GeForce Kepler 104 (GK104) Packs 256-bit GDDR5 Memory Bus, 225W TDP

[Benetanegia]

Originally Posted by KooKKiK

i didn't see anything in the article that prove your argument.

may be u should "try again"


oh, and you said you didn't claim anything ???

what is this ???




If i had Kepler IN HANDS and benched it right now, i'm sure u gonna make an excuse like "it's only an engineering sample" anyway.

You don't see anything in that article that proves my point? Hahahaha. Nice try, but stop trolling.

My point: Kepler might not be memory bandwidth limited, just as countless of previous cards that AMD and Nvidia surprised us with, that had much less bandwodth than their predecesor. <-- (stating posibilities/probabilities, without stating or asserting how things are going to be only how they may be == no claim)
Proofs: the article, 8800GTX vs 9800GTX, GTX480 vs GTX570, and several cards in the article and many many other cards before and after.

Your claim: Kepler will be memory bandwidth limited. (stating what it will be == claim)
Proof: NONE.
what you think it's "proof": Your GTX580, which is NOT Kepler by any means or stretch of imagination, suffers a 3% penalty when creating an artificial 15-20% gap between stock/balanced GPU clocks and memory clocks. That's it, every 20% less memory BW, degrades performance by 3% on the GTX580, which is not GK104.

I'm still awaiting your proof. The burden of proof in on your side, as always has and you have ZERO proofs so far. Of course you won't have any proof until Kepler is released, but you'll figure it out.

On the positive side, you are a good troll. Mamma troll is probably proud of you.

[KooKKiK]

oh boy, mr. slippery


i proved it in the same way of your first argument ( hi-end gpus have excessive bandwidth and that for computing, and even 6970 BW is enough for 7970 performance bla... bla... )

and then you changed it ( maybe Kepler won't handle bandwidth in the same way as Fermi/7970 )

WTF !!!


your article say it straightly that when everything is about the same, difference in bandwidth mainly affect the overall performance.

( see 3870 vs 4670 and 360M vs 435M comparison )

and 570 vs 480 is not the case coz GTX480 is a partially shader disabled chip, but NOT in the memory part.

and this difference is only about 10 - 15% ( 9800 vs 8800 either ) NOT as much as when you compare 6970 to 7970 BW.


anyway, i've finished this, let the people see and judge by themselves whos right and whos wrong.

[Benetanegia]

Originally Posted by KooKKiK

i proved it in the same way of your first argument ( hi-end gpus have excessive bandwidth and that for computing, and even 6970 BW is enough for 7970 performance bla... bla... )

You proved nothing. 3% of change with 20% of relative memory change is not something to even take into account. 3% is NOT bottleneck. I never said BW does not affect performance at all, I said it does not affect it significantly. Learn to read and notice the subtle differences.

and then you changed it ( maybe Kepler won't handle bandwidth in the same way as Fermi/7970 )

I didn't change anything. It's all part of the same point. Memory bandwidth does not work as you think AT ALL. BW is not a wall against which the GPU hits and stops. BW bottleneck/limitation is an efficiency curve, where low values affect performance a lot and higher and higher values have diminishing returns.

And of course different architectures/chips react differently to BW. Even thinking it's any different than that is stupid. Example right in the article I posted, there's 5 cards with 51.2 GB/s memory bandwidth and all of them have very different performance:

By brands (more similarities in architecture)

GTS 160 M - 3374
9800M GTS - 3700 (+9%)
9800M GTX - 4123 (+22%)

HD4670 - 2552
HD5830 - 4243 (+66%)

your article say it straightly that when everything is about the same, difference in bandwidth mainly affect the overall performance.

When everything is the same... when everything else is the same... of course the only factor that remains (memory bandwidth) affects performance. Right. But as the article also points out, it does not affect anything close to linearly. In fact in many cases it's completely minimal.

So when everything is the same...

Too bad it's never ever the same between 2 different chips, even if they have the same or similar specs, as the article shows. And Kepler will definitely be different to Fermi, while still mantaining a lot of similarities when it comes to overall architecture, but still the chips are going to be very very different, and as such, yes it is posible, no, probable, that Kepler won't be bottlenecked by 256 bit. And we don't even know if it has 256 bit anyway, that's why I never claimed anything as fact and always carefully chose my words. It is probable that Kepler won't be bottlenecked and yes, why not it is also posible that it will be bottlenecked, but if talking about probablities, IF it really is 256 bit, it's far more probable that it won't be affected much, OR IT WOULDN'T RELEASE with 256 bit!! Or do you think they randomly choose specs?? pff

EDIT: http://translate.google.es/translate...bit_in_spielen

HD7970 with HD6970 bandwidth, 15% slower than stock HD7970. Still 20% faster than HD6970. And this is with horrendously high latencies (which is why I said you can't just underclock for comparison, it's not 100% accurate). 1800 Mhz can certainly run with much reduced latencies compared to 2700 Mhz memory.

[Casecutter]

Originally Posted by ZakkWylde

This was on my facebook feed, thought others here might like a look. Some un-based claims of Nvidia dominance.

http://www.maximumpc.com/article/new...i_architecture

From the Maximum PC post...
"even Nvidia's mid-range cards will give AMD's high-end GPUs a run for their money"
"claiming Nvidia's mid-range cards will have the moxie to challenge AMD's higher end GPUs"

Those I can correspond, but this might be a little strong... "Nvidia is going to "win this round on just about every metric" with its Kepler architecture, which will trump AMD's Tahiti "handily."

Read what I'm thinking we'll see... when the catchphrase was, "NVIDIA is gunning for the performance crown from AMD Radeon HD 7900 series with this chip"
http://www.techpowerup.com/forums/sh...06#post2521506

[crazyeyesreaper]

all those articles are just reposts with edits of Semi Accurate article written by Charlie, id wait for real numbers before worrying about anything.

[phanbuey]

wow... the Charlie I had read before literally would start frothing at the keyboard at the mere mention of nvidia.

One of the comments suggested that he was fishing for a content thief... probably true lol.