Sound_Card
Regular
Yes but has not the meaning of RV changed with RV being single and R being dual?
AMD: R7xx Speculation
- Thread starter Unknown Soldier
- Start date
- Status
Yes but has not the meaning of RV changed with RV being single and R being dual?
I must have missed the official announcement on that one.
Still, I'm having a hard time fitting V and single together over V and Value.But for competitions sake I sure do hope I am wrong this time around.
Sound_Card
Regular
I'm not saying RV is high end, obviously that would be the dual GPU R.
$349 and $269 does not scream high end to me. If that chart is to be correct.
Sound_Card
Regular
I must have missed the official announcement on that one.
Just a trend.
Also Richard H.'s speech.
Last edited by a moderator:
You lost me. There's no "R" at the moment...there's only RV. Stick two RV's on a board and you get an X2. What's the "R" you're referring to?
Sound_Card
Regular
R700?
3870X2=R680 in ATi nomenclature. He's extrapolating that 4870X2=R700.
Yeah I got that but I don't think it's enough to deduce that RV will always be single chip or that all future high-end cards will be multi-chip. For example...what's stopping future midrange cards from being multi-chip designs? I think the naming has a lot more to do with market positioning than with the number of chips on the board.
Yeah, sure, I wasn't speculating that it is or that it is not.
aaronspink
Veteran
No idea - 32 is so boring though!
[note: this is a caricature of R7xx Rumours. Stop believing everything you read and randomly combining unreliable rumours please?
I say we only allow prime multiples of primes in the future!
aaronspink
Veteran
>< a prime multiple of a prime
Any new ETA on release date?
it was released yesterday, didn't you see all the coverage? And in following the trend set by the 9800GTX, the MSRP is actually LOWER than the 9800GTX!
Seriously, Q3/Q4 are most likely for general availability.
Fast14?
But the die size is possibly the least reliable part of all current rumours; why are you and so many others taking that as a starting point?
Anyway, personally, I'm expecting 40 TMUs...
Exactly, we dont really know anything about the chip until it's released or a week or two before maybe as usual. All these transistor counts, die shots, etc indicating a possible small chip are as likely to be wrong as any of the other floated stats. It could be a very big chip, honestly.
Berek
Regular
I have a hard time believing it (if the performance is to be as great as expected), but this French site reports full technical specs and suggested pricing for the RV770 series, which peg the RV770XT at $349. Is this not a more higher "mid-range" price?
http://www.hardware-infos.com/news.php?news=2008
Although this doesn't suggest any X2 variants, which would probably be their "high-end" card at $599.
http://www.hardware-infos.com/news.php?news=2008
Although this doesn't suggest any X2 variants, which would probably be their "high-end" card at $599
.
I sincerely hope, they get RV770 out the door much sooner than that.
Berek
Regular
May... next month, and they are doing a hard launch, so availability will also likely be in May. We cannot say on how "tight" the supply might be, however.
So.... what do you think of Fuads latest authorative statement? 512-bit wide memory subsystem and GDDR5 - at 2500 MHz, no less.
http://www.fudzilla.com/index.php?option=com_content&task=view&id=6994&Itemid=1
I'm just starting to wrap my head around RV770's performance and it's shaping up to be a nice part.
If I could briefly outline what I believe it is fair to say are the likely specifications of RV770 and the performance gains that could accompany it.
First and foremost let's discuss the ROPs and the impact the changes from RV670 to RV770 are likely to have on performance in scenarios which are wholely-limited by the performance of the ROPs.
1) 30% higher clock over RV670 - 775MHz -> 1050MHz
2) 2x Z-rate
3) restored AA resolve with 4 samples per cycle
Combine the 30% higher clockspeed with the doubled Z & AA rates and it becomes clear that up to 160% performance improvement can be realized in Z/AA-bound scenarios.
Next up comes the TMUs.
Pretty straight-forward here, unless ATi has done something more than just double the TMUs (and up the clock 30%) we're once again looking at up to a 160% performance increase, this time in texture-bound scenarios.
On to the shader processors.
This time we get a more modest 50% increase in functional unit count/instruction throughput, by way of 480 SPs. Once again thanks to that 30% upclock however, we actually realize up to a 95% performance increase (architectural changes and unknown "other" tweaks aside) in shader-bound scenarios.
Lastly comes RAM bandwidth.
This is sort of a mixed-bag. Compared to the GDDR3 version of RV670, RV770 will see a small bump in memory clock from 1.8GHz to 2.2GHz, resulting in 22% more bandwidth, up from 57.6GB/s to 70.4GB/s. Unfortunately though, this is actually a slight decrease in bandwidth from GDDR4-equipped RV670 (which operates @ 2.4GHz and provides 76.8GB/s). It remains to be seen if this will hold back the performance of the RV770. Given recent comments on these very forums I believe RV770 may very well be deserving of a 512-bit bus or at least some ~4GHz GDDR5.
So it all boils down to sort of a mixed bag... It has been stated that performance is up over RV670 at least 50%, which is certainly respectable. Let's just hope 50% is more of a worst case than an average, because 50% ain't enough to retake even the single-GPU performance crown. Nvidia already makes a $350 GPU that performs at least 50% better than the HD 3870 in most scenarios - the 9800 GTX. Don't take my word for it, read this X-bit review for proof.
I have a feeling that the individual spec increases are going to lead to a lot more than just 50% performance gains in many common gaming scenarios. I'm just going to hope for the best. If nothing else, at least that gives NV incentive to get GT200 out the door ASAP and also gives price pressure for the GF9 series.
A dual-GPU RV770 + GDDR5 would be one badass graphics card.
If I could briefly outline what I believe it is fair to say are the likely specifications of RV770 and the performance gains that could accompany it.
First and foremost let's discuss the ROPs and the impact the changes from RV670 to RV770 are likely to have on performance in scenarios which are wholely-limited by the performance of the ROPs.
1) 30% higher clock over RV670 - 775MHz -> 1050MHz
2) 2x Z-rate
3) restored AA resolve with 4 samples per cycle
Combine the 30% higher clockspeed with the doubled Z & AA rates and it becomes clear that up to 160% performance improvement can be realized in Z/AA-bound scenarios.
Next up comes the TMUs.
Pretty straight-forward here, unless ATi has done something more than just double the TMUs (and up the clock 30%) we're once again looking at up to a 160% performance increase, this time in texture-bound scenarios.
On to the shader processors.
This time we get a more modest 50% increase in functional unit count/instruction throughput, by way of 480 SPs. Once again thanks to that 30% upclock however, we actually realize up to a 95% performance increase (architectural changes and unknown "other" tweaks aside) in shader-bound scenarios.
Lastly comes RAM bandwidth.
This is sort of a mixed-bag. Compared to the GDDR3 version of RV670, RV770 will see a small bump in memory clock from 1.8GHz to 2.2GHz, resulting in 22% more bandwidth, up from 57.6GB/s to 70.4GB/s. Unfortunately though, this is actually a slight decrease in bandwidth from GDDR4-equipped RV670 (which operates @ 2.4GHz and provides 76.8GB/s). It remains to be seen if this will hold back the performance of the RV770. Given recent comments on these very forums I believe RV770 may very well be deserving of a 512-bit bus or at least some ~4GHz GDDR5.
So it all boils down to sort of a mixed bag... It has been stated that performance is up over RV670 at least 50%, which is certainly respectable. Let's just hope 50% is more of a worst case than an average, because 50% ain't enough to retake even the single-GPU performance crown. Nvidia already makes a $350 GPU that performs at least 50% better than the HD 3870 in most scenarios - the 9800 GTX. Don't take my word for it, read this X-bit review for proof.
I have a feeling that the individual spec increases are going to lead to a lot more than just 50% performance gains in many common gaming scenarios. I'm just going to hope for the best. If nothing else, at least that gives NV incentive to get GT200 out the door ASAP and also gives price pressure for the GF9 series.
A dual-GPU RV770 + GDDR5 would be one badass graphics card.
- Status
