The gap shrinks some if you include a 1.5GB 7950, and then there's a rumor of a Tahiti LE which I guess would weigh in with 1536 shaders and around that $349 range.
AMD: Southern Islands (7*** series) Speculation/ Rumour Thread
- Thread starter UniversalTruth
- Start date
The gap shrinks some if you include a 1.5GB 7950, and then there's a rumor of a Tahiti LE which I guess would weigh in with 1536 shaders and around that $349 range.
AMD be lame? How about you be realistic - of course they charge as much as they can when there's no competition in sight, any sensible company does that when there's no superior motive to do something else (like HD5 & DX11 marketshare)
Silent_Buddha
Legend
Not sure how much wiggle room there is in the performance/midrange part of the market. We already know that Enthusiasts will pay just about whatever a company wants them to pay (AMD and Intel in the past with 1k+ USD CPUS for example).
But when you get to performance/midrange. The price side of the price/perf ratio starts to become more and more important. Is that market going to be willing to jump over the 300 USD barrier? Or will they just wait until whatever card filters down to their price range? I have a feeling it's going to be the latter and whichever company goes above 300 for the performance/midrange is going to have a tough sell.
On the other hand you can try to get some budget minded enthusiasts with a 7930 in that price range. Assuming there's enough supply of Tahiti chips that can't cut it as 7950 or 7970.
Regards,
SB
But when you get to performance/midrange. The price side of the price/perf ratio starts to become more and more important. Is that market going to be willing to jump over the 300 USD barrier? Or will they just wait until whatever card filters down to their price range? I have a feeling it's going to be the latter and whichever company goes above 300 for the performance/midrange is going to have a tough sell.
On the other hand you can try to get some budget minded enthusiasts with a 7930 in that price range. Assuming there's enough supply of Tahiti chips that can't cut it as 7950 or 7970.
Regards,
SB
itsmydamnation
Veteran
i found this interesting:
http://hashcat.net/oclhashcat-lite/
seems that scaling is a little bit better then just raw flops but then there are one or two things ( like des ) where its way faster.
http://hashcat.net/oclhashcat-lite/
seems that scaling is a little bit better then just raw flops but then there are one or two things ( like des ) where its way faster.
itsmydamnation
Veteran
yep i keep telling/joking with the defense net ops guys i work with im going to take my optical sniffer and brute force there traffic. Funny thing its starting to get closer to reality. the recent updates to crypto standards have been about fixing key exchange etc, so there hasn't been a lot of movement to protected against brute force style attacks.
According to Ailuros at 3dcenter forum Cape Verde only has 10 CUs. That's not a whole lot and gives it raw flops rate below that of HD5770 (though higher texturing rate due to higher clock). It is also less than a third of Tahiti which sounds a bit shabby (I was hoping more for something like 12 CUs, a bit more than a third, this is a much larger difference than from Juniper to Cypress or Cayman). But since it is said to be below 150mm² I guess we'll see how efficient this architecture really is (could have similar transistor count as Barts though of course needs to live with half the bandwidth).
With 10 and 8 CUs respectively at least I get my more than 1 CU per group
Though I'm actually no longer convinced the CUs will really be organized like that.
With 10 and 8 CUs respectively at least I get my more than 1 CU per group
Though I'm actually no longer convinced the CUs will really be organized like that.
I am under the impression that the cache-sharing for the Scalar Units between CUs is an option (maybe for increased efficiency), not a must - and probably this collaboration is facilitated on a per-case basis not requiring fixed wiring (GDS-employment case?).With 10 and 8 CUs respectively at least I get my more than 1 CU per group
So, basically, it all depends on how AMD decides to lay down their GCNs. Only thing that really seems like a given would be an equal amount of GCNs for each Raster-Engine.
We should be making the comparison to Pitcairn not Cape Verde. Remember AMD has since demoted all its model numbers down a notch to make room for x970 at the top. Pitcairn should be closer to where Juniper was positioned.
http://www.xbitlabs.com/news/graphi...adeon_HD_7000_Incoming_in_February_March.html
According to that slide It's 2/15 for Cape Verde. Pitcairn in March.
caveman-jim
Regular
If Cape Verde comes out we'll probably quite a good hand on how and what pitcairn will be.
And speaking of that...
I wonder if anybody is brave enough to gulp a few of those sugar pills
I wonder if anyone is actually brave enough to put those pills in the fan intake
DarthShader
Regular
It gets funny when one realizes "verde" means "green" in spanish.
And the backside is of the package is great too, look here:http://www.computerbase.de/news/2012-02/gruener-troll-pillen-von-amd-mit-1-ghz/
http://verdetrol.com/
Silent_Buddha
Legend
Bummer, only 7 days worth of benchmarking pills?
Regards,
SB
Regards,
SB
itsmydamnation
Veteran
not bad... the Fermi add was better
Similar threads
