[Boards: 3 / a / aco / adv / an / asp / b / bant / biz / c / can / cgl / ck / cm / co / cock / d / diy / e / fa / fap / fit / fitlit / g / gd / gif / h / hc / his / hm / hr / i / ic / int / jp / k / lgbt / lit / m / mlp / mlpol / mo / mtv / mu / n / news / o / out / outsoc / p / po / pol / qa / qst / r / r9k / s / s4s / sci / soc / sp / spa / t / tg / toy / trash / trv / tv / u / v / vg / vint / vip / vp / vr / w / wg / wsg / wsr / x / y ] [Search | Extra juicy! | Home]

Pascal

This is a blue board which means that it's for everybody (Safe For Work content only). If you see any adult content, please report it.

Thread replies: 38
Thread images: 3

> 4 TFLOPS double precision floating point

are your anuses prepared?
>>
p neat i guess idk desu
>>
nvidia could literally cancel all pascal shit and go straight for volta, but they need to suck us dry every 2 years :(
>>
File: vf1-ex1-461x346.jpg (35KB, 461x346px) Image search: [iqdb] [SauceNao] [Google]
vf1-ex1-461x346.jpg
35KB, 461x346px
>>51538216

assuming the marketing hype is close to reality and it will be able to achieve close to 4 tflops peak on fp64, does that mean it will be able to do ~8 tflops on fp32? cause fp32 is pretty useful for gaming.
>>
>>51538216

kek, Volta IS NOT coming out in 2017
>>
>>51538330
Current GPUs already do 8+ TFLOP SP
>>
>>51538216

>meaningless comparison to x86 CPUs

nvidishill/10
>>
>that big jump from 2016 to 2017 on x86 CPU
I fucking wish.
>>
>>51538822
Very possible with double the core count.
>>
>>51538713

that slide is from SC15, so the comparison to x86 makes sense in context.
>>
>>51538566

no
>>
which is more important for games, memory bandwidth or latency?
>>
>>51538878

Not really. Everyone knows that graphics cards absolutely fucking destroys x86 in floating point, integer maths and memory bandwidth.

It's just a "WOAH" factor for marketting.
>>
>>51538216
Couldn't a fully enabled Fiji do 4.3Tflops dp64?
How is this impressive?
>>
>>51538933
Generally: latency.
In practice: it depends.
>>
>>51538933
Depends on the game.
You can get around low latency somewhat, you can't get around small bandwidth.
>>
>>51538216
fp64 is irrelevant for gaems and even fp32 is ridiculous overkill for most parts of the usual GPU burner effects.

Though from what I hear the new cards will be pretty damn fast at fp16 too. Shader langs without explicitly sized types are so dead (or on driver profile life support).
>>
>>51539449
>even fp32 is ridiculous overkill for most parts of the usual GPU burner effects.

what does this mean?
>>
>>51539069
powersavings
>>
>>51538911
>The Fury X is capable of a peak rate of 8.6 GFlops (Single Precision)
>>
>>51539889

sauce?

http://wccftech.com/nvidia-pascal-volta-gpus-sc15/

> As for single precision performance, we will see the Pascal GPUs breaking past the 10 TFLOPs barrier with ease.

how does single precision translate into gaming performance?
>>
>>51540456
>asking for source on a fucking simplistic formula
Go back to grade school.

4096 x 1000 x 2 = Over 8 TFLOPS
>>
>>51540456
>how does single precision translate into gaming performance?

It doesn't. None of this has any relevance to gaymen whatsoever. If it did, the Fury X would crush the 980 Ti.
>>
>>51538304
Explain to me how they can just go straight to Volta please, very detailed.
>>
>>51541201

in reality the "2" is just a "1" ;)

> FMAC
>>
>>51541322

you're talking out of your ass. fp32 is the defacto standard in gaming (fp64 in hpc).
>>
>>51539852
But everyone goes 16 nm next gen, so it really is marketing and nothing more.
>>
>>51542687
>>51541322
>>51540456

Only if game is using COMPUTE all time
Gamming is ROP, TMU, bandwidth, arch
>>
>>51539341
>You can get around low latency
low latency is a bad thing right
we want more delay, more is always better :^)
>>
>>51539852
A same-size die of Fiji made in 16nm would have over 14Tflops single/7Tflops double precision at the same power draw

Again, how is this impressive?
>>
>>51538304
>they should make us wait like retarded amdumbs for non-rebranded shit every 3 years
Eat shit
>>
And nothing to do with them because games are made for console and ported to PC.

Quadrofags pls leave.
>>
I will always buy Intel™ because I only play games with Intel Inside™. Intel also pioneers innovative new technologies like Hyper Threading Technology™, Intel Rapid Start Technology™ and the highest quality chipsets to ever grace motherboards.
When I boot up with a brand new Intel™ i7™ with the latest Z chipset, I can enjoy the games the way they where meant to be with Intel Inside™. Intel™ also delivers a far more silkysmooth experience with its Hyper Threading Technology™.
Intel i5™ is also very power efficient. A processor is the most power hungry device in your house. Air conditioners, water heaters, lights, etc all use less power than a processor. Which is why Intel™ puts gamers first by ensuring that their gaming experience is of the highest quality while looking out for gamers by giving them the most value in their electrical bill.
At this point in time, there's really no reasons to consider an AMD processor at all. I tried once, it caused so much heat that it exploded and nearly burnt down my house. It also consumed so much power that it produced an EMP and destroyed not only the rest of my computer but my entire neighborhood.
Intel™ also pioneered how useless MORE CORES is with the i™ series processors. Years ago, everyone thought MORE CORES were the future. Now, Intel™ has debunked that myth entirely and increased efficiency. Now you can save thousands a year in electricity thanks to Intel™ with its powerful IPC. MORE CORES will never be part of Intel's™ line up.
It's quite clear that OPs an AMD shill trying to convince you to settle on something less than the optimal experience with Intel Inside™. Intel™ is the only real way to play games. We have seen recently that they offer incredible libraries for software developers like Intel C++ Compiler. He is probably too poor to afford the Intel Inside™ experience and can not afford to play any games.
Don't be a poor gamer with bad chipsets and a huge power bills. Play games with Intel Inside™
>>
>>51538933
Memory access by GPUs generally follows a predictable pattern, so prefetch works well and latency isn't too bad. That's why GDDR* and HBM are optimized for a high bandwidth over everything else.
>>
>>51538216
Why should anyone care about this shit when video games aren't becoming more fun?

I'd rather play Super Metroid than most 2015 games
>>
>>51538216
Is there a time frame or something of when to expect this?
>>
https://en.m.wikipedia.org/wiki/FLOPS

Built using commercially available parts. Intel Celeron G1830 and AMD Radeon R9 295x2 tops out at over 11.5TFLOPS at a grand total of US$902.57.[55][56]

AMD master race
>>
>>51547750

probably next summer
Thread posts: 38
Thread images: 3


[Boards: 3 / a / aco / adv / an / asp / b / bant / biz / c / can / cgl / ck / cm / co / cock / d / diy / e / fa / fap / fit / fitlit / g / gd / gif / h / hc / his / hm / hr / i / ic / int / jp / k / lgbt / lit / m / mlp / mlpol / mo / mtv / mu / n / news / o / out / outsoc / p / po / pol / qa / qst / r / r9k / s / s4s / sci / soc / sp / spa / t / tg / toy / trash / trv / tv / u / v / vg / vint / vip / vp / vr / w / wg / wsg / wsr / x / y] [Search | Top | Home]
Please support this website by donating Bitcoins to 16mKtbZiwW52BLkibtCr8jUg2KVUMTxVQ5
If a post contains copyrighted or illegal content, please click on that post's [Report] button and fill out a post removal request
All trademarks and copyrights on this page are owned by their respective parties. Images uploaded are the responsibility of the Poster. Comments are owned by the Poster.
This is a 4chan archive - all of the content originated from that site. This means that 4Archive shows an archive of their content. If you need information for a Poster - contact them.