This. Just like with the 7970 fake stuff, can people actually apply some common sense please.
-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
Nvidia Kepler vs 500 series specs released - **HOLY CRAP**
- Thread starter DamoTourettes
- Start date
This. Just like with the 7970 fake stuff, can people actually apply some common sense please.
Associate
- Joined
- 19 Dec 2011
- Posts
- 114
what are the differences between the crappy graphs that came out for the 7970 and the actual results posted now? I remember seeing some stupidly exaggerated graphs.
There always are... For the last few GPU releases there have been a wave of obviously faked charts "leaked" all over the place - the vast majority of them completely and utterly fake.
Once we get some specs (HOPEFULLY this month), then we should have a rough idea of performance. Beyond that we'll have to wait until cards start showing up in the wild before we get any *real* info (so, a couple of weeks before release).
Associate
- Joined
- 6 May 2011
- Posts
- 248
- Location
- London
They'll likely reserve the 6xx naming scheme for the low end cardS (GT620 etc).
i like how people arnt noticing the % on the side, they seem to think the card will be 2x as fast just because the bar is twice as big
Where is this % on the side? Are we looking a the same (fake) graph?
Where is this % on the side? Are we looking a the same (fake) graph?
Exactly, it could be the amount of cornflakes the fan will shred whilst playing Crysis 2 in a timed run of 60 seconds for all we know
oh those are fps, ithought they were percent, and i meant op
Presumably they're fps, who knows? There's nothing on the graph that actually states what the numbers stand for. I'm still putting money on my Cornflake theory posted above, or perhaps Frosties?
I really don't think AMD's the reason. Note that this series started with a process /architecture jump named the GTX 200 Series. 100 series was reserved for OEM and/or mobile parts iirc.
Then they skipped 300 which was again OEM and mobile parts and named their retail line GTX 400 series. GTX 500 was a tweak of 400 so no big jump there. (Same process and architecture).
But again GTX 700 series will be a new architecture on a new process, and 600 will be OEM/mobile parts.
So a pattern's starting to emerge. It may not be precise or strictly consistent, but there seems to be some order to (at) the (edge of) chaos. The one thing that sorta disturbs this pattern is when they then did refreshes of GTX 280 on a new process and called it GTX 285. And it's only the 280 that became the 285. The 275 had no counterpart and neither did the 295. And there was no GTX 265.
It seems to me to have been the theme from the beginning, with a bit of faltering along the way. Betweennew architectures they have always stuck to the jump of 200 for each cycle. Within these architecture jumps they've been inconsistent. (So while short-term predictability seems precluded, there is some long-term predictability
).If they wanted to have a higher number than AMD they already had that when they were running the 8000GTx series alongside AMD's Radeon 3000 series.
Last edited:
I think it's about time they stopped using bloody numbers, let's break the mould! What about calling them, I dunno, something like Gaylord Focker for the Epic, monster top end card and work down from there? All in agreement shout "aye"
aye!
Caporegime
- Joined
- 18 Oct 2002
- Posts
- 33,188
It's really actually looking more and more likely that the 7970 is really exactly as good as it should be, just heavily underclocked due to power constraint.
The most interesting thing really about the +70-80% performance per "real" gen, is that power usage has roughly speaking gone up at about half that rate.
Someone did a nice list on semiaccurate.
So basically speaking over the past 3-4 big generational jumps we've seen essentially power increase relatively to performance increase, IE donkeys years ago power increased by roughly 1/3 of performance, the next big gen power increased by roughly half of performance, and last time power increased by a bit more than half of performance.
THe only thing you could say for almost certain is that to maintain the 70-80% performance increase we would likely have required, bare minimum, a 35% power increase also.
That is likely why Kepler will either blow through the 300W single gpu barrier, or be heavily clock limited.
AMD has, in my mind, female genitalia'd out of this fight by releasing a 7970 at 20-30% lower clocks than it could apparently easily have had, to keep power at similar level to last gen while a "real" 7970 could likely have used 300W in gaming and potentially more than that in furmark/gpgpu stuff.
From what I can tell almost every reviewer could overclock their 7970 to the max possible with default voltage, which is 1125Mhz, and with a new bios and a mild voltage bump everyone seems to max out at 1250Mhz, and that is on the stock cooler AND with decent temps still, that is over 30% faster, and memory is overclocking well also.
I think the only "disappointment" in the 7970, is that AMD have tried to force it to be a 200-225W card in normal usage, when in reality as per every previous gen, it needed a likely 40-50% power bump to get its 70-80% faster.
With 30% higher clocks it would quite easily thrash both the 580gtx and destroy the 6970.
But yes, ultimately the problem is how do we have 40-50% power usage increase every single new gen. Years ago it meant going from 40 to 60W, to 90W, to 150W, but that scaling is now getting a bit out of control. Idle power usage is improving dramatically and I think we're well before the point where running a single gpu with 350-400W is somehow dangerous or expensive when it can use sub 20W at idle.
But again the big question for kepler will likely be more a case of, will it be big and neutered to under 300W. Because for Nvidia, well Nvidia tdp is 250W, and in stuff like furmark both could go basically to 300W and a bit over, a 40-50% power increase would be even more problematic.
IS that one of the reasons we're seeing rumours of hotclocks going away, one of the fundamental reasons for higher leakage on Nvidia cards than AMD, quite possibly.
This is what becomes a reasonable power usage, someone with 2x 6970 now using rougly 400-500W, rest of the system and uses a 800-900W PSU, a 7970 single could probably get close "unhooked" power wise, but to double his performance you'd be talking about two cards using potentially 800-900W overclocked on their own.
I think the big kepler could be either, disappointing, or as default use way more than 300W, the ultimate "who did better this gen" will likely only be fairly and really seen with both cards overclocked on air on the stock cooler and see where the cards might be without power becoming the limiting factor.