|
would it kill them to just use 5 different variations of a new chip in the product stack instead of the entire chip library dating circa 2011
|
# ? May 14, 2016 01:04 |
|
|
# ? Jun 8, 2024 16:00 |
|
~the SeCrEts of PaScAl~ http://vrworld.com/2016/05/10/pascal-secrets-nvidia-geforce-gtx-1080/
|
# ? May 14, 2016 01:05 |
fozzy fosbourne posted:~the SeCrEts of PaScAl~ http://vrworld.com/2016/05/10/pascal-secrets-nvidia-geforce-gtx-1080/ garbage article.
|
|
# ? May 14, 2016 01:24 |
|
Anime Schoolgirl posted:would it kill them to just use 5 different variations of a new chip in the product stack instead of the entire chip library dating circa 2011 lovely hypothesis: the M400 series is a way to get rid of the very last of their 28nm chip stock (Grenada excluded), maybe even most of the tuned Fury cards (retired into either Pro Duo or M495X), and is exclusively a mobile line up; let Grenada go on super firesale. The 500 and M500 series are Polaris/Vega. Possible supporting evidence is HXCF for Bristol Ridge?
|
# ? May 14, 2016 01:26 |
|
AVeryLargeRadish posted:garbage article. quote:By Theo Valich
|
# ? May 14, 2016 01:26 |
|
fozzy fosbourne posted:~the SeCrEts of PaScAl~ http://vrworld.com/2016/05/10/pascal-secrets-nvidia-geforce-gtx-1080/ https://youtu.be/Ix2cvYQPpQM
|
# ? May 14, 2016 01:46 |
|
SlayVus posted:That just doesn't make sense to me. NDA the date the NDA gets lifted. 1080 is going to perform high enough that it's just positioned relative to that top performance slot that warrants its price. 1070 has the potential to be a lot closer to polaris 10, so keeping the clocks and stuff in flux so they can tune them where they need to be to beat it is more important.
|
# ? May 14, 2016 05:36 |
|
FaustianQ posted:lovely hypothesis: the M400 series is a way to get rid of the very last of their 28nm chip stock (Grenada excluded), maybe even most of the tuned Fury cards (retired into either Pro Duo or M495X), and is exclusively a mobile line up; let Grenada go on super firesale. The 500 and M500 series are Polaris/Vega. Possible supporting evidence is HXCF for Bristol Ridge? The other way around would be saner, use all the Polatis 11 chips on mobile and stick with rebrands for low end desktop cards because they're low margins as is and no one gives a gently caress anyways.
|
# ? May 14, 2016 09:25 |
|
EdEddnEddy posted:If that has any chance of happening again, I am ok with waiting till the end of the year for a Ti that literally curbstomps the 1080 with sheer performance for $200 But then Volta comes out soon after and then that'll be better but then AMD might have something which might be better and
|
# ? May 14, 2016 09:52 |
|
There won't be anything new on the very high end until 2018 or so and 2017 is only a little more than 6 months away. On top of that AMD might've pulled forward Vega to this Oct. Waiting over a year is fairly silly but waiting 5-6 months or so isn't a terrible idea. Particularly if you already have a high or mid range card and want a big jump in performance to last you a while.
|
# ? May 14, 2016 10:46 |
|
OC3D thinks there's going to be a watercooled 1080 clocked at 2.5ghz
|
# ? May 14, 2016 14:08 |
|
lol. They are sourcing a guy on their forums, who sourced this, who are sourcing wccftech, who were just sourcing that crazy "secrets of pascal" article I posted earlier
|
# ? May 14, 2016 14:45 |
|
So much misinformation jesus This NDA's expiry cant come soon enough.
|
# ? May 14, 2016 14:46 |
|
fozzy fosbourne posted:lol. They are sourcing a guy on their forums, who sourced this, who are sourcing wccftech, who were just sourcing that crazy "secrets of pascal" article I posted earlier oh, lol
|
# ? May 14, 2016 14:52 |
|
fozzy fosbourne posted:lol. They are sourcing a guy on their forums, who sourced this, who are sourcing wccftech, who were just sourcing that crazy "secrets of pascal" article I posted earlier The ouroboros of garbage journalism is complete
|
# ? May 14, 2016 16:19 |
|
AMD releases experimental AMDGPU driver for GCN1 Thought some good news would be in order. This is in combination with a fairly massive update in general for their linux drivers, as well as adding OverDrive support. That and them working furiously to get Vulkan up and running for all GCN versions makes me hopeful. It's kind of too bad VLIW will be forgotten, but I get it they're old and it's a large enough workload to get everything done for GCN let alone another µarch.
|
# ? May 14, 2016 16:48 |
|
but why hasnt amd released a video graphics cards yet this is reporter for GPU megathread news reporting, asking the hard hitting questions nobody dares to ask
|
# ? May 14, 2016 18:05 |
|
I wonder if we could go to an expo or something and say we're here on behalf of Serious Hardware/Software Crap, and get press passes.
|
# ? May 14, 2016 18:20 |
|
Maybe SHSC should start doing benchmarks and reviews, with the usual goon autism, insufferableness and snark.
|
# ? May 14, 2016 18:25 |
|
FaustianQ posted:Maybe SHSC should start doing benchmarks and reviews, with the usual goon autism, insufferableness and snark. Gonna be honest, it'd be above a decent amount of places. I'd do it. Granted part of the reason I'd do it is to have somebody other than Computerbase.de doing 3440x1440 benches.
|
# ? May 14, 2016 18:50 |
|
I've been thinking. Apart from the Furies, which were bad for various reasons, this is going to be AMD's first new GPU launch in like 6 years, where all the stock isn't going to get sucked into the black hole that is cryptocurrency mining.
|
# ? May 14, 2016 19:07 |
|
What makes you think P10 and P11 won't also all be bought by butters for mining new idiotic flavors of cryptomonopoly money After all, you can put more of them in a PC without having to buy a bigger powersupply
|
# ? May 14, 2016 19:09 |
|
Wait, they're still doing that? I thought GPU mining was over. Why did I buy an nvidia
|
# ? May 14, 2016 19:11 |
|
It's fine, it's not anywhere near as profitable as it used to be plus it's been largely discredited by anyone with actual assets now.
|
# ? May 14, 2016 19:13 |
|
As soon as Polaris launches there's going to be a new bitcoin, you'll see! IT'LL MAKE US RICH.
|
# ? May 14, 2016 20:02 |
|
Deleted Tweet from LinusTechTips: "Durr.. Just got a memo from NVIDIA that you can get a software key of some sort to unlock 3way and 4way SLI on 1080!" Can we not go down that road Nvidia?
|
# ? May 14, 2016 20:32 |
repiv posted:Deleted Tweet from LinusTechTips: "Durr.. Just got a memo from NVIDIA that you can get a software key of some sort to unlock 3way and 4way SLI on 1080!" What do you mean?
|
|
# ? May 14, 2016 20:39 |
|
I think maybe they mean the 980Ti DLC Edition road.
|
# ? May 14, 2016 20:40 |
|
HMS Boromir posted:I think maybe they mean the 980Ti DLC Edition road. Exactly. He doesn't say they're charging for the unlock key, but if it's free then what's the point of locking it in the first place?
|
# ? May 14, 2016 20:42 |
repiv posted:Exactly. He doesn't say they're charging for the unlock key, but if it's free then what's the point of locking it in the first place? I figured that they don't have 3 way and 4 way SLI working the way they want yet so it's software locked. They would tell the press this so that people stop spreading the "1080 can only do 2-way SLI!" rumor, the 1080/1070 might not have it at launch but the hardware can do it once Nvidia gets it working properly.
|
|
# ? May 14, 2016 21:07 |
|
I hope your guess is right. Would hate to see consumer products adopt the madness of the test instrument industry, where you can buy a product with a colour LCD but have it display in solid green unless you buy a loving licence code
|
# ? May 14, 2016 21:26 |
|
Really, if they needed more bandwidth for SLI, why not increase the contact density and/or size of the fingers? Because as it is now, 26 contacts don't have enough bandwidth for SLI. So now you can use 52 contacts on the 1080, but limit expandability. Comedy option - Use displayport as a SLI interconnect. One SLI finger does 1GByte/s, DP1.3 does 3.24GByte/s after overhead. SlayVus fucked around with this message at 21:44 on May 14, 2016 |
# ? May 14, 2016 21:37 |
|
SlayVus posted:Comedy option - Use displayport as a SLI interconnect. One SLI finger does 1GByte/s, DP1.3 does 3.24GByte/s after overhead. The bidirectional data channel is only 720Mbit/s sadly, and half-duplex. It would be hilarious to see Nvidia resort to the same kludge ATi used to rush first gen Crossfire to market though.
|
# ? May 14, 2016 21:57 |
|
FaustianQ posted:Kind of leaves the door open on 2017 custom 1080s that perform noticeably better using 2 8 pin connectors and 14Ghz GDDR5X. This, except it will be called 1080Ti
|
# ? May 14, 2016 22:24 |
|
I've got a really fun idea. They could go through the PCIe bus like sane people rather than just doubling the inadequate SLI finger connection.
|
# ? May 15, 2016 00:41 |
|
but then they'd have to pay amd and intel licensing fees
|
# ? May 15, 2016 00:49 |
|
Wait, is that seriously why they keep doing SLI through fingers? o.o
|
# ? May 15, 2016 00:51 |
|
froody guy posted:This, except it will be called 1080Ti I guess there is precedent for a GX104 chip to be considered flagship so that's really not outside possibility. But rather than a 1080ti, just the new Titan (G? X2? Y? K? P?) or something, 900$.
|
# ? May 15, 2016 03:03 |
|
FaustianQ posted:Titan (G? X2? Y? K? P?) My money is on "Titan 10".
|
# ? May 15, 2016 03:05 |
|
|
# ? Jun 8, 2024 16:00 |
|
Subjunctive posted:My money is on "Titan 10". Or they move to other colors on ROYGBIV plus White. Somehow I think they'd skip the Titan Red, though.
|
# ? May 15, 2016 03:08 |