixbt review. :p
 
Except that no matter how much green fairy dust you try to pour on it, your fully enabled GM204 at 400mm^2 does not perform close to a fully enabled 600m^2 Fiji XT.

It does perform quite close if you don't cherry pick only 4k results and actually fully enable them as in overclock them. No need to even raise the voltages .
 
No, you're supposed to understand that no one is going to buy this "Fiji is so very bad" crap because you decided to compare performance/area between a chip that is fully enabled and another that salvaged by disabling execution units, and are somehow trying to pass this as a legitimate comparison.
Even if they'd disable 100% of the functional units and achieved 100% yield, the die is still 50% larger. (And requires an interposer. And requires HBM. And requires a beefier cooler. And requires a beefier power circuit.) That is what makes it such a shitty proposition for AMD.

How come the 390X gets about 29% more performance than a GTX 970 in Shadows of Mordor 4K with only a 10% larger die? Wow, GM204 is such a bad chip!!111one
When all else fails, let's resort to cherry picking individual results. It's a sign of desperation. Just go to the TPU average results and you'll see that even at 4k, the difference between a 970 and a 390X (with 8GB!) is only 12%, and only 8% for 2560x1440.

You mean those huge memory price differences that you so far have completely made up in your head and insist bringing up in every. single. post. about Fiji?
Without HBM, and with 100% yields on either side, and at $6000 per water, the difference between Fiji and gm204 is $26 right there. I think that's a very nice starting point. Why don't you, for a change, come up with correction factors and make some quantitate estimates about how this becomes a good deal for AMD.
 
Back
Top