Some new Fury X pics from PCPer
http://www.pcper.com/news/Graphics-...X-Graphics-Card-Pictured-Uses-2-x-8-pin-Power
http://www.pcper.com/news/Graphics-...X-Graphics-Card-Pictured-Uses-2-x-8-pin-Power
AMD Radeon Fury X Series | HBM, Small Form Factor And Water Cooling | June 16th
- Thread starter wachie
- Start date
King Gilga
Member
Want to see how the fan version/pro look more tbh.
Me too. I have a feeling AMD is going to make a ton of money from Fiji Pro.
That's one chunky fan
LQX
Member
Wow, that fan looks huge and might be right up on some CPU coolers like the Cooler Master Hyper 212. Also the tubing being on the side my also cause conflicts as you have to then bend it below the CPU cooler. I hope they throughly tested that thing with different settups.
fizzelopeguss
Banned
Fan looks like a Gentle Typhoon.
Like it has been said already Ryan is one of the most obvious pro-Nvidia anti-AMD reviewers out there. He gets exclusives from Nvidia, why should he expect the same from AMD?
Probably not new news and probably already known but, Radeon Fury X reportedly reaches 8.6 TFLOPS FP32 (single precision) performance.
64 compute units each with 64 SP (4096 SP).
http://www.tweaktown.com/news/45851...-rocks-4gb-hbm-4096-bit-memory-bus/index.html
http://wccftech.com/amd-radeon-fury-x-fiji-based-graphics-cards-synthetic-benchmarks-revealed/
Who will be the first with a 10+ TFLOP gaming GPU:
AMD with 400 / Arctic Islands or Nvidia with Pascal ?
64 compute units each with 64 SP (4096 SP).
http://www.tweaktown.com/news/45851...-rocks-4gb-hbm-4096-bit-memory-bus/index.html
http://wccftech.com/amd-radeon-fury-x-fiji-based-graphics-cards-synthetic-benchmarks-revealed/
Who will be the first with a 10+ TFLOP gaming GPU:
AMD with 400 / Arctic Islands or Nvidia with Pascal ?
Probably Nvidia? Flop race does not always translate into gaming performance but its still a usefull metric for other applications of these big GPUs.
I know that, just wanted to guess who'll be the first to get those bragging rights, as that's all it would mean.
AMD have usually been ahead in the FLOP race though. 980 ti and R9 290X are basically equal in FLOPs and if you factor in die size the R9 290X is more efficient than the Titan X.
As you said it is pretty much useless. We generally game with these cards so all it can do is hint at untapped potential
And this is the damned reason I've been on my Core i5-750 since September 2009, pretty much every new CPU's has been "not the game changer we were looking for". Yes yes, upgrading to even a Haswell now would be a boost for me, but Skylake was supposed to be the thing to wait for, so I waited, and now, it's not, and Cannonlake might be the one to wait for... blah
The last few CPU and GPU generations have really put a wrench in my computer upgrade schedule, which had up to now firmly been on a 3-4 year cycle for a full upgrade, with motherboard/CPU upgrade on year 1, and 2 years later a GPU upgrade, alternating like that, back and forth.
/rant.
DDR4 will be the gamechanger along with Cannonlake (or Zen).
My strategy is to upgrade when the RAM is upgraded - by then you'll get a massive performance boost.
We don't know yet - but since it's Intel probably not. Hence why better to wait. With the extra money saved up you could go Crossfire/SLI off the bat! Or upgrade the monitor at the same time (Freesync/G-Sync monitors would be more common and hopefully cheaper).
It should be though. Cannonlake is just going to be the usual same architecture/smaller node like what Ivy Bridge vs Sandy Bridge was.
I'm probably going to do the same thing. Originally I was considering upgrading to Skylake if the jump's big enough to ditch my 2500k, but with Zen announced I might as well see what gives.
A guy at Something Awful has a 390X as well and...
(This is Firestrike Extreme 1.1.)
Maybe it's just bad drivers, right? ....Right?
I have read lots of conflicting rumours but one has been that the 390X is based on the Grenada core. This is essentially Hawaii but respun at GF and utilising enhancements in Tonga. GF has much higher transistor density than TSMC so AMD could make a die that is exactly the same as Hawaii with the Tonga enhancements and have it take up less area making it cheaper to produce / die.
If those firestrike numbers are genuinely of a 390X it does suggest the installed driver is far from optimal and the only reason the driver would be that far from optimal is if the 390X is not just a simple rebrand. Of course it could be a fake graph as well so it is not conclusive in any way shape or form.
It will likely be AMD. I predict AMD will get their 16/14nm card out before NV because AMD have a track record of hitting the die shrinks sooner than NV anyway. On top of that NV have to figure out HBM where as AMD will already have experience of that from the Fury line.
Of course comparing NV flops to AMD flops to predict gaming performance is stupid and does not correlate as flops are just one part of a much bigger picture.
Coulomb_Barrier
Member
Doubt it, there's zero performance boost from DDR4 in gaming right now. Currently the only tangible benefit from it seems to be lower power draw compared to DDR3. When manufacturers really start pushing the frequencies of the memory (past 3200Mhz) then things may change. And I'm speaking as someone who has 16GB 2400Mhz DDR4 in my system.
EDIT: Zen is going to be a really interesting proposition. I'm personally very excited to see AMD's Zen APUs with HBM controllers.
Don't APUs / iGPUs tend to benefit from higher RAM speed? Or will things like HBM or other workarounds to more to benefit those?
michaelius
Banned
AMD have usually been ahead in the FLOP race though. 980 ti and R9 290X are basically equal in FLOPs and if you factor in die size the R9 290X is more efficient than the Titan X.
As you said it is pretty much useless. We generally game with these cards so all it can do is hint at untapped potential
Both companies measure flops diffrently - AFAIK Amd measures them at maximum clock speed while Nvidia measures them at basec non boost speeds.
Yh that seems to be the only use case u see big improvements in gaming. Given the bandwidth HBM can give it should be interesting to see what kind of performance igpus/apu with get with some embedded.
Coulomb_Barrier
Member
Well think about it - current AMD APUs have graphics processing cores but unlike discrete GPUs with GDDR5 memory, APUs are fed by slow-ass DDR3 memory. It is a bottleneck for their performance, which will be lifted almost totally when AMD produce their Zen APUs with HBM (likely HBM2) sometime at the end of next year or later.
jfoul
Member
Coulomb_Barrier
Member
With the red LED writing on the side and general compactness, the Fury X has got to be the best looking card AMD has produced, regardless of it's performance.
With that big Gentle Typhoon fan, I'm guessing it's going to be very quiet too. I think they've learned from their disastrous R9 290/X launch.
With that big Gentle Typhoon fan, I'm guessing it's going to be very quiet too. I think they've learned from their disastrous R9 290/X launch.
jfoul
Member
Yeah, the card looks really nice. Nice finish, braided tubes, quality fan, nice thick rad.
Not sure about that just yet, DDR4 will have the same issue all new RAM versions have had, the higher latency will eat up the performance advantage of the higher clock. So it's still a ways off.
Holy shit that radiator is huge, thought it was going to be the size of the Corsair h80i or something.
Its a rad and fan, so maybe a 40mm rad. I suppose its that or a thinner 240mm one that less people would have space to mount.
dgrdsv
Member
I've said this each time a rebrand happened - it doesn't really matter what it was before if the card is competitive.
The issue with 300 series rebrands though is that FL12_0 support looks to be somewhat sparse in it which is bad since I fully expect most of titles coming on Xbox One as well as PC to switch to FL12_0 by the end of this year.
Thing is there is no confirmation not only for Fiji Pro but for a non-reference air cooled Fiji XT even. I have a feeling that a lower positioned Fiji based card is still some months away.
Nah, AMD is a more likely candidate. Their FLOPs are not as affective as NV's so they have to push higher on them to remain competitive. This was the case since G80.
It puzzles me that the card itself is two slot wide for some reason.
You expected three slots? I thought most AIO coolers without VRM fan were two slots so this doesn't come as a surprise.
RenegadeXV8
Banned
Actually the score is just fine. He was benching Firestrike Extreme (http://i.imgur.com/bDr6gif.jpg). It is virtually equal to my 290X clocked to 1000MHz:
390: 4959, 5165
290X: 4953, 5192
So if the 390 is a straight up rebrand of the 290X, then is the 390X simply a slightly overclocked 290X? If so, why would anyone pay $80 more for that?
ScepticMatt
Member
Not sure about biased but they are obviously Nvidia focused.
They have exclusive podcasts with new Nvidia technology, and they have made bullshit claims about FreeSync ('no frame doubling') and overhyped the Titan X ('twice as good as 980' my ass)
I favor Techreport and their 99% frametime benchmarks. And computerbase for German websites.
frontieruk
Member
I've been pondering that too..
I imagine we're going to find that out when AMD actually announce them and we get actual benchmarks from reliable sources on new drivers.
I cant find any of those 390x results using the search feature on futuremark's website, does anyone else have a direct link to em, and not just a screenshot of a result?
Maingear solves dual/triple aio like this
Maingear solves dual/triple aio like this
Sure, but the air passing trough the radiators wont heat up enough to become a problem for the motherboard or any other component that it passes over.
It is, as they all probably are.
It's all sending heat outside then. Still that wouldn't work for me due the HDDs taking space in front of the Front fan.
If I do get the Fury X, I'll probably place it on the Rear Fan (since it has cooling tubes cutouts and I can place the radiator outside the case).
For the CPU, I'll probably move it to the top of the case, blowing outwards as well.
If I do get the Fury X, I'll probably place it on the Rear Fan (since it has cooling tubes cutouts and I can place the radiator outside the case).
For the CPU, I'll probably move it to the top of the case, blowing outwards as well.
dgrdsv
Member
I expected one, like on the most WC solutions I know.
GPUz uses look up tables from manufacturers nowadays. That's why nobody knew about the 970 until we forced Nvidia to tell us.
I've just decided to make myself a watercooled rig for the first time in about 10 years instead. The only one I ever had was a waterchill kit and I can't even remember if it did my GPU or not. The rads sat outside the case just dangling about lol.
Got myself a nice Phantex Evolv ATX on the way in about 3 weeks and will have a 360mm x 60mm rad and a 280mm x 30mm rad doing the work for me. Just not sure yet which GPU it'll be cooling!
Of course, if I have to waste extra money on a Fury that already comes with a rad and stuff then that'll influence my decision.
Got myself a nice Phantex Evolv ATX on the way in about 3 weeks and will have a 360mm x 60mm rad and a 280mm x 30mm rad doing the work for me. Just not sure yet which GPU it'll be cooling!
Of course, if I have to waste extra money on a Fury that already comes with a rad and stuff then that'll influence my decision.
It needs a heatsink that covers all the components and also AIO pumps tend to be thick. I've never seen a thin pump, where the thinnest i've seen is just slightly under the thickness of two pcie slots.
Most WC solutions don't have an integrated pump on the heatsink.
All the components are all stacked in the same place now. Thats why the card is very short
The previous poster wasn't asking why the card is short, but rather, why it was two pcie slots thick.
Coulomb_Barrier
Member
I never expected it to be anything less than 2 PCIE slots thick.
So excited to find out in a few days whether the Fury X is overrall faster than a Titan X or not.
So excited to find out in a few days whether the Fury X is overrall faster than a Titan X or not.
Didnt Nvidia already tape out GP100? (as per the rumor) And there are no rumors yet on any AMD chip taping out on 16FF, if that's the case then Nvidia will win the flop race. Whether they bring that to the gaming market is another question.
General Lee
Member
I believe Nvidia has stated they'll use TSMC 16 nm FF+ for their GPUs, and the predictions about that seem to vary widely. Most optimistic think Nvidia would be ready in Q1 2016, and the most pessimistic think 2017. I wouldn't bet much on big Pascal being out all that soon just based on previous track record with new nodes.
Granted it's one post but it's something - https://forum.beyond3d.com/posts/1848329/
I havent read the same for AMD even though the general expectation is for them to adopt the new process first.