Maybe, but if Fury X ends up being a beast of a card Nvidia don't be able to do anything about it no matter how high they clock up their GPU. And I don't think Nvidia wants to end up with a space heater of a card they've cultivated a reputation of cool and quiet.
I would only hope for GPU market to become like the good old days.
Back then NV released 6800 Ultra, but that card was beaten by X800XT Platinum Edition, to which NV responded with a faster 6800 Ultra Extreme. Then ATI reclaimed the crown again with an X850XT Platinum Edition. Those were fun times! :biggrin:
If the Titan X is beaten by Fiji XT by even 5%, first, NV will start the world's biggest 6GB-12GB VRAM > 4GB HBM1 viral marketing and review campaign (sites like TechReport and PC Perspective will do "special" tests). At the same time if Fiji uses 290-300W of power, the 40-50W of power differences will be spun like the polar ice caps melting with Fiji XT cards and that Fiji XT would use as much power as an African village.
Then they'll release a fully unlocked 980Ti Black Edition/Metal Edition in the fall or so and while at it also might clock it higher. Don't you worry, NV isn't sweating this one at all. That is before we even get to GameWorks which has proven to swing NV's performance 25-50% at times under the veil of "helping the developers make more advanced graphics."
Except this is nothing like poker. All you do by withholding information is lose potential sales. Sure you get to make a splash but all the people who went out and bought a 980 Ti aren't going to upgrade for 10% more performance. AMD is not apple and can't play this game.
If AMD has a winner on their hands I see no reason why they should not try and trumpet to the rooftops how good this card is: "But wait...look at whats in store consumers. Wait 2 weeks and get a much better deal".
Probably because they made decisions to release the information on June 16th with PC Gamer months ago. You don't just back off from your business obligations and start changing the rules of the game.
Or water cooling isn't the only variable in the equation. One GPU running at 1GHz isn't going to have the same amount of heat output as a different GPU design (or process) at 1GHz.
What part of his 500W statement did you not understand? No single chip card will use 500W of power. How well does a single 120mm AIO CLC handle a 500W R9 295X2? Better than probably 90% of all 980Ti coolers, and on a completely different level to a 980Ti/Titan X reference blower.
The statement you made that Fiji XT might run hot is a physical impossibiilty with an AIO CLC unless the CLC is broken or the system is running in a sauna or something. The chance of a reference 980TI/Titan X running cooler
and quieter at the same time than a Fiji XT AIO CLC is 0%.
Plus, most of the heat from the GPU is being exhausted out of the case.
980Ti reference OC - pure jet engine. Everyone who criticized an R9 290/290X reference for years and is running an overclocked 980Ti is a 100% hypocrite.
Right now Newegg has just 1 after-market 980Ti - EVGA version, the rest are all reference cards. NV came unprepared for the 980Ti launch. They released another hot and loud reference card that the media isn't ripping apart, cuz NV.
At 58% fan speed, the 980Ti is a freaken jet engine. Can't imagine how loud this thing sounds at 100%. :sneaky:
http://www.computerbase.de/2015-06/geforce-gtx-980-ti-test-nvidia-titan/8/
Would it be more believable or would people just assuming its fud?
AMD has said at Computex, Fiji is the fastest GPU in the world. AMD also said at Computex, it won't be more power hungry than R290X.
Isn't that indicative enough?
That massive ~ 600mm2 die size, likely 250-300W TDP, only reinforce my view that those early Chiphell performance leaks and supposed AMD leaked slides showing 4096 shaders, 256 TMUs just may have been right all along, albeit they may have used an older revision of Fiji with lower GPU clocks (1Ghz vs. say 1.05Ghz for the final version as an example).
R9 290X beats 780Ti at 1440P and 4K and that's despite 438mm2 vs. 561mm2 die size.
Per TPU, Titan X is
42% faster than R 290X at 4K
Per Sweclockers, Titan X is
48% faster.
Per Computerbase, Titan X is
41% faster at 4K.
Per Hardware France, Titan X is
44.6% faster at 1440p, and 980TI is 42.4% faster.
That means if AMD releases a card 50% faster than R9 290X, it will beat both the 980Ti and the Titan X.
1050mhz 4096 shaders, 256 TMUs, 64 Tonga-style ROPs, 512GB/sec HBM + 40% Tonga's colour compression, we'd be looking at 52% faster. But R9 290X suffers severely with tessellation and its 64 ROPs are slower than Tonga's 32 (damn). That means if AMD can increase tessellation performance 2-2.5X and real world pixel fill-rate performance 2-2.5X, the extra shaders and textures won't be as bottlenecked by the geometry and ROP units/engines. We could then see performance 55-65% faster in those games where R9 290x was pixel-fill rate and/or geometry bottlenecked.
Facebook
Twitter