AMD Fury X Reviews

[Jaydip]

IIRC, "higher performance per watt" was referring to AMD cards, not the 980ti?

I saw the slides now, yup both you and the cap are correct. I was wrong in assuming that it was referring to 980 Ti. Still HBM should have bought down the power usage correct?

 

[Jaydip]

Actually it's pretty common to compare new products vs your old products to show progress made. NV does it all the time comparing Tesla vs Fermi vs Kepler etc.

Yup saw the slides now.

 

[Azix]

http://tpucdn.com/reviews/AMD/R9_Fury_X/images/perfrel_2560.gif[/IMG

vs.

[IMG]http://tpucdn.com/reviews/Gigabyte/GTX_980_Ti_G1_Gaming/images/perfrel_2560.gif[/IMG

980Ti = 87%
Gigabyte G1 980Ti = 100% or 15% faster (1.15x)

If Fury X = 100%
GTX980Ti = 109%
Gigabyte G1 980Ti = 109% x (100%/87%) = [B]125%[/B]

$40 more for 25% higher factory-warrantied performance at 1440P, 50% more VRAM as a bonus. Even if Fury X was $549, I don't think I would buy it over the after-market 980Ti.[/QUOTE]

Thats just an overclocked 980ti. We have yet to see a properly overclocked fury X to be making those judgements.

one scenario could be the partners have to tune overclocking through the bios so we see most of the OC from factory overclocks with still limited control for DIY overclocking. I'd assume AMD sent out the base card setup.

Did any reviews show the voltage the GPU currently uses?

 

[Jaydip]

their comment was that Fury X would be up to 1.5X performance / watt over 290x & Nano would be up to 2.0X performance / watt over 290x

using TechPowerUp's charts

http://www.techpowerup.com/reviews/AMD/R9_Fury_X/32.html

@ 1080, it's 1.33
@ 1440, it's 1.39
@ 4K, it's 1.43

That is very accurate then, kudos to AMD. I was still hoping HBM would tilt this battle in their favor however.

 

[Stormflux]

Certainly a shaky launch. Good but not great, which AMD had to be. The month old 980TI was really the curveball for everyone, especially the Aftermarket OC variants. Still too rich for me and the Canadian dollar.

I agree with others, with these reviews the FuryX would have been a solid recommendation at $599. As I'm questioning the validity of the $549 Fury. Is it a FijiXT but just lacking the AIO cooler? Or is it a FijiPro? If it is a FijiPro, than extropalating performance from these reviews would put it slightly ahead of 980 performance.

 

[RussianSensation]

Thats just an overclocked 980ti. We have yet to see a properly overclocked fury X to be making those judgements.

Ok but when I go out to buy a new card, I see a $650 Fury X and a $690 Gigabyte G1 that's 25% faster out of the box with 0 overclocking required. What am I going to buy/recommend? Overclocking is not guaranteed but that 25% faster performance is warrantied by Gigabyte. That's not even the fastest 980Ti as Zotac AMP! 980Ti should have 1355mhz Boost out of the box.

AMD pulled a 6970 vs. GTX580. Performance is very good at high rez stock vs. stock but once the latter is overclocked, it's all over. That generation though AMD priced HD6950 at $299 that unlocked and HD6970 at $379 I believe. Today they are trying to pull off a similar thing but price Fury X at $650. Getting crushed by an after-market 980Ti. Fury X needed to be 5-7% faster than a TX to survive the onslaught of after-market + overclocked 980Ti. Since it can't even beat a reference 980Ti at 1440P, it's hopeless now without a dramatic 10-15% driver improvement and some incredible voltage overclocking. But the latter could result in Fury X using 150W+ more power than an after-market 980TI OC.

 

[RaulF]

Thats just an overclocked 980ti. We have yet to see a properly overclocked fury X to be making those judgements.

one scenario could be the partners have to tune overclocking through the bios so we see most of the OC from factory overclocks with still limited control for DIY overclocking. I'd assume AMD sent out the base card setup.

Did any reviews show the voltage the GPU currently uses?

Factory overclock card. All you have to do is plug it in your computer and go. Should not have to mess with AB or any type of bios firmware switching.

 

[Headfoot]

Ok but when I go out to buy a new card, I see a $650 Fury X and a $690 Gigabyte G1 that's 25% faster out of the box with 0 overclocking required. What am I going to buy/recommend? Overclocking is not guaranteed but that 25% faster performance is warrantied by Gigabyte. That's not even the fastest 980Ti as Zotac AMP! 980Ti should have 1355mhz Boost out of the box.

AMD pulled a 6970 vs. GTX580. Performance is very good at high rez stock vs. stock but once the latter is overclocked, it's all over. That generation though AMD priced HD6950 at $299 that unlocked and HD6970 at $379 I believe. Today they are trying to pull off a similar thing but price Fury X at $650. Getting crushed by an after-market 980Ti. Fury X needed to be 5-7% faster than a TX to survive the onslaught of after-market + overclocked 980Ti. Since it can't even beat a reference 980Ti at 1440P, it's hopeless now without a dramatic 10-15% driver improvement and some incredible voltage overclocking. But the latter could result in Fury X using 150W+ more power than an after-market 980TI OC.

We don't know how Fury X overclocks until we get voltage unlock. This is fundamental, and basic.

Max overclock vs Max overclock is always what enthusiasts care most about. If it uses more power that's pretty much completely immaterial. Enthusiasts dont buy top end gear and overclock it to save power. We also don't know if Fury will be cut down or full chip, just on air. Factory OC Fury is almost a certainty. It could be all the cut down chips are going into Nano.

The picture is not complete yet, and all this "The Sky is Falling" nonsense is just noise until ALL the Fury cards have dropped and its full lineup versus full lineup comparison.

 

[Azix]

Factory overclock card. All you have to do is plug it in your computer and go. Should not have to mess with AB or any type of bios firmware switching.

And we could expect factory overclocked fury x cards as well. If we can't change the voltage right now, it is possible board partners will. I think AMD said they can't change the cooler most of all, but they still build the cards so if there is OC to be had, it will be had. They won't all be selling the same card I think.

I mean its a water cooled card and the VRMs themselves are watercooled. Things could get either very interesting with cards from the partners or just be very boring. I think more interesting than boring. The review samples might be just a sample.

 

[krumme]

Ok but when I go out to buy a new card, I see a $650 Fury X and a $690 Gigabyte G1 that's 25% faster out of the box with 0 overclocking required. What am I going to buy/recommend? Overclocking is not guaranteed but that 25% faster performance is warrantied by Gigabyte. That's not even the fastest 980Ti as Zotac AMP! 980Ti should have 1355mhz Boost out of the box.

AMD pulled a 6970 vs. GTX580. Performance is very good at high rez stock vs. stock but once the latter is overclocked, it's all over. That generation though AMD priced HD6950 at $299 that unlocked and HD6970 at $379 I believe. Today they are trying to pull off a similar thing but price Fury X at $650. Getting crushed by an after-market 980Ti. Fury X needed to be 5-7% faster than a TX to survive the onslaught of after-market + overclocked 980Ti. Since it can't even beat a reference 980Ti at 1440P, it's hopeless now without a dramatic 10-15% driver improvement and some incredible voltage overclocking. But the latter could result in Fury X using 150W+ more power than an after-market 980TI OC.

Looking at the 596mm2 and far less space used for mem controller i think we are in for the 10-15% improvement relative vs980ti. At least. But it will take half a year.
I have no doubt amd will have to lower price if they want to move any cards in numbers that matters. The pro and nano is for that. The bom must be quite lower. Especially the nano. How expensive can a gfx so small be..lol.
My guess is nano for 350 within end this year. Pro for 400. At most.

 

[lefty2]

Does anyone know why Tom's Hardware has the card using less power than 980 Ti and all other reviews show more power draw?
I notice that Tom's Hardware seem to take a lot more effort to measure the power precisely, producing a huge amount of graphs, while the other reviews just produce a single number.

 

[exar333]

And we could expect factory overclocked fury x cards as well. If we can't change the voltage right now, it is possible board partners will. I think AMD said they can't change the cooler most of all, but they still build the cards so if there is OC to be had, it will be had. They won't all be selling the same card I think.

I mean its a water cooled card and the VRMs themselves are watercooled. Things could get either very interesting with cards from the partners or just be very boring. I think more interesting than boring. The review samples might be just a sample.

This might get interesting when OCing the card. According to Tom's, the VRM temp was 'acceptable' at gaming but very high during a stress test. It will be interesting to see what this looks like when voltage is upped and OC'd to the max.

http://www.tomshardware.com/reviews/amd-radeon-r9-fury-x,4196-8.html

 

[Enigmoid]

Judging by the few reviews that overclocked, it looks like fury will not overclock well and even if the clockspeed is cranked up something else is holding the card back.

IMO this is a $599 card at best. The CLC cooling is nice and quiet but with the limited OC it seems rather pointless for the additional cost.

The limited OC also has me scared for the Fury (vanilla version), especially if its more cut down as it won't be able to compete with the X via. OC.

 

[Azix]

The specs from sapphire seem standard. Time will tell if they have better models I guess. The other sites list the same 1050 Mhz.

http://www.sapphiretech.com/productdetial.asp?pid=69ED4799-7518-434C-80CD-3FF8811F8648&lang=eng

https://www.visiontek.com/graphics-...k-radeon-r9-fury-x-detail.html#specifications

http://xfxforce.com/en-us/products/...md-radeon-r9-furyx-liquid-cooled-r9-fury-4qfa

I will be surprised if they are all the same with no special models (specs-wise). But seems more likely now.

 

[dark zero]

The problem is not the HBM. It was OK, the problem is the AMD weak core. That's why they failed hard

 

[Piroko]

I couldnt find any in europe with stock. And most specify july as delievering dates (2nd and 6th).

alternate.de seems to have stock, at 729€. --never mind, just rechecked and stock's gone.

Does anyone know why Tom's Hardware has the card using less power than 980 Ti and all other reviews show more power draw?
I notice that Tom's Hardware seem to take a lot more effort to measure the power precisely, producing a huge amount of graphs, while the other reviews just produce a single number.

Noticed that as well. The PCIE power draw is astonishingly low, guess that's why they gave that card two 8-pin connectors.

 

[RaulF]

And we could expect factory overclocked fury x cards as well. If we can't change the voltage right now, it is possible board partners will. I think AMD said they can't change the cooler most of all, but they still build the cards so if there is OC to be had, it will be had. They won't all be selling the same card I think.

I mean its a water cooled card and the VRMs themselves are watercooled. Things could get either very interesting with cards from the partners or just be very boring. I think more interesting than boring. The review samples might be just a sample.

That would be nice. But i can't recall the 295X2 getting an OCed model, and if there was i am pretty sure it was very mild (not worth it).

 

[AtenRa]

There are people working in AMD that (i will put this mildly) doesnt know how to play the game.

1: You give your new shiny GPU to review sites for evaluation with a 2-3 page review guide that specifically says with capital letters "YOU DO NOT TEST ANY GAMEWORK TITLE" and "YOU INCLUDE AS MANY AS POSSIBLE MANTLE SUPPORTING GAMES". And you imply that anyone not following those guidelines will not receive a sample again.

2: You dont release the damn card after NVIDIA release the GTX980Ti if you cannot have the same performance as TITAN-X. They had to release it right after TITAN-X at the $649 price. That way Fury would make a splash.

3: If you couldn't release before the GTX980Ti then you wait for Windows 10 and better drivers to increase performance (including number 1 above) and get what you after.

Now everyone will remember that Fury X is slower than GTX980ti and the work to undone this is tremendous because you are AMD and not NVIDIA.

And this Fury X launch is coming too close to R 300 fiasco. Heh, i would have sworn they are doing it in purpose and they afraid to be in the lead.

 

[bryanW1995]

sorry but I don't listen to hardOCP, they've made a few too many journalistic blunders to be trusted as a neutral authority

do you have any FPS graphs to show of this?

LOL

Journalistic blunder = "Shattered Glass" or just "Incompetence"?

 

[badb0y]

We don't know how Fury X overclocks until we get voltage unlock. This is fundamental, and basic.

Max overclock vs Max overclock is always what enthusiasts care most about. If it uses more power that's pretty much completely immaterial. Enthusiasts dont buy top end gear and overclock it to save power. We also don't know if Fury will be cut down or full chip, just on air. Factory OC Fury is almost a certainty. It could be all the cut down chips are going into Nano.

The picture is not complete yet, and all this "The Sky is Falling" nonsense is just noise until ALL the Fury cards have dropped and its full lineup versus full lineup comparison.

What? I don't see how AMD Fury X is going to catch a GTX 980 Ti just through overclocking since GTX 980 Ti is a very good overclocker on it's own right.

 

[bryanW1995]

That very same sentiment is used whenever AMD cannot keep up with Nvidia in performance for a given feature.
The very same was said about PhysX. Tesselation.
Getting old.

Perhaps things changed after I left in 2012, but, as I recall, NV used some dirty tricks to artificially hinder performance on AMD cards....and once a hacked driver came out with a workaround, the AMD cards did just fine with it. Am I misremembering, or did things change in the past couple of years?

The gains are obviously worth it... especially when you don't pay for your cards.

I can't remember the last time i used a gameworks feature on my system.

Is that really necessary? He was a huge NV fan long before he joined FG. Veiled criticisms like this are not conducive to a healthy debate on graphics cards.

 

[Azix]

This might get interesting when OCing the card. According to Tom's, the VRM temp was 'acceptable' at gaming but very high during a stress test. It will be interesting to see what this looks like when voltage is upped and OC'd to the max.

http://www.tomshardware.com/reviews/amd-radeon-r9-fury-x,4196-8.html

Looks like they did need a fan on that then. I wouldn't assume its the VRMs specifically causing the increased temps but they are likely hot. Tiny card with lots of components not being cooled by a fan. The VRMs themselves could still be benefiting from the water cooling and be lower. They should have taken off the back for a better look

Software readings would be huge help

 

[LTC8K6]

alternate.de seems to have stock, at 729€. --never mind, just rechecked and stock's gone.

Noticed that as well. The PCIE power draw is astonishingly low, guess that's why they gave that card two 8-pin connectors.

It's generally not good to pick one review that has a different result and use that to say something definite.

Typically if one reviewer has a different result from the rest, it is suspect?

 

[bryanW1995]

That's the thing a lot of people are not recognizing the significance of. AMD was the value entity, make smaller GPU that offer good bang for buck but yielded the battle to NV for the performance crown.

People even said they couldn't make a big GPU. The reasons were lower R&D investment, low yields, low returns, no HPC market like NV has to subsidize development of such a monolithic beast...

This narrative leaves out the fact that AMD had the same opportunity to go after that HPC market and, instead, they chose to go after Intel in cpus/apus. Smart move, guys. You have to give JHH lots of credit for aggressively pursuing that high growth market (one that didn't have much competition, either).

and............AT goes live with a motherboard review instead of being on time with Fury X review

EDIT: Apparently Ryan Smith is sick and should have it up in the next few days

LOL

He should just hire BFG10k to do all of his reviews. AT doesn't need to be first with their review of things like this, but to be several days late is pretty bad. I'm guessing that a similar snafu is what convinced Anand to step back from doing all of the Video Card reviews bitd.

 

[omek]

This might get interesting when OCing the card. According to Tom's, the VRM temp was 'acceptable' at gaming but very high during a stress test. It will be interesting to see what this looks like when voltage is upped and OC'd to the max.

http://www.tomshardware.com/reviews/amd-radeon-r9-fury-x,4196-8.html

It would be nice to know what that stress test was and what the VRM temperature was while gaming. You could imagine that Furmark uncapped could easily coax the GPU into using quite a bit more power, if that's true the 100*C VRM temperature isn't all that bad given that the closed loop was overcome and the fan also wasn't ramping up.

Very interesting and most likely easy for the AIB's to overcome by beefing up the power delivery or the user by adding a 120mm screamer.