Kepler postponed 1 year

T

Trackr

[H]ard|Gawd
Joined
Feb 10, 2011
Messages
1,786
According to this news post, Kepler won't be coming this quarter, but rather Q3/Q4 2012 and into 2013.

133a.jpg


Boy, this is a real slap in the face to everyone (including me) who's been waiting for GTX 680 since a year ago.
 
That would suck, I need to upgrade my ageing GTX 295.

Do nVidia seriously expect users to continue buying their GTX 580 as the fastest single GPU card for another year? This is really weird.
 
That chart is already wrong. It shows the 550ti released at the beginning of 2011 but it came out in March. It also shows the 520 at the beginning of 2011 and it came out in April.
 
I think the actual release date for the stuff being replaced is irrelevant in this chart?
 
gk112 is a single gpu card on the top, slotted with the 590
gk110 is a dual gpu card slotted with a 580

that just seems odd that the top card isnt a dual gpu card. usually the top card is a dual card. the dual card listed is like a 560x2. they is not 680x2 (690) on the road map.
 
Hopefully wrong. I'll be itching for something new by summer.
 
They are in clear violation of Moore's law and should be arrested.
 
According to that the 28nm uses GDDR5 which is wrong as it going to use XDR2. So thats as old as the mountains "I presume and wont bother about that
 
Spooony said:
According to that the 28nm uses GDDR5 which is wrong as it going to use XDR2. So thats as old as the mountains "I presume and wont bother about that
Click to expand...

Where did you hear that? I heard amd is using xdr2, but I haven't heard anywhere nvidia was. Xdr2 on a 512 bit bus would give you about 500gb of memory bandwidth.
 
Any fool with MSPaint can draw a picture. Is this what people consider "news" these days?
 
Well, this is almost enough of an excuse for me to go out and buy one of those 3 GB 580s with a waterblock...
 
While it doesn't specifically say what information they are working off of, the story at TechPowerUp *does* say that it's what *could* come out and when it *could* be released.

They don't do a very good job of saying it, but they pretty much say it's conjecture. In my mind it's a foolish thing to do since people are going to take it and run with it. Rather irresponsible on their part, particularly with how badly the table is designed.
 
For me the later the better as hopefully Ivy Bridge, Windows 8 and the 680 or whatever it will be called will come out about the same time and I can build a new rig with the latest and greatest all at once.
 
I doubt that... if AMD releases their cards at beginning of 2012, Nvidia will have to follow. I doubt 1 year of having AMD sellin top cards without any oposition is nvidia's dream scenario.
 
The CEO just confirmed they are coming in the last stock call. This is a fake slide. These cards will hit Q1 for sure.

My forum famous crystal ball confirms it.
 
tricky0502 said:
Where did you hear that? I heard amd is using xdr2, but I haven't heard anywhere nvidia was. Xdr2 on a 512 bit bus would give you about 500gb of memory bandwidth.
Click to expand...
Now where did you get that info from? nvidia uses a bigger bus than Amd yet their memory performance and the connection to the ram sucks. Their better shader engine and design make up for that though. So a huge bus doesnt give you huge bandwidth. You still need to connect the thing to the engine and Nvidia is lacking in that department
 
the leaked specs for the 7970 with xdr2 memory are 8000 mhz for the memory on a 256 bit bus (256GB/s of bandwidth). So if the memory on kepler ran at 8000 mhz (same as amd) and was on a 512 bit bus it would have 512GB/s. a little bit of over kill?



amd_hd7900_series_specifications.jpg


Spooony said:
Now where did you get that info from? nvidia uses a bigger bus than Amd yet their memory performance and the connection to the ram sucks. Their better shader engine and design make up for that though. So a huge bus doesnt give you huge bandwidth. You still need to connect the thing to the engine and Nvidia is lacking in that department
Click to expand...

amd uses faster ram and a slower bus speed, nvidia uses slower ram speed on a faster bus. the memory bandwidth is about equal (slight edge to nvidia in the current gen cards). memory bandwidth has nothing to do with shader performance.
 
I hope that this time NVIDIA will release their cards on time, unlike when GTX 480 was released 6 months late after ATI released 58xx.
 
Something that baffles me is the absence of GK100. Back with Fermi, the first flagship Fermi GPU was codenamed GF100, and it was the subsequent revision GTX 580 that was codenamed GF110

Something tells me that this GK112 may be a revision of the first gen Kepler, with more stream processors once TSMC's 28nm HP has matured. I think we may very well still get the GK100 GPU next year, with a revision in the following year.
 
Why would an Nvidia employee put "nvidia desktop graphics" on a presentation? Wouldn't they just simply put "desktop graphics chip roadmap" as it's obvious who the person giving the presentation would be talking about?
 
I REALLY hope this isn't true, I was hoping to get a couple Keplers early 2012. I'm sick of AMDs driver bullshit.
 
tricky0502 said:
amd uses faster ram and a slower bus speed, nvidia uses slower ram speed on a faster bus. the memory bandwidth is about equal (slight edge to nvidia in the current gen cards). memory bandwidth has nothing to do with shader performance.
Click to expand...

It's not a faster bus, it's a wider bus. Same effect though.
 
I don't believe that chart. The fast parts are first which squeeze the most money from impatient buyers, and then the cheaper/slower parts trickle out top to bottom.
 
Spooony said:
According to that the 28nm uses GDDR5 which is wrong as it going to use XDR2. So thats as old as the mountains "I presume and wont bother about that
Click to expand...

Nvidia NEVER uses new technology before/same time as AMD/ATI.. they always use AMD as the guinea pig so theres no way in hell Nvidia is going to use XDR2 until AMD tests and releases it to the mass market. plus i'm pretty sure AMD has at least a couple years of exclusive rights to XDR2 otherwise they are idiots.
 
Fakes. Besides they wont be late on this, they're currently able to blame TSMC for delays with the new 28nm process, it would shame for them to still not be ready when TSMC can deliver.
 
Last edited:
DeathPrincess said:
Why would an Nvidia employee put "nvidia desktop graphics" on a presentation? Wouldn't they just simply put "desktop graphics chip roadmap" as it's obvious who the person giving the presentation would be talking about?
Click to expand...

Have you ever worked in a large corporation or institution? Most I've been associated with like to see their name and logo on everything and in every sentence of any text or speech associated with them. So that per se isn't a reason to doubt it.
 
Gah, I spent the last 2-3yrs dealing with ATI drivers...plz don't make me have to do it again /sob.
 
Supposedly Kepler is going to be 3+ screen support on single card. If thats true I am definitely 680 bound. If not, I am going to be fighting ati drivers with you inglewood.
 
You must log in or register to reply here.
Back
Top