AMD RV770 refresh -> RV790

Interesting. Both the Dutch and Australian sites use the same Sapphire manufacturing number: 21150-00-40R. Australian even mentions the part number: VGASAP4891. I doubt that they're using the same database or that it's coincidence that they're using the exact same numbers....
 
4890

Cheers, I'm a long time lurker on this message board and today I decided to sign up because of the listing of HD4890 @ Salland Automatisering. ;)

I believe the RV790 is in fact a 850Mhz part with 960SP's and 48TMU's. Assuming perfect scaling this combination would be 36% faster than RV770 and that would fit very well with the rumored 20-30% faster than RV770.

An up-clocked RV770 @ 950Mhz would only be 26% faster assuming perfect scaling and that would give us something like 15-20% faster than RV770. Considering the thermal issues at such a high clock speed, it just doesn't sound feasible to me. Not to mention it took AMD/ATI 10 months to get this part out.

Please take my ramblings with a truck-load of salt. Everything I say is just based on rumors anyway so anything can happen. Buthis is not the first time AMD has been sand-bagging. Remember the 480SP's rumored for RV770 when it in fact had 800. :smile:

/Kef
 
Info on the retailer sites seems correct. Sapphire will be shipping cards at the end of this week.
 
Kef said:
Cheers, I'm a long time lurker on this message board and today I decided to sign up because of the listing of HD4890 @ Salland Automatisering. ;)

I believe the RV790 is in fact a 850Mhz part with 960SP's and 48TMU's. Assuming perfect scaling this combination would be 36% faster than RV770 and that would fit very well with the rumored 20-30% faster than RV770.

An up-clocked RV770 @ 950Mhz would only be 26% faster assuming perfect scaling and that would give us something like 15-20% faster than RV770. Considering the thermal issues at such a high clock speed, it just doesn't sound feasible to me. Not to mention it took AMD/ATI 10 months to get this part out.

Please take my ramblings with a truck-load of salt. Everything I say is just based on rumors anyway so anything can happen. Buthis is not the first time AMD has been sand-bagging. Remember the 480SP's rumored for RV770 when it in fact had 800. :smile:

/Kef
Click to expand...

Yes I remember, Jawed even was convinced it would only have 16 TMU's. 480 SP's was also the max speculated (or was it 600-800 with some dedicated to physics?). If you'd told us back then it would sport 48 TMU's AND 800 SP's. minds would have been blown :LOL:

Of course, it didn't turn out all that amazingly fast as the stats sound, although fast enough to give Nvidia headaches.
 
Last edited by a moderator:
Kef said:
Please take my ramblings with a truck-load of salt. Everything I say is just based on rumors anyway so anything can happen. Buthis is not the first time AMD has been sand-bagging. Remember the 480SP's rumored for RV770 when it in fact had 800. :smile:

/Kef
Click to expand...

RV790 was also up to recently supposed to be at 40nm. I guess it takes one step at a time to realize a couple of things.

RV790 is not just a higher clocked RV770 that one I consider almost a certainty at this point, meaning your thermal considerations are in vain to start with.

Don't let me spoil the party; some folks haven't had a chance to even get used to their signatures yet *cough*
 
Just for kicks if the 4890 is just an upclocked 4870 at 850 mhz, (13% faster) and you apply it to this Firing Squad GTX 285 review with ideal scaling (scaling all the 4870 1GB benchmarks + 13%) you'd get these numbers vs stock GTX 285 (with whatever X AA they used, 8X in most cases, and at 1920X1200)

------------- 285 -- 4890
Fallout 3 60.7 57.6
COD4 113.1 91.1
Crysis High 34.9 34.5
Left 4 Dead 77.3 83.5
Clear Sky 28.3 24.4
Far Cry 2 53.5 51.1
Bioshock 80.1 75.9
WiC 50 44.1


Not too shabby, better than I initially thought. The 4890 manages to win one game and be within close spitting distance in all others except COD4. That's not considering all the overclocked 285 variants though. Still, it shows ATI doesnt have to do much else besides this 13% clock bump to be competitive with the 285. In fact given the rumored 4890 price points, I'd be a bit surprised if they did anything else. Anything much more (either 960 SP's or OC'ed to 950 variants) and it will equal or exceed the 285, meaning they would also price it accordingly.

If there's one thing I've learned is foolproof in these speculation games, it's price point. If the 4890 is to be 299 or less, thats a strong hint it's nearly as good as a 285 (349) and nothing more.

If it is just a few percent slower than 285, 299 price trending down towards 249 would then be highly attractive. So it would all make sense.
 
Last edited by a moderator:
Hmm, it has 2x 6-pin power connector. Not 8-pin + 6-pin, as some sources implied. Only the PCB is 8-pin ready.
 
"HD4895" for the 950MHz version?

HD4850/70 had sensible names - not sure what's going on here though.

Jawed
 
4890...
powercolor4890.jpg
 
Uh oh, new pin layout around GPU probably confirms that this is indeed a new layout / shader config

Edit:

Just noticed that RV770 has them in 6 x 12 k the back, whereas RV790 shows them at 4? x 11?

Let the speculation begin!
 
You must log in or register to reply here.
Back
Top