AMD Fury X Postmortem: What Went Wrong?

[Silverforce11]

All this back and forth about which card is more powerful yadda yadda when pc games are starting to become nothing but halfbaked ports of console games. Titan x, Fury x and 980ti are pointless cards at the moment and going forward because as 2015 shows games are not truly taking advantage of their power. You may argue that 4k is bringing some of these cards to their knees but whats the point when you need a high resolution monitor to actually distinguish pc quality from consoles? With previous games like crysis 3, metro last light etc. one could actually see the difference, but apart from running at higher resolutions and fps we have nothing. Top cards power are grossly underutilized if something does not change going forward. The PCmasterrace moniker is fading with each crappy games that are ported over.

I agree. That's why I'm not going to waste $ to upgrade this gen because its a total non starter to pay so much for so little. I rather take the family out for an extra few nice dinners.

 

[PieIsAwesome]

All this back and forth about which card is more powerful yadda yadda when pc games are starting to become nothing but halfbaked ports of console games. Titan x, Fury x and 980ti are pointless cards at the moment and going forward because as 2015 shows games are not truly taking advantage of their power. You may argue that 4k is bringing some of these cards to their knees but whats the point when you need a high resolution monitor to actually distinguish pc quality from consoles? With previous games like crysis 3, metro last light etc. one could actually see the difference, but apart from running at higher resolutions and fps we have nothing. Top cards power are grossly underutilized if something does not change going forward. The PCmasterrace moniker is fading with each crappy games that are ported over.

This.

Nvidia/AMD fanboys: "Our cards can play console ports better than your cards can! Look at these benchmarks!"

lol.

 

[AtenRa]

Even Titan X can do 15-20% OC on its crap reference cooler.

Fury X does ~75mhz extra, it's a joke.

It did by raising the voltage automatically.
Seams to me AMD took a step back with Fury-X so OEM Custom designs will be better. Like custom OC GTX980Ti models that are better than Titan-X.

 

[Headfoot]

Doesn't really matter, they messed up. Initial impressions stick. Remember the hot & noisy R290/X?

Now most gamers will think Fury X can't OC at all or its vcore locked permanently.

Yes, apparently 90% of this forum is too daft or willfully blind to understand ASICS OVERCLOCK BETTER WITH MORE VOLTAGE.

Apparently its VERY difficult to understand that it will overclock more when voltage raising utilities come out.

 

[.vodka]

It's even more difficult to grasp the fact that kepler/maxwell cards also up the voltage once you get the clocks higher automatically. They don't do 1200-1400 MHz on "stock voltage" if you get what I mean.

http://www.techpowerup.com/reviews/NVIDIA/GeForce_GTX_980_Ti/35.html

We see 1200MHz on 1.2v here for this particular 980Ti. TPU's info on Fury states 1.22v as stock voltage, and Fiji is doing about 1150MHz on that. I don't see it too far behind.

Once overvolting becomes available we'll see what Fiji is capable of. The dust still has to settle to draw any useful conclusions. Still, I agree about the first impressions. We all know what the first impression of the 290x did to its image. Fury isn't anywhere as bad, though. Needs just a little more polishing that should come with time. Not getting overvolting tools available on launch day, however, after hyping the VRM and seeing the AIO keep the card chilly... come on, you can't possibly miss that opportunity...

 

[railven]

Once overvolting becomes available we'll see what Fiji is capable of. The dust still has to settle to draw any useful conclusions.

What? Come on, seriously? Here is my specific buying check list that I like to use for buying hardware.

It's slower, uses more power, only runs cooler cuz it has a beefier heatsink once that's gone I imagine power consumption to go up some more, it's slower, it has less VRAM for similar price of faster cards...

But we can't make any "useful conclusions" until we get overvolting? I'm pretty sure I'm not the only one who will reach this conclusion:

"it will overvolt, use even more power and STILL be slower than a similar OC'ed/OV'ed 980 Ti"

 

[pj-]

I've seen several people mention the VRM hitting 100C at stock clocks, and asking if that will be a limitation when OCing.

How big of a deal are VRM temps?

 

[.vodka]

Fiji is new silicon. We don't know what it can do with more voltage. It could do much better than Hawaii, for example. It could very well have a similar behavior. We don't know. As it is, FuryX needs a price cut, that's all. Going forward it's gonna get better, as GCN did in the 7970 and 290x.

Consume more power after getting more voltage? Yes, this also happens to every piece of silicon out there. It could be worthwhile. It could be not. Still, it's a product that gets AMD on the 980Ti range of performance. It was needed, they couldn't have possibly gone much longer relying only on Hawaii. It's no bulldozer as it's been called in this thread. Subpar, I agree. A train wreck? No, not at all. A 2900XT vs 8800GTX repeat would've been a train wreck. It isn't this time.


If you're in a hurry there are a lot of juicy >1300MHz 980Tis out there that can get you going.


We can all agree nV has won the DX11 era, on both performance and drivers once Maxwell arrived. There's a new chapter about to be started in a month with W10+DX12 and AMD's hardware seems to be tailored for DX12 from day one, back to the 7970 almost four years ago. Things could get *very* interesting once DX12 games arrive. Meanwhile, Maxwell is still king, especially when overclocked, and Pascal/Arctic Islands are far, far away. In between, anything could happen.

 

[Azix]

The most detailed power consumption analysis I have seen had it using less power than the 980ti (tomshardware). Most don't do that kind of analysis and probably do not have that kind of resolution. The second closest was techpowerup but I am not sure about the resolution or method on that.

4GB is fine for single cards currently. Its crossfire where it might matter more. Doesn't matter what cooler it has, it runs cooler. Nobody would excuse the 290x reference for running hotter just because its cooler wasn't as good.

Price performance meh.

 

[pj-]

The most detailed power consumption analysis I have seen had it using less power than the 980ti (tomshardware). Most don't do that kind of analysis and probably do not have that kind of resolution. The second closest was techpowerup but I am not sure about the resolution or method on that.

4GB is fine for single cards currently. Its crossfire where it might matter more. Doesn't matter what cooler it has, it runs cooler. Nobody would excuse the 290x reference for running hotter just because its cooler wasn't as good.

Price performance meh.

Most sites report total system power consumption. If Fury X has higher DX11 overhead, would that cause more CPU usage and negate the lower draw of the GPU itself?

I don't see how else 10+ sites would be wrong in observing that a system equipped with a Fury X uses more power than an identical system with a 980ti.

 

[Squeetard]

When's the last time the underdog has put it to the big guy? 15 years ago with the Athlon and the ATI 9700. That was it. Now the playing field has tilted to Intel/nVidia and it will never tilt back. They got caught out once and they never will again.

The Fury is a good card. Now that all the launch hype has died down I feel a bit disappointed. They needed a great card. With new tech I expected much better results than this.

I've had my 7970's in xfire for 2 years now. Before that it was a 590 hydrocopper. Both have now soured me for SLI and Xfire. I'm looking for a good single gpu card and now find myself waiting for next year and the node drop. None of the offering make me want to pull out my wallet.

 

[exar333]

You could have asked the same thing when the R290/X launched. They learnt NOTHING at all from the horribly noisy 7970 and especially the 7970Ghz Ed.

They also learnt nothing from mistakes of NV with Fermi. Seriously who thinks 94C on a GPU is a good idea for consumers?

AMD is pure fail when it comes to marketing & management. I was hoping Lisa Su may change course but releasing Fury X with a water cooler with ample thermal room WITHOUT vcore mod tools is idiotic.

@tviceman
A 15% OC is still ~1.2ghz or ~150mhz. Better than the 50-75Mhz that review sites get. It would look less crap at OC for sure. Heck, maybe it can do a 20% OC with vcore for people who don't care about power use! Atm, it's not a possibility because they decided to launch it crippled!

Totally agree. They finally fix the sound/temp issues and then forget to even include the basic tools for overclocking. This is AFTER they touted the amazing OC ability of the chip, at a huge e3 presentation, just a week before.

NV learned a hard lesson with Fermi and Kepler was obviously based from those mistakes.

I wonder if NV releasing the 980Ti forced AMD's hand in releasing this a bit early? There wasn't a hard launch date communicated, so I wonder if this was released a few weeks early to help with the bleeding, so to speak. The re-brands didn't really matter, but AMD needed something on the higher end to compete. This feels very rushed.

Drivers? Sketchy
OC tools? Non-existent
Pro version? Wait a month
Nano? TBD
X2 version? TBD

This was essentially a paper-launch for Fury as a whole, with some Fury X's shipped. Everything else is still on the roadmap and purchasers feel like beta participants.

Who knows, maybe 2-3 months from now this all matures nicely and performance, OC tools and more products really make it look 'good'.

Unfortunately, the impression has already been made (again) that its not great. You can't execute like that as the underdog...

 

[railven]

When's the last time the underdog has put it to the big guy? 15 years ago with the Athlon and the ATI 9700. That was it. Now the playing field has tilted to Intel/nVidia and it will never tilt back. They got caught out once and they never will again.

The Fury is a good card. Now that all the launch hype has died down I feel a bit disappointed. They needed a great card. With new tech I expected much better results than this.

I've had my 7970's in xfire for 2 years now. Before that it was a 590 hydrocopper. Both have now soured me for SLI and Xfire. I'm looking for a good single gpu card and now find myself waiting for next year and the node drop. None of the offering make me want to pull out my wallet.

As a former Xfire 7970 user, if I was still rocking that hardware I'd drop it in a heart beat for either a fury X or a 980 Ti.

The power saving alone would (in my neck of the woods) be a god send. The noise/heat from my old 7970's (with custom coolers too) was unbearable in the summer (use more electricity having to turn on the A/C so I can game comfortable with my rig).

Then the lack of strong Xfire profiles for games I play. I was playing with 1 card as often as I was with 2.

 

[exar333]

When's the last time the underdog has put it to the big guy? 15 years ago with the Athlon and the ATI 9700. That was it. Now the playing field has tilted to Intel/nVidia and it will never tilt back. They got caught out once and they never will again.

The Fury is a good card. Now that all the launch hype has died down I feel a bit disappointed. They needed a great card. With new tech I expected much better results than this.

I've had my 7970's in xfire for 2 years now. Before that it was a 590 hydrocopper. Both have now soured me for SLI and Xfire. I'm looking for a good single gpu card and now find myself waiting for next year and the node drop. None of the offering make me want to pull out my wallet.

The whole 48xx/58xx/6970/7970 were all fantastic products. I didn't see AMD behind much at all during those years, and in many ways ahead. Its been since then that they have slipped.

Agree on the CPU front....the X6 was pretty solid though, until that got replaced by the worse BD arch.

 

[Azix]

Most sites report total system power consumption. If Fury X has higher DX11 overhead, would that cause more CPU usage and negate the lower draw of the GPU itself?

I don't see how else 10+ sites would be wrong in observing that a system equipped with a Fury X uses more power than an identical system with a 980ti.

System power consumption can introduce that kind of thing yeah. Though I guess it matters

I think this is the best analysis so far of the power consumption

http://www.tomshardware.com/reviews/amd-radeon-r9-fury-x,4196-7.html

 

[Squeetard]

As a former Xfire 7970 user, if I was still rocking that hardware I'd drop it in a heart beat for either a fury X or a 980 Ti.

The power saving alone would (in my neck of the woods) be a god send. The noise/heat from my old 7970's (with custom coolers too) was unbearable in the summer (use more electricity having to turn on the A/C so I can game comfortable with my rig).

Then the lack of strong Xfire profiles for games I play. I was playing with 1 card as often as I was with 2.

You make a good point but my whole rig is watercooled. And the tower is in the closet of the spare room next to my office. Zero heat/noise in my office. So I'm waiting for the node drop to really make a difference in power.
I had just as many issues running SLI as I do xfire.

 

[Headfoot]

It's even more difficult to grasp the fact that kepler/maxwell cards also up the voltage once you get the clocks higher automatically. They don't do 1200-1400 MHz on "stock voltage" if you get what I mean.

http://www.techpowerup.com/reviews/NVIDIA/GeForce_GTX_980_Ti/35.html

We see 1200MHz on 1.2v here for this particular 980Ti. TPU's info on Fury states 1.22v as stock voltage, and Fiji is doing about 1150MHz on that. I don't see it too far behind.

Once overvolting becomes available we'll see what Fiji is capable of. The dust still has to settle to draw any useful conclusions. Still, I agree about the first impressions. We all know what the first impression of the 290x did to its image. Fury isn't anywhere as bad, though. Needs just a little more polishing that should come with time. Not getting overvolting tools available on launch day, however, after hyping the VRM and seeing the AIO keep the card chilly... come on, you can't possibly miss that opportunity...

Well said, I agree with you 100%. Good post.

 

[Headfoot]

System power consumption can introduce that kind of thing yeah. Though I guess it matters

I think this is the best analysis so far of the power consumption

http://www.tomshardware.com/reviews/amd-radeon-r9-fury-x,4196-7.html

Tom's had the best Fury X review IMO. Frametime analysis, very in depth power analysis, good selection of games (there could be more, but you can't get everything you want). Really well done

 

[IEC]

System power consumption can introduce that kind of thing yeah. Though I guess it matters

I think this is the best analysis so far of the power consumption

http://www.tomshardware.com/reviews/amd-radeon-r9-fury-x,4196-7.html

Tom's actually has a proper testing setup for power consumption. I would trust their numbers to be more accurate than the competition.

 

[Jaydip]

Tom's review could have been considered decent but no, just look at this
http://www.tomshardware.com/reviews/amd-radeon-r9-fury-x,4196-3.html

 

[Enigmoid]

Tom's review could have considered decent but no, just look at this http://www.tomshardware.com/reviews/amd-radeon-r9-fury-x,4196-3.html

Tom's is using all different drivers. Its a mess.

Their power numbers are questionable too. Very different from other sites. I remember they also messed up the 980 power numbers and ended up pulling them (simulated 970/980 clocks and power usage with an aftermarket 970/980 with a much much higher power limit, naturally results were incorrect).

The idle power is unrealistically low and uncorroborated by any other site.

 

[sm625]

The best way to measure power consumption is at the wall with one GPU, and then at the wall with the other GPU. Tom's is playing too many games in their methods.

 

[RussianSensation]

Yes, apparently 90% of this forum is too daft or willfully blind to understand ASICS OVERCLOCK BETTER WITH MORE VOLTAGE.

Apparently its VERY difficult to understand that it will overclock more when voltage raising utilities come out.

This forum is 100% full of double standards, but specifically when it comes to favouring NV products. I was about to start a new thread but didn't even bother because I'd just get called out for "Damage control" or some other non-sense.

Here are the facts:

HD7970Ghz = $499 with a 3-game bundle, 50% more VRAM, huge overclocking headroom/scaling, 9% faster at TPU at 1440P on launch date than $499 GTX680 2GB. I recommended HD7970Ghz back then and today I am recommending the 980Ti. Coincidentally that is definitely not the case for the 980Ti fanclub which back then downplayed all key advantages of HD7970Ghz from launch.

#1. The amount of bashing GTX680 2GB received for having 50% less VRAM from the same people talking about 4GB vs. 6GB today on the NV side - practically nil.

#2. The amount of bashing GTX680 2GB received for being noncompetitive by 9% at the same price resulting in worse price/performance from the same people today bashing Fury X for not being fast enough at 1440p - practically nil.

#3. How many people during HD7970Ghz vs. 680 era discussed how one can buy after-market HD7970Ghz cards? Almost none of the people today who only want to compare factory pre-overclocked 980Ti cards vs. a reference Fury X. That's right cuz they don't want to discuss how a reference 980Ti is a failure in terms of noise levels and temperatures but back then the theme was to ignore after-market HD7970Ghz cards' noise levels, temperatures and overclocking at all costs and continue with the mantra that all HD7970Ghz cards run hot and loud despite not 1 reference HD7970Ghz ever sold in retail channels like Newegg or Amazon, etc. It's sad to see these insane double standards being applied today but 680 was somehow excused from this? Right....

While objective gamers have every single right to criticize Fury X for not delivering enough (until we see its full overclocking potential with voltage control), nearly every single person who recommended a 680 over an HD7970Ghz back then is a true hypocrite. I'll never forgot how certain posters on here would constantly, and I mean constantly mislead with reference HD7970Ghz noise and temperature levels but now they are eerily quiet about how the reference 980Ti's cooler is a giant failure, especially when it comes to overclocking. :sneaky:

Did Fury X live up to the hype? No, for now it didn't but the amount of bashing happening and how it's a total failure is ludicrous considering HD7970Ghz had 50% more VRAM, cost the same and outperformed 680 by the same 9%. Yet, how many of the same people called 680 a failure and recommended 7970Ghz over it? Practically none of the NV owners/Fury X bashers. It's pretty easy to spot blind fans on this forum to be honest.....

Fury X 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, we still got a card that at stock is nearly as fast as a reference GTX980Ti, but runs way cooler and quieter. Maybe with full voltage control, maybe it can overclock to 1250mhz? We can't say yet, but either way there seem to be driver issues because the scaling from R9 390X is simply too low at the moment to make sense unless the card is almost 100% ROP bottlenecked.

At 4K it basically trades blows with GTX980Ti/Titan X which means Fury X CF should be fairly competitive against 980Ti SLI.

Still, it's hilarious that the clueless/biased review sites are trying to say that 980 is a true competitor to Fury X in terms of gaming feel when objective sites with more than 5 games (or which have a spread of games and not just mostly GW titles) are finding Fury X destroying a 980 by 25-30% at high rez gaming and more or less trading blows with the reference 980Ti. Where Fury X starts to suffer is against after-market 980Ti cards but those are factory pre-overclocked 15-18% which makes sense that they would win since AMD's AIBs have so far didn't release any factory pre-overclocked Fury cards.

Right now, the logical choice is an after-market 980Ti but it remains to be seen if overclocking will improve with voltage control and if AMD can get another 5-10% performance increase with better drivers over time; and if AMD will allow AIBs to release faster clocked versions of the Fury X.

Generally speaking the biggest problem with this card right now is the price. I am not sure why there is so much bashing for 1080P performance since we know AMD has a DX11 driver overhead. Also, I would personally never buy a $650 GPU late June 2015 to play at 1080P. I think the key resolution for comparison is 1440P (or Ultra wide 3440x1440) for single chip and 1440P/1600P/3440x1440/4K for CF/SLI.

I agree. That's why I'm not going to waste $ to upgrade this gen because its a total non starter to pay so much for so little. I rather take the family out for an extra few nice dinners.

I am sticking to my point that this is a stop-gap generation. None of the cards out are fast enough for 4K on their own and with next gen we should get close to 980Ti SLI in a single chip. Also, most likely NV will drop SLI bridges and we will get at least 8GB HBM2 on flagship cards. This is a good time for HD7950/HD7970/680/770 owners to upgrade but for 970/290/290X owners in CF, the only way to truly 'upgrade' is to buy 2x 980Ti or 2x Fury Xs. That's a LOT of money for what smells like a stop-gap generation. I am not going to be surprised if 970's successor comes very close to the 980Ti in performance in 2H of 2016. For anyone who isn't in an urgent need to upgrade their GPUs, I would say there is no rush right now. On the NV side, their GM200 lacks 4K hardware decode/encoding which means it'll be outdated for 4K anyway when next gen GPUs launch.

This is a double-whammy: Fury X is slower than after-market 980Ti but once Pascal comes out NV is likely to focus all of is driver efforts on that architecture and performance for Maxwell cards may suffer like Kepler and Fermi cards did. That means spending $1300+ on 2 of these cards is going to look silly come 14nm/16 HBM2 gen.

 

[railven]

Tom's is using all different drivers. Its a mess.

Their power numbers are questionable too. Very different from other sites. I remember they also messed up the 980 power numbers and ended up pulling them (simulated 970/980 clocks and power usage with an aftermarket 970/980 with a much much higher power limit, naturally results were incorrect).

The idle power is unrealistically low and uncorroborated by any other site.

You must be confused. If Tom reports negative power consumption, it's a fact. The other 10+ review sites - all wrong.

Tom says Fury X uses -23 W, you better call your electric company because they owe you a check.

Sarcasm aside, seriously if one out of a handful of site is reporting something completely different - you can't err saying that one site is right. Clearly something is off and I'd hope the other sites (or Toms) are taking a double look.

 

[Jaydip]

Tom's is using all different drivers. Its a mess.

Their power numbers are questionable too. Very different from other sites. I remember they also messed up the 980 power numbers and ended up pulling them (simulated 970/980 clocks and power usage with an aftermarket 970/980 with a much much higher power limit, naturally results were incorrect).

The idle power is unrealistically low and uncorroborated by any other site.

The most important thing is they are using built in benches , do we play benches or do we play the game? incompetence at it's best.

The funny thing is just 2-3 days ago these sites were listed as NV shill sites, THG,HWC,PC-PER, now they are the best reviewers on planet :biggrin: