AMD: R7xx Speculation
- Thread starter Unknown Soldier
- Start date
- Status
Its hard for me to believe they get that much of a boost at those resolutions when you'd think bandwidth would be a limiting factor (which favors 3870).
Does rv670 performance tank at high AA/AF?
Sqrt(256)=16.00
Sqrt(256)=16.58
!6.00/16.58=0.965 or 96.5%
I think a difference in linear dimensions of 3.5% is within the margin of error at this stage. The original resolution of that RV770 die shot (before resizing it for comparison purposes) was low enough to allow for the discrepancy. As is, for that matter, someone measuring 16.x mm with a ruler, and calling it 16.
The amusing thing about running the numbers out...
9800GTX = 604
4850 = 647.5
4870 = 777(±2.5) *funny number btw...
Vantage Texture Performance Increase-
777/604 = 1.29
647.5/604 = 1.07
Goes to show where the vast majority of the performance is coming from...
Now I am excited to see how good performance will scale with overclocks!
40% overclock anyone?
Last edited by a moderator:
So we need some high-rez photos to be sure...
Maybe with LN².
Rumors: Lie or true ?? I don't know....
http://www.tomshardware.com/forum/249173-33-more-rv770-details-chiphellThose guys in China seem to have special way to get info.
Today chiphell release more detail about coming RV770.
You know that both R600 and RV670 have 64*5D units.
Now, sit tight, don't scream.
RV770 will has 160*5D :bounce:
The TMU units and ROP will increase "An awful lot".
Whatever it means "An awful lot" I would never think - it will be over 40TMU's and 24ROP's.
Edit: based on speculation An awful lot
RV770XT
55nm
750MHz core
1050MHz shaders
160 *5D (800 Stream Processors)
32/40 TMU's
20/24 ROP's
GDDR5 1800MHz (~115GB's bandwidth)
$349.00 US dollars
Last edited by a moderator:
Why would ROPs/RBEs need to increase? I haven't heard them referred to as a limiting factor, except maybe their Z rate, which could be doubled without affecting the # of RBEs. Not to mention that a 256-bit memory bus with GDDR3 for the 4850 doesn't seem to be screaming for more pixels per clock relative to RV670.
TUs is what we're all hoping and expecting to be increased. Sure seems like they have been, given that Vantage MT fillrate #. I'm guessing much of the improvement in CoD4: Modern Wharf is due to more TUs more than anything else.
TUs is what we're all hoping and expecting to be increased. Sure seems like they have been, given that Vantage MT fillrate #. I'm guessing much of the improvement in CoD4: Modern Wharf is due to more TUs more than anything else.
http://www.amazon.com/Diamond-4850P...f=sr_1_59?ie=UTF8&s=pc&qid=1213253657&sr=1-59
Amazon claimed 480SP's for HD4850....
lol: Maybe HD4870 will have 800SP's and HD4850 480SP's.
Silent_Buddha
Legend
U wrong..
Call of Juarez 87%
Call of Duty 130%
Crysis 81%
Er so you're trying to say that 48xx in those charts is slower than 3870 in 2 games?
I think I'll go with Jawed's numbers.
Remember those numbers are x% of 3870. So 187% of 3870 is a bit less than twice as fast (200%).
Now, just need to know if one small rumor is true or not. When the Inq reported they held a 4870 card with only 1x6 pin power connector.
If that's true...jebus ATI pulled a miracle out of their bag of tricks.
Somehow I doubt it's true though.
Regards,
SB
leoneazzurro
Regular
http://www.amazon.com/Diamond-4850P...f=sr_1_59?ie=UTF8&s=pc&qid=1213253657&sr=1-59
Amazon claimed 480SP's for HD4850....
lol: Maybe HD4870 will have 800SP's and HD4850 480SP's.
They also write "RV7000" as the new GPU name
Er so you're trying to say that 48xx in those charts is slower than 3870 in 2 games?
I think I'll go with Jawed's numbers.
Remember those numbers are x% of 3870. So 187% of 3870 is a bit less than twice as fast (200%).
Now, just need to know if one small rumor is true or not. When the Inq reported they held a 4870 card with only 1x6 pin power connector.
If that's true...jebus ATI pulled a miracle out of their bag of tricks.
Somehow I doubt it's true though.
Regards,
SB
I forgot the plus
Sorry Jawed, u were right.
http://www.tgdaily.com/html_tmp/content-view-37907-135.html
AMD’s physics secret revealed: It’s Havok
The two companies said that they are also “investigating” the use of “AMD’s
massively parallel ATI Radeon GPUs to manage appropriate aspects of physical
world simulation in the future.”
AMD’s physics secret revealed: It’s Havok
The two companies said that they are also “investigating” the use of “AMD’s
massively parallel ATI Radeon GPUs to manage appropriate aspects of physical
world simulation in the future.”
Berek
Regular
Looks like the 4870 is due to strike against the 9800gtx. This is interesting news, but I'd still like to see how it will do against the 9800gx2 or the GT200 cards. Either way, at the price point they are going with this card, it sounds like it will be a real winner.
http://www.fudzilla.com/index.php?option=com_content&task=view&id=7854&Itemid=1
On the flip side (if there is one), Nvidia's GT200 cards might end up priced lower:
http://www.fudzilla.com/index.php?option=com_content&task=view&id=7853&Itemid=1
http://www.fudzilla.com/index.php?option=com_content&task=view&id=7854&Itemid=1
On the flip side (if there is one), Nvidia's GT200 cards might end up priced lower:
http://www.fudzilla.com/index.php?option=com_content&task=view&id=7853&Itemid=1
Jawed
Legend
Except that bandwidth seems to have rarely been an issue for HD3870 (if ever, in games). In comparison with G92, which is clearly bandwidth limited, the relative performance of HD3870 indicates it would be happy with considerably less bandwidth. I've suggested 50GB/s, but that is a guess.
Anyway, with HD3870 not being bandwidth efficient while HD4850 is bandwidth limited, the relative difference seems unsurprising to me.
HD4850 being 20% more bandwidth efficient than 9800GTX is a big deal. We're expecting GT200 to be more bandwidth efficient too.
Going from 4xAA to 8x isn't what you'd call tanking, generally. It's usually the first step, to 16xAF/4xAA, that does the damage.
Jawed
- Status
