Charlie says the interposer is made by UMC on 65nm. Had forgotten they even existed
http://semiaccurate.com/2015/06/22/a...odd-bits-tech/
Charlie says the interposer is made by UMC on 65nm. Had forgotten they even existed
http://semiaccurate.com/2015/06/22/a...odd-bits-tech/
Yup, there's no need for a cutting-edge process for the interposer. Even at 65nm the wiring is far more intricate and complex than the traces out to G5.
DanceswithUnix (24-06-2015)
AMD mucked this launch up IMHO. If there were driver optimisations to be had,they should have delayed the launch and if not they should have launched this at around £400 to £450.
Even under £200 they simply just rejigged a cut down Tonga a bit,instead of launching a fully enabled one which would have given the GTX960 some better competition.
I really hope for their sake,that they can launch their next generation GPUs on the process node before Nvidia does.
The only redeeming feature for me is potentially their OpenCL support,and maybe some aspects of their DX12 support,but TBH I can see my still staying with Nvidia for my next upgrade!
Yeah the interposer is quite interesting IMO - 1011mm^2!!!
I wonder roughly what it costs to make? Only the top metal layers, 65nm* and no transistors would give extremely high yields I imagine.
*WRT that 65nm though - that's a number used to refer to the gate pitch but if there are no transistors I guess it's a bit of a misnomer, only using the 65nm back-end?
Edit: Again WRT die size, I wonder if it's a special process they're using for interposers since I'd think 1011mm^2 would be way above the reticle size of a normal node? Even so, Charlie says they're only a few mm of the max. Even with greater memory support with switching or whatever, there might not have been room on the interposer!
Good to see a review finally. Going to digest all the text a bit more later and take a wider view of other reviews (along with online pricing) before deciding between that and the 980ti. If for some reason (maybe later down the line) you needed to crossfire 2 of these, would the radiator footprint allow both to be attached to 2x the front intakes? If Hexus is able to have a closer look at the radiator then that would be appreciated.
Both the Fury and 980ti are superb cards, more than capable of playing most games at high frame rates (ok maybe not totally at full 4K)
Fury has real potential in a mini-itx system.
It seems to me that AMD missed a big trick. Transpose the placement of the watercooling and power sockets would make (at least for me) a big improvement in tubing and cable runs.
Sadly, it looks at moment if green team have the better card and less power use. But it is not by much and if AMD can produce a 6GB or 8Gb card the performance might be better.
If buying tomorrow I would go green, by Xmas who knows?
The cooler is great, although the card right now is a bit of a poor performer for the cost. Might as well get a 980Ti for around the same price since they overclock FAR better.
That being said, once voltage unlocks come for the fury, we might be able to overclock it so that it laughs at the 980Ti, in which case, its amazing value for money, being faster, cooler, quieter and cheaper than the 980Ti... we will have to wait and see I think though.
I expected more but it's still a beast GPU.
I will definitely wait at least until Christmas to see what will happen with the prices.
The innate problem is that AMD probably realised it could not win this round by a healthy margin. Getting 4,096 cores up to 1,100MHz-plus is no easy task, especially as the cooling is about as good as it's going to get. Even if it had, thereby diminishing the yields substantially, Nvidia could have come back with, say, a GTX 980 Ti-X - a fully core-enabled GPU clocked in at 1,100MHz core.
The way I see it, the efficiencies of Maxwell have trumped the benefits of HBM and disadvantages of GCN... for this iteration of graphics cards at least.
AMD marketing and market research just comes across as half baked when compared to Nvidia in recent years.
Seriously,are they that dumb?? What did they expect from reviews?? They should have dropped the price to £450 RPP with the current set of drivers,so it was not competing directly with the GTX980TI but more with the GTX980. Heck,even only release a limited quantity just for halo reasons if they don't make much money just to strengthen the brand.
Or if there were driver improvements to be had,why launch now?? They should have launched it on the 14th of July FFS.
Plus no voltage unlock on launch day. Seriously,WTF?? The GTX980TI cards are awesome overclockers.
The only way I can see this as a positive is if the actually custom Fury cards are overclocking monsters and make the Fury X look redundant in some way,ie, a bit like the GTX980TI against the Titan X.
Now is the best time to pick up a 295x2
@CAT: I agree it would be a crying shame if there's some untapped potential in driver improvements; release-day reviews count as we saw with the 290X cooler fail. The updated-driver reviews get far less attention, the GCN series in general still seems to be climbing up the performance charts lately compared to release but it's still the initial reviews people will go to.
However I'm impressed with what they've done with power consumption, actual card power being below the 290X from the reviews I've seen so far. I guess I wasn't the only one half-expecting those power claims to be a bit on the optimistic side but they nailed it TBH. It seems they've also done some work on idle and multi-monitor consumption: http://www.tomshardware.co.uk/amd-ra...w-33235-7.html
Either way, this is a good pipecleaner for HBM for AMD - they're near-enough maxed out on both die and interposer size now so there's not much more they could have done hardware-wise without a uArch change or a die shrink. However they now have production experience with HBM so there's less risk than combining it with a new node etc.
Wow, I hope there is something wrong with those benchmarks......super-expensive RAM, water-cooler, 2x 8pin PCI-E connectors.....it should have been a beast but those benchmarks makes it look like a power-hungry flop.
I'd happily take the same-priced, more efficient with more RAM 980Ti over the card with the fancy tech
I can see them being discounted fast to sell them and I wonder where that leaves AMD after the cost of HBM and WC units for each card.....
Main PC: Asus Rampage IV Extreme / 3960X@4.5GHz / Antec H1200 Pro / 32GB DDR3-1866 Quad Channel / Sapphire Fury X / Areca 1680 / 850W EVGA SuperNOVA Gold 2 / Corsair 600T / 2x Dell 3007 / 4 x 250GB SSD + 2 x 80GB SSD / 4 x 1TB HDD (RAID 10) / Windows 10 Pro, Yosemite & Ubuntu
HTPC: AsRock Z77 Pro 4 / 3770K@4.2GHz / 24GB / GTX 1080 / SST-LC20 / Antec TP-550 / Hisense 65k5510 4K TV / HTC Vive / 2 x 240GB SSD + 12TB HDD Space / Race Seat / Logitech G29 / Win 10 Pro
HTPC2: Asus AM1I-A / 5150 / 4GB / Corsair Force 3 240GB / Silverstone SST-ML05B + ST30SF / Samsung UE60H6200 TV / Windows 10 Pro
Spare/Loaner: Gigabyte EX58-UD5 / i950 / 12GB / HD7870 / Corsair 300R / Silverpower 700W modular
NAS 1: HP N40L / 12GB ECC RAM / 2 x 3TB Arrays || NAS 2: Dell PowerEdge T110 II / 24GB ECC RAM / 2 x 3TB Hybrid arrays || Network:Buffalo WZR-1166DHP w/DD-WRT + HP ProCurve 1800-24G
Laptop: Dell Precision 5510 Printer: HP CP1515n || Phone: Huawei P30 || Other: Samsung Galaxy Tab 4 Pro 10.1 CM14 / Playstation 4 + G29 + 2TB Hybrid drive
There are currently 1 users browsing this thread. (0 members and 1 guests)