AMD Radeon Fury X Series | HBM, Small Form Factor And Water Cooling | June 16th
- Thread starter wachie
- Start date
Irobot82
Member
Paying $80 for double the VRAM from 4GB to 8GB I guess.
Those benchmarks have to be wrong. A 390x with the improvements of Tonga should be a very decent GPU.
http://gpu.userbenchmark.com/Compare/AMD-R9-285-vs-AMD-R9-280/3174vs2241
http://gpu.userbenchmark.com/Compare/AMD-R9-285-vs-AMD-R9-280/3174vs2241
Yeah it's all conjecture and speculation until at least the 16th and even then it depends on how much AMD tells us at the PCGamer show at E3. A lot is still going to depend on when the various gaming sites get their cards and can actually run real world benchmarks AND if the product is actually available for purchase on shelves.
Well tape-out is just start of fabrication of the chips and there are many more steps to a working graphics card.
I remember reading back in the day it takes about six months to go from tape-out to full production. I wonder if that's still the case nowadays when smaller nodes are harder to ramp up. Then there's also HBM2 that needs to be ready.
I figured both AMD's and Nvidia's new cards would launch around mid to late 2016, but who knows it might be sooner.
anchorman128
Member
It'll be around this time frame next year that we will be seeing these new cards. That is a fairly accurate estimation in my opinion, considering they have already been taped out. However I could see HBM2 and fabrication on a new node possibly giving manufacturers trouble. I would say Q2-Q3 2016 new cards from both companies will be out. No way they let it slide into 2017, or we are going to be getting ANOTHER round of refreshed cards.
NumberThree
Member
There may also be multiple tapeouts before the chip is in its final, productized form.
That's weird. If I had the card I could run a simple device query and find the number of compute units (OpenCL). I figured GPUZ is able to query the GPU the same way.
Unknown Soldier
Member
All GPU-Z does is query the BIOS string and then it fills in the blanks with what it knows the card's official specs are.
https://www.youtube.com/watch?v=6kEcbLRnbVI
Unboxing guy from yesterday benched the 390X, slightly faster than the 970 in 3DMark, same as the 290X basically.
Linus also weighed in on the rumors - http://linustechtips.com/main/topic...-time-for-the-16th-reveal/page-8#entry5210707
Unboxing guy from yesterday benched the 390X, slightly faster than the 970 in 3DMark, same as the 290X basically.
Linus also weighed in on the rumors - http://linustechtips.com/main/topic...-time-for-the-16th-reveal/page-8#entry5210707
That is one person. Seemingly Anandtech have got review GPUs (going by Ryan Smith's tweets.)
https://twitter.com/RyanSmithAT/status/608414895684595712
I compiled a small C program under windows. (This took me a bit as I use OSX/Linux for C programming.) This program queries the GPU for info concerning execution units, OpenCL version, etc...It seems to me that GPU-Z may be looking up the info (based on what I heard) versus actually querying the device.
I have read the argument that it's just a rebrand moreover, I find this odd. If this is a straight up rebrand, why are the latest beta drivers causing this video card to fail? Not sure if 3DMark is also pulling the information from its database versus querying the device. However if it is, the information it's pulling up may be wrong.
If anyone knows anyone with a 390 vanilla (the one I'm curious about) or the 390x, I have a small program they can run. Not sure how I would get the executable to them moreover, if they have visual studio I have a project they can compile. I understand the concern of asking someone to run a program they got from a stranger on the internet.
I have read the argument that it's just a rebrand moreover, I find this odd. If this is a straight up rebrand, why are the latest beta drivers causing this video card to fail? Not sure if 3DMark is also pulling the information from its database versus querying the device. However if it is, the information it's pulling up may be wrong.
If anyone knows anyone with a 390 vanilla (the one I'm curious about) or the 390x, I have a small program they can run. Not sure how I would get the executable to them moreover, if they have visual studio I have a project they can compile. I understand the concern of asking someone to run a program they got from a stranger on the internet.
XFX 390X
Looks like a reference 290 but with the shitty VRM controller XFX likes to use.
SuperEpicGuy
Banned
Wait, is this the highest/most powerful graphics card for AMD right now? Or is fiji going to be another brand? Cause if the 390x is the highest they've got, they'll be dead once nvidia whips out their 16nm FF+ GPUs.
Fiji is Fury (the title of this thread). 16nm is next year for both Nvidia and AMD.
Unknown Soldier
Member
Keep an eye on this guy:
http://hardforum.com/showthread.php?t=1865280
He's the one with the 390X, he's going to run 3DMark on it and report back. Also he gave us that BIOS dump, anybody with a 290X flashed it yet?
http://hardforum.com/showthread.php?t=1865280
He's the one with the 390X, he's going to run 3DMark on it and report back. Also he gave us that BIOS dump, anybody with a 290X flashed it yet?
Won't say much. It will probably just replicate the poor performance of the other 390x 3dmark benchmarks.
SonataDusk
Banned
My money is ready. More pictures at the above link.
Im very new to water cooling, so I ask, based on these pictures is this water cooling method self contained within the block under the fan? I always thought you had to hook everything up to a central pump which can have concerns for leaking.
This system is what is known as an All in One or AIO. Typically these systems are entirely self contained with the pump integrated into the heatsink and the reservoir integrated into the radiator (where the fan is).
They tend to be safe these days and would require a very dedicated fool to get them to leak.
Insane Metal
Gold Member
Some other dude tried it on his 290X
So just about THE SAME as a 290X. lol AMD
KillaKalz347
Banned
Pretty disappointed to see the rebrands, I would've expected AMD to use the improvements from Tonga and port them over to the Hawaii gpu. Whatever though, of the Fury X launches at $850 and is only a few fps better than the Titan X, I'll just go with Gigabytes 980ti G1 Gaming.
i can get a r9 290x for £210 today. i don't want that wheezing beast chewing at my wallet every time an electricity bill turns up, nor do i want to pay a wedge more for an overclocked version with a bit more RAM that i'll never use on my pissant monitor, nor do i want to move up to the £400+ club.
fucking AMD. the 970 nailed the price point that the mid tier sorta enthusiast goes for, and because of it is a sales monster despite the gimped RAM.
fucking AMD. the 970 nailed the price point that the mid tier sorta enthusiast goes for, and because of it is a sales monster despite the gimped RAM.
exmachina64
Banned
What did you expect them to do?
This is the most exciting product they have lined up till Zen and their financials are abysmal.
They had to follow Nvidia's lead and release a product targeting the Titan's market eventually.
metalhead79
Neo Member
I went ahead and ordered a GTX 970 G1 Gaming. Now that we know the 300 series are just rebrands, I've no interest in dealing with the heat and noise I had from my R9 290 again. I'd been holding out because of the rumors that they *could* be refreshed with GCN 1.2.
Also, I have read that NVENC is better than VCE for Steam in home streaming. Which is something I use heavily.
Also, I have read that NVENC is better than VCE for Steam in home streaming. Which is something I use heavily.
I think it's a bit too early to be benchmarking the 300 series. People should wait for the new drivers, which will probably be released after the cards are actually, y'know, announced.
metalhead79
Neo Member
I wasn't following the news to heavily until my 290 died 4 weeks ago, as I had no intention of getting a new GPU until 2016. I fucked up putting a Kraken G10/H105 on it trying to get it quieted down. The mounting screws on the G10 were too short. I put too much pressure on the PCB trying to get it mounted securely.
But yeah, I guess it was wishful thinking that AMD would update the 290/290X with GCN 1.2. There's no precedent from past rebrands.
Can't wait to try and flash that BIOS to my 290 and see. Free clock bump incoming!
It was a fact that AMD made a deal with Global Foundries, which offered a better 28nm Fab process than TSMC 28nm. Something like a 1 billion dollar deal if I remember correctly. Carraizo, their latest APU uses that process, and a lot of doubt that it used up all their orders. Maybe Fiji now too but, rumors also spread that they would use this to respin Tonga and Hawaii, to more give a boost to efficiency.
Quickest article I found:
http://hexus.net/tech/news/industry/78757-amd-switching-28nm-process-globalfoundries-2015/
I doubt that will work. Those cards are probably of a much higher bin than even the better binned 290s. Plus the bios probably won't contain timings for memories found in the 290 series cards (you should note that the 280x bioses don't work on the 7970 cards).
But hey, do it and report back.
Even then, the new node might be immature and expensive.
It's not the worst idea to get the best 28nm card you can now and then sit out the first Pascal generation for the same-node refresh.
RenegadeXV8
Banned
If we get official specs at E3 then we can easily estimate performance since GCN scales almost linearly.
ThoseDeafMutes
Member
My guess also.
Yes. Either way, it was never going to max games barring some miracle (at least at 60fps). Just wait for benchmarks.
My question is, why didnt AMD just release the Fury line stand alone? Did they know they were going to basically torpedo themselves by putting out a whole new numbered product line with basically zero differentiation from the 200 series?
OEMs want new products AKA new numbers even if they are essentially the same, to coincide with the Windows 10 launch. This isnt the first time this happened and certainly wont be the last.
Nvidia did the exact same thing with the 800 series of notebook cards (well besides the final 860m 2GB which was a maxwell chip)
I'd still be waiting for benchmarks from proper websites that have been given new drivers before making a decision, but each to their own.
If it is indeed true and AMD have literally just changed the box and put the price up with a small clock bump then they should hopefully get slammed in the reviews for it.
If it is indeed true and AMD have literally just changed the box and put the price up with a small clock bump then they should hopefully get slammed in the reviews for it.
It all comes down to price, honestly. If they dont move them down (in price), they will get slammed.
Based on intiial reports, it seems like these are slightly binned chips so even with the core clock bump and additional VRAM, the TDP appears to be same.
Unknown Soldier
Member
Well AMD are the masters of self-defeating
The story is currently that the usual websites aren't being sent review samples before release, which is unusual for video card launches. I suspect this isn't coincidental with the increasing certainty that the entire 300 series are straight-up rebrands of the 200 series.