I was hoping to see a Keplar card by March or April at the latest just so that I had a choice of which would be best for me. I am not a fanboy of any manafacturer but with rivalry comes price wars....this is what I want 
-
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.
Jen Hsun on Kepler (CES 2012)
- Thread starter Havana_UK
- Start date
Caporegime
- Joined
- 18 Oct 2002
- Posts
- 33,188
No one quite knows with the 7870, as said the 7970 "unlocked" for want of a better term IS 80% faster than a 6970 and 50-60% faster than a 580gtx. The 7870 could be a bang on 6970 performance, I think it could end up higher, though again maybe not at stock. The architecture clearly has a lot of clock speed potential, will the VRM's and pcb be built well enough to show that on the 7870, time will tell. It might be things like MSI lightning versions of the 7870 that could be killer.
But while the difference between the 6970/6870 isn't massive, that was due to die size limits, that speed difference could well be similar this gen but due to tdp. THe 7970 is really a 350W card stuck at 225W.
In terms of Nvidia pulling a top end big core out the bag on short notice, its certainly possible, but I think fairly unlikely. The 7990 should, realistically be out a long time before the big Kepler, and likely before the midrange Kepler. Hopefully at least for 7990 buyers it will return to the "old school" dual gpu cards, where the 3870x2/4870x2 were massively better value than the single card. IE £180 for a 512mb 4870 while a 1gb per core 4870x2 was £330 on launch, far below the price of 2x 512mb 4870's.
But while the difference between the 6970/6870 isn't massive, that was due to die size limits, that speed difference could well be similar this gen but due to tdp. THe 7970 is really a 350W card stuck at 225W.
In terms of Nvidia pulling a top end big core out the bag on short notice, its certainly possible, but I think fairly unlikely. The 7990 should, realistically be out a long time before the big Kepler, and likely before the midrange Kepler. Hopefully at least for 7990 buyers it will return to the "old school" dual gpu cards, where the 3870x2/4870x2 were massively better value than the single card. IE £180 for a 512mb 4870 while a 1gb per core 4870x2 was £330 on launch, far below the price of 2x 512mb 4870's.
I don't buy any of this crap about "not showing for business reasons".
If Nvidia had working high-end silicon to show, then I'm fairly sure that they would have shown it - or at the very least "discussed" it in a way that cost sales from AMD.
I still believe that Kepler will significantly out-perform the 7970 when it arrives; even the mid-end part may be able to compete well with it. But, it seems that we will be waiting quite a while for the cards. If we see the mid-high end cards before May / June time then I will be surprised.
If Nvidia had working high-end silicon to show, then I'm fairly sure that they would have shown it - or at the very least "discussed" it in a way that cost sales from AMD.
I still believe that Kepler will significantly out-perform the 7970 when it arrives; even the mid-end part may be able to compete well with it. But, it seems that we will be waiting quite a while for the cards. If we see the mid-high end cards before May / June time then I will be surprised.
Come February and Nvidia starts losing their market share unless they can come up with some good promises or price drops.
They may lose market share in the tiny fraction of the market where the 7970 sits, but the real work of securing market share for the next gen architecture is already done (i.e. nVidia's claimed 300 mobile Kepler design wins out of an approximate 550 sockets on offer).
Soldato
- Joined
- 25 Sep 2009
- Posts
- 10,244
- Location
- Billericay, UK
There's not a dogs chance there going to rework the design for Kepler it's been in development for years so changing it now would mean massive delays and them being at competitive disadvantage for a lot longer.
If I recall Fermi was April 2010 but was due Q3 2009, Kepler from the reports is another large complex chip which I would guess is one of the reasons why it's late (Nvidia did officially say this was due 2011) which points to Nvidia not learning from past mistakes. Meh, I think it's time for Mr Huang to go, most CEO's have a usefulness for about 8 to 10 years and he's long past he's expiry date some fresh blood might be what the doctor ordered for the green team.
Nvidia still has deals with many OEM's so I doubt they will lose much share (which isn't all that important) but I would imagine their bottom line will be squeezed a lot harder. As for the retail side history tends to show that Nvidia always manages to make one decent card that appeals to a lot of consumers (6600GT, 8800GT, GTX460 etc) by offering near top end performance for mid range money.
Last edited:
No, they will lose real market share where 7950s and lower cards will land. Of course I'm not talking about mobile market here.
This, why do people believe this bullcrap that Nvidia would adjust their technology according to AMD's releases. They've had their designs in for years, technology is developed years ahead and future steps are just modular achievements.
The only thing that Nvidia will bear in mind is release timing and polishing up the production process. Gibbo has mentioned weeks ago that GTX 500 cards were no longer in production for a while.
Nothing huge in the grand scheme of things.
How many aftermarket cards are sold compared to OEM PCs? A few percent of the total market at most.
Soldato
- Joined
- 21 Sep 2009
- Posts
- 5,267
- Location
- London
Meanwhile AMD are raking it in as the main OEM supplier to Apple now, where it used to be Nvidia.
ASUS Shows Off Ivy Bridge Notebooks featuring Nvidia Kepler Mobile GPU, Launching in April 2012
http://wccftech.com/asus-shows-ivy-...vidia-kepler-mobile-gpu-launching-april-2012/
So...how "new" is Kepler really going to be?
Seems that the product nomenclature will be GTX 6xx then...
http://wccftech.com/asus-shows-ivy-...vidia-kepler-mobile-gpu-launching-april-2012/
So...how "new" is Kepler really going to be?
Seems that the product nomenclature will be GTX 6xx then...
Soldato
- Joined
- 13 Oct 2004
- Posts
- 2,774
- Location
- Teesside
lol at fatality in the pic, is he not dead yet?
Soldato
- Joined
- 11 Oct 2008
- Posts
- 3,833
- Location
- London
From all the numbers that have come out, the HD4000 will be swinging for Llano levels of IGP performance (about 480 ATI stream processors, roughly 128 CUDA cores for reference).
If the HD2500 is going to be sporting half the EUs of the 4000, that still puts it ahead of the current HD3000, barely.
Considering that kind of graphics power is going to be coming as standard on nearly every model of Ivy Bridge chip, how many budget machines will forego a discrete card and stick with the very capable integrated graphics?
The more I think about it, the more sure I am that, going for the low-end first is Nvidia defending it's market share, rather than chasing the headline numbers.
Last edited:
Nv and AMD are absolutely miles away from Intel on graphics parts shifted...and the gap is increasing slightly, not decreasing. NV has gone all mobile, AMD has gone all integrated, and both these strategies are to sell bucketloads of chips. I really am thinking that these latest two architectures could be the final huge and fast and hot pc gpu releases for a while. AMD looks like it will be into modular designs, and NV into SoC designs, so in future your HD8890 might possibly be 4 chips on one board. Or your GTX890 might be 8 SoC's.
Don't forget that nVidia are chasing after the MPP market as well, so I think it's likely that they'll keep developing bigger and faster products with more and more cores.
Caporegime
- Joined
- 18 Oct 2002
- Posts
- 33,188
8 soc's would waste a HUGE amount of silicon on repeated parts, Nvidia has no reason to do this. Likewise AMD go modular? Every gpu since a 2 pipeline product is already modular............ that is basically what a GPU is and always has been. Right now a 7970, 7870, 7770 are all the same chip but various modular versions of it.
The biggest problem with the 7970 is power, not die size at the moment. Split it into 4 separate chips and you'd still have the same amount of power and be unable to add a 5th or 6th.
The biggest limit, is power, and as electricty gets more expensive people are likely to be less and less willing to increase power by 50% per generation, which isn't surprising.
To be honest, we've only got realistically 3 processes left max before silicon gets so expensive, till the production process takes so many stages and so long to finish that the increase in cost makes it essentially pointless for all but critical applications in very high price sectors. If power goes up 40-50% per gen till then, the end power usage isn't "that" bad for the potential performance and would always still be lower power/performance than previous gen xfire.
But will AMD/Nvidia release a 450W 22nm and then a 600W 14-16nm chip........ I really don't know.
People, especially certain people on these forums will say power efficiency of certain mobile devices will bring massive power savings to the desktop. They won't, mobile parts are magnitudes slower, one game that runs on a pc and has a "mobile" varient isn't the same engine, isn't the same quality. On the rails crap on a tablet with prebaked effects, even if you can prebake the effects well, still loses quality and would be a MASSIVE step backwards in PC graphics. You aren't more power efficient, those chips are simply doing massively less work.
Soldato
- Joined
- 25 Sep 2009
- Posts
- 10,244
- Location
- Billericay, UK
All you doing is quoting numbers. Intel might be able to design a graphics chip as good as a Llano but Intel's software isn't know near as good. If Intel had some mid range video cards they wouldn't sell because their drivers wouldn't be good enough.
they are even more modular than they have been...however, I didn't say that is what would happen. these parts are going to have to get more power efficient, the 7970 is basically clock limited because of power requirement, as are 580's, and cpu's in some respects. The market is diminishing, people are buying less and less high end parts and games are showing this in how they are coded ie. the console ports.
Which way will they go? I don't know.
AMD going more modular means they are going to have the same sort of family architecture completely across the range, right down from 17w apu's to 300 watt monsters. it's not like that now, and as already has been mentioned nv's new mobile parts are fermi+keplar designs. Announcements at CES along the lines of mobile graphics parts with much much better specs mean they will be forced into changing their tack as the iphone 8 or whatever will possibly replace your gaming rig and console. Consoles are already struggling in some respects and being replaced by tablets and the like...it's a natural thing. I feel you're off the mark regarding mobile parts. Yes they are slower, the devs will just move away from console and pc games and start on mobile first, like angry birds et al. You won't even need a fast card to play the games then will you?
Which way will they go? I don't know.
AMD going more modular means they are going to have the same sort of family architecture completely across the range, right down from 17w apu's to 300 watt monsters. it's not like that now, and as already has been mentioned nv's new mobile parts are fermi+keplar designs. Announcements at CES along the lines of mobile graphics parts with much much better specs mean they will be forced into changing their tack as the iphone 8 or whatever will possibly replace your gaming rig and console. Consoles are already struggling in some respects and being replaced by tablets and the like...it's a natural thing. I feel you're off the mark regarding mobile parts. Yes they are slower, the devs will just move away from console and pc games and start on mobile first, like angry birds et al. You won't even need a fast card to play the games then will you?
)