According to those figures, no not really. 27fps after 4-5 years of the game being out is poor.
AMD: Southern Islands (7*** series) Speculation/ Rumour Thread
- Thread starter UniversalTruth
- Start date
According to those figures, no not really. 27fps after 4-5 years of the game being out is poor.
According to those figures, no not really. 27fps after 4-5 years of the game being out is poor.
Yeah, Crysis can still challenge single cards if you turn all the settings up and 2560, etc (though, I imagine stuff like 8XMSAA is basically useless, broken? all it will do is half your FPS)
Such angst
I know. I'm just an ass wanting real data!
Man from Atlantis
Veteran
Introduction
Tahiti architecture
GNC Compute Unit
level cache Tahiti
Tahiti's ROP, texture units and the memory controller
Tahiti's Tessellator
any general-purpose computing performance improve?
GCN and Partially Resident Textures
Tahiti power control
VCE video coding technology
AMD Steady Video 2.0
Eyefinity 2.0 and DDMA
RADEON HD 7970/7950 and the larger size
PCINLIFE
Man from Atlantis
Veteran
It's probably the main reason but maybe yields and power dont allow them to use all.. heck intel and nvidia are far away from 1.8TF DP so they may keep it for CI ?
I'd be surprised if FirePro variants of Tahiti were not 1:2 DP. 1:4 is most likely just for consumer cards. High enough to beat NV's consumer cards (edit: and Tesla/Quadro as well, actually) with ease, low enough to make FirePro cards more attractive to those who need more DP-throughput.
Introduction
Tahiti architecture
GNC Compute Unit
level cache Tahiti
Tahiti's ROP, texture units and the memory controller
Tahiti's Tessellator
any general-purpose computing performance improve?
GCN and Partially Resident Textures
Tahiti power control
VCE video coding technology
AMD Steady Video 2.0
Eyefinity 2.0 and DDMA
RADEON HD 7970/7950 and the larger size
PCINLIFE
Hmmm… that's a pretty large gap between the 7970 and 7950.
Funny thing, I could mention countless users saying the exact same thing about GeForce drivers
Man from Atlantis
Veteran
15mins left..
D
Deleted member 13524
Guest
Really?
If they keep the clocks, it should be something like a 10% difference. Not that much.
BTW people, please stop reposting those huge slides over and over...
Oh yeah, forgot about clocks. 10-25% slower then...
It's actually slightly smaller than the gap between 6870 and 6850 (28/32=87.5%, 12/14=85.7%).
It's also possible the relative difference in core clockspeed will be smaller in return, I'd expect 850 +/- 25 MHz for the 7950.
Yes, but they could write anything and we'd just have to take their word for it. Or do you think it was 874?
It's four CUs, but I forgot that they have to be grouped by 4, so yes, it's as small as can be. It's just larger than the gap between 6970 and 6950 (4CUs vs 2SIMDs, or 12.5% vs 8.33%).
It may well be a hardware limitation. After all, 1/4 is the cheapest way to do it.
Similar threads
