There's no block available anyway so that pretty much makes the decision for me. I might have a look and see if I can squeeze in another 7970.
-
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.
AMD working on Malta dual-chip card
- Thread starter Final8y
- Start date
Caporegime
- Joined
- 18 Oct 2002
- Posts
- 37,804
- Location
- block 16, cell 12
Agreed they've put a huge effort into improving their driver sector and if they keep it up I'll move back to team red
They've had 20 years to improve their drivers what makes you think they will now?
Don't turn this into a troll thread please...
AMD's drivers have issues, but nowhere as bad as people make out. If you have genuine driver issues instead of a trollollol post then please mention them...
They've already started, they stated last year that they would be putting a lot more effort into their driver team. As above though lets just stop this now as it will escalate.
This or:
Have you two not seen my build thread? A hacksaw and a Dremel are nothing compared to the mills and lathes I usually use!
I believe i said
"Thing will get interesting if AMD release the 13.3 drivers and pull up alongside titan performance. Its within the realms of possibility looking at the benchmarks posted so far."
Soldato
- Joined
- 21 Sep 2009
- Posts
- 5,267
- Location
- London
Kepler was just a progression from the previous gen whereas GCN was a whole new architecture, hence why the Keplers were better optimised from the start and the GCN cards are improving with each driver.
I don't think its very realistic other than some application specific opptimisation the cards are now performing about inline with an extrapolation of their on paper capabilities - which leads to a potentially interesting point because Titan is performing considerably below its on paper potential - whether thats just that the drivers aren't upto speed, nVidia deliberately holding back performance to see what AMD do/other reason or that the on paper capabilities are inflated by the compute potential but it can't be brought to gaming useage I don't know, could even be due to having issues managing the thermal/power envelope.
I think AMD have potentially played their cards wrong here tho - if everything I've heard is true the "8970" would be performing in the region of 35-40% faster than the 7970 and they could afford to sell at around £500 a considerable saving over Titan. Pushing it back may have been a kneejerk reaction to when it looked like Titan was going to be around 60% faster than the GTX680 on average.
Titan paper capabilities.
GTX 680 = 1536 Cores
Titan = 2688 Cores (+75%)
Cores scaling estimated 80% = +65% performance
Add to that the cores are not 100% of the performance of a GPU, Guesstimating an actual performance gain over the GTX 680 of 50% clock for clock. about what it turned out to be.
Thats just my Guesstimation doing a few sums in my head, which i would accept could be way out.
But IMO, i think a lot of those paper capability estimates that were floating around were way out.
GTX 680 = 1536 Cores
Titan = 2688 Cores (+75%)
Cores scaling estimated 80% = +65% performance
Add to that the cores are not 100% of the performance of a GPU, Guesstimating an actual performance gain over the GTX 680 of 50% clock for clock. about what it turned out to be.
Thats just my Guesstimation doing a few sums in my head, which i would accept could be way out.
But IMO, i think a lot of those paper capability estimates that were floating around were way out.
Soldato
- Joined
- 1 Jun 2010
- Posts
- 7,058
- Location
- Bedford
I can't remember the wording correctly, but when I had a 4870X2, there was talk in video reviews of a future driver update allowing the gpu's to communicate more efficiently (sideport?), though it never transpired. I'm assuming this hasn't changed since, and I think it would be very interesting if something like that was put into practice from release day.
7970X2?
7970X2?
Imagine a HD 7770 on crack.
Sideport was a bit mis-understood, it kind of worked like (tho quite different) to the command shortcut abilities of the nVidia PCI-e splitter (NF200, etc.) the realworld gains are in the region of about 5%. I've actually forgotten the exact details on it now but I did a longer post awhile back on it.
