NVIDIA Kepler speculation thread

I think he was referring to influencing Apple's decision making.

Past problems that are known to be corrected probably carry much less weight than ongoing problems.
 
Last edited by a moderator:
Kaotik said:
Software can be fixed, and it's not like nVidia drivers are without fault. However, software can't fix chips breaking what happened before they moved to ATI/AMD
And let's not forget the 3DMark 03 cheating!
 
No one sane would use inferior AMD products after being burnt like that. And the fact Kepler swept up almost all of the Ivy Bridge notebook design wins is quite telling.
The article is taking quotes from 2009, not at all relevant to the products that were (and still are) shipping from the last cycle.
 
I understand it's not a very titillating explanation, but what if Apple simply chose its components based on the technical merits of new generation? You know: forward looking instead of backward? For a company that puts battery life front and center, Fermi was not a very compelling product line. We don't know enough power numbers for meaning comparisons (especially since they are using GDDR5!), but with tons of others having chosen Nvidia for their notebooks as well, it's fair to say that their perf/W must be competitive against AMD.
 
We don't know enough power numbers for meaning comparisons (especially since they are using GDDR5!), but with tons of others having chosen Nvidia for their notebooks as well, it's fair to say that their perf/W must be competitive against AMD.

Thus far GK107 is not exactly covering itself in glory in that area - in this case peak TDP difference lower than the performance differential and an idle power that is higher despite a smaller die size.
 
That is a DDR3 version vs GDDR5 cards. The 7750 has 2.5x the bandwidth of the 640m there; surely a more accurate assessment could be made of the chip itseslf with GDDR5.
 
DDR3 will also be significantly lower power draw than GDDR5, and probably have more in common than most of the notebook implementations. The idle power is a very important factor for notebooks that are driving the display from the discrete graphics as well.
 
Dave Baumann said:
DDR3 will also be significantly lower power draw than GDDR5
Oh please. Less yes, relative to the potential performance gain from having 2.5x the bandwidth... not much.

The GT640m DDR3 isn't the only GK107 card. There is also the GT650m (which I believe Apple chose). Perhaps it would be prudent to wait for some benchmarks of the GT650m before condemning GK107 based solely on a single implementation.
 
Last edited by a moderator:
http://micgadget.com/21980/apple-is-close-to-finally-updating-the-mac-pro/



No one sane would use inferior AMD products after being burnt like that. And the fact Kepler swept up almost all of the Ivy Bridge notebook design wins is quite telling.


I will be honest i have no idea from where thoses things are coming, i m offtly on a french site who turn to be with a lot of Mac users..and half is a side mac site, I like how the guy in this article cite problems who date from 2009, when in reality im not even sure this was AMD related cause:

- monitors problem, it was impossible to say if it was from AMD as most monitors was just broken.
- Hard drive failure ( non amd related )
- overheating ? its not amd who is responsible of the design of the Imac of Apple.
- Monitor with side broken at delivery or the screen was damaged...
- the list is long.
- I dont even speak about the flash problem who was there with Apple since many years, causing the CPU to work at 100% usage. ( no acceleration on gpu ). Due to the faulty support of flash with Apple.

In fact the famous 27" Imac was just completely faulty at start... i have never read anywhere this was due to AMD gpu ... The most funny is i have never read anywhere those problem have occur with AMD HD4000 series at this time ... nor even with HD6970M used on new Imac, macbook ...

Most funny enough, is if Apple had change his contract with nvidia was coming from problem who look similar at thoses describe with previous Nvidia hardware on Apple laptop ( including the overheating problem and gpu dying .
 
Last edited by a moderator:
Oh please. Less yes, relative to the potential performance gain from having 2.5x the bandwidth... not much.
Well at idle the power draw savings are much more significant than the performance decrease :).
I dunno though if idle power draw is still that important now that pretty much every notebook design has some sort of optimus or whatever nvidia, amd, apple call it (even the high-end chips this round...).

The GT640m DDR3 isn't the only GK107 card. There is also the GT650m (which I believe Apple chose). Perhaps it would be prudent to wait for some benchmarks of the GT650m before condemning GK107 based solely on a single implementation.
There's quite some benchmarks on notebookcheck already for the gt650m. Some with ddr3, some with gddr5 memory (though I think in some benchmarks they've got the wrong memory listed, I have no idea if apple uses gddr5 or ddr3), and quite surprisingly imho the practical performance difference isn't THAT big - something like 10-25% (which is of course more than the difference between ddr3 gt640m and ddr3 gt650m but still less than I expected). Maybe the ddr3 version clocks higher because the ddr3 memory uses less power or something.
 
The article is taking quotes from 2009, not at all relevant to the products that were (and still are) shipping from the last cycle.
I am sorry , but it's quite obvious that NVIDIA enjoys an advantage with their software suite , even their software AA (FXAA) works better and smoother and with less bugs than AMD's software one (MLAA).
 
Thus far GK107 is not exactly covering itself in glory in that area - in this case peak TDP difference lower than the performance differential and an idle power that is higher despite a smaller die size.

Does THG explain somewhere how especially their load numbers are obtained? Wall power, after all, increases also with higher bus, memory and CPU load which in turn can be caused by a faster GPU that is not stalling the rest of the system as much.
 
I am sorry , but it's quite obvious that NVIDIA enjoys an advantage with their software suite , even their software AA (FXAA) works better and smoother and with less bugs than AMD's software one (MLAA).
For one, where on earth does this stem from? Second, exactly what relevance to Apple does this have?
 
Does THG explain somewhere how especially their load numbers are obtained? Wall power, after all, increases also with higher bus, memory and CPU load which in turn can be caused by a faster GPU that is not stalling the rest of the system as much.
That favours the power numbers of Verde more, as too would if GK107 is following GK104 it is limited to PCIe Gen 2 on X79 where Verde would be running Gen 3.
 
Thus far GK107 is not exactly covering itself in glory in that area - in this case peak TDP difference lower than the performance differential and an idle power that is higher despite a smaller die size.
If we only look at 16x10 and 19x10 resolutions, a GT640 is 24% behind a 7750 in perf/W. (Assuming all tests run at the indicated power.) That's doesn't seem to be an insurmountable gap by switching to GDDR5, considering the large amount of ROP and shaders and pathetic bandwidth.

It's not as if one has to beat the other into submission. Competitive is sufficient. Difference in brand reputation, perceived driver quality, and pricing (smaller die, as you pointed out) can do the rest.

Edit: FWIW, the only straight comparison I could find between DDR3 and GDDR5 for otherwise identical cards, is this one. If you average the performance of gain for all 16x10 results, you get a speedup of 24.7%. (For this particular case, power consumption was surprisingly identical, but you need many samples to come to a firm conclusion on that one. That's true also for those THG results, of course.) It's a different architecture with different unit ratios, but it shows that a 25% improvement is not some number from lala land and it's better than pretending there is no difference.
 
Last edited by a moderator:
That favours the power numbers of Verde more, as too would if GK107 is following GK104 it is limited to PCIe Gen 2 on X79 where Verde would be running Gen 3.

Probably - and I did not follow the agenda of making your argument invalid. :) I just genuinely wanted to know if THG details it somewhere.
 
Heh. I think by "hard evidence" he meant "mushy, unsubstantiated drivel".

You are so lucky. Only 24 hours ago there was 0 in stock at newegg, now you can choose between 4 of them. Suddenly the demand dropped so much that you have something in stock. :LOL:

Actually, the reaity shows something like: you have few days with non availability at newegg and only few cards available at hardwarevertsand, at the next period of few days things reverse- you have some models in newegg and nothing at hardwareversand. :mrgreen:
I begin to forget when its launch was, so long ago. :LOL:
 
Back
Top