Nvm, made in china... low quality..
bleh, sometimes people need to relax abit, or i just suck at jokes!
The Official G84/G86 Rumours & Speculation Thread
- Thread starter Arun
- Start date
Nvm, made in china... low quality..
bleh, sometimes people need to relax abit, or i just suck at jokes!
Float HDR does truly have a higher bandwidth hit. Given that neither Doom3 or Fear don't have HDR, where's the relevance to that?
Let me take a different perspective: if bandwidth was truly the primary bottleneck in Doom3/FEAR in that review, then why isn't the trend the same in other applications? Because Doom3 consumes supposedly more bandwidth then a game like Riddick or even STALKER?
Nvm, made in china... low quality..
bleh, sometimes people need to relax abit, or i just suck at jokes!
You should try to detect the intended sarcasm at times. Open up any product manufactured in "Germany" today and we'll see where the majority of parts comes from after all
You got me wrong, I wasn't saying it's the bandwidth.
The relevance is that 06 SM3.0 test is one thing it seems to be doing clearly worse than X1950Pro. The Pro would score lower on the SM2.0 test buth higher on SM3.0 test.
You got me wrong, I wasn't saying it's the bandwidth.
My point was that in that case the primary bottleneck might actually be bandwidth, because of float HDR being used in the SM3.0 tests.
Au contraire to the Doom3/FEAR results where there's either some sort of weird driver bubble or the ROPs are bottlenecking performance. If the 3dmark03 score above in trinibwoy's INQ link had comparable scores with 4xAA in 1600 on X1950/7900s it would help to see what would be at fault here.
You should try to detect the intended sarcasm at times. Open up any product manufactured in "Germany" today and we'll see where the majority of parts comes from after all
Hmm, *checks sarcasm detector*
O its broke..
Anyway, on topic
Could we potentially see the 8600 series alot earlier than first thought? So many leaks of pics/benches and im guessing theres going to be alot to go around during launch.
But all benches so far point to a bottleneck in the 8600 series. The cause could be because of
a) bandwidth
b) ROP
c) or driver issues.
I think we are comparing these cards to X1950XT, 7900GTX level, instead of the 7950GT or the X1950pro level which sounds more appropriate.
Err, what? What was your "first thought"?
It looks halfway decent to me. But then again, so does my 1900GT
This was never meant to be more than halfway decent, or else it would eat into 8800 territory.
NV certainly doesn't want to lose those higher margins due to internal competition from a mainstream part.
That and I'd also expect in due time to see from both IHVs Performance GPUs. Considering those parts usually arrive somewhere around the $250 MSRP, I don't think these =/<$200 GPUs have a bad price/performance ratio in the grander picture of a D3D10 top to bottom product family.
Comparisons with former generation Performance GPU aren't invalid, yet I would expect users that own such GPUs to upgrade actually to a Performance D3D10 in the foresseable future then to a mainstream part.
http://www.vr-zone.com/?i=4883&s=2
Now exactly why would they bench it overclocked vs a stock 7900 GT with no AA?
Now exactly why would they bench it overclocked vs a stock 7900 GT with no AA?
Ah, so that 3DM06 score is with the default 1x4 settings, not the imaginary settings where a 128b card whips 256b ones at high res with AA. Makes more sense.
The extra bandwidth seems to help the GTS a lot. It's got a ~12% higher core speed, ~43% higher RAM speed, and it scores ~19% faster than the GT. Slightly strange is that it does so across the board, both at 10x7 1x1 and at 16x12 4x16.
Are the shader units clocked at core speed or at a multiple of ~2? I'm guessing multiple, given the superlinear framerate increases and assuming that the shader units aren't starved for bandwidth at 10x7 1x1. Which I guess might possibly explain underperforming in FEAR (Ail's test is the obvious check). No idea about D3--I didn't think it was a bandwidth hog, but if they tested at the IMO silly ultra HQ setting, I can see b/w (or framebuffer? I forget Ail's framebuffer #s, and his site isn't cooperating).
But, man, that thing's kickin' some ass in Riddick, even over a 7900GT. I thought the GF7 series handed the RX1 series its ass across the board in that game?
The extra bandwidth seems to help the GTS a lot. It's got a ~12% higher core speed, ~43% higher RAM speed, and it scores ~19% faster than the GT. Slightly strange is that it does so across the board, both at 10x7 1x1 and at 16x12 4x16.
Are the shader units clocked at core speed or at a multiple of ~2? I'm guessing multiple, given the superlinear framerate increases and assuming that the shader units aren't starved for bandwidth at 10x7 1x1. Which I guess might possibly explain underperforming in FEAR (Ail's test is the obvious check). No idea about D3--I didn't think it was a bandwidth hog, but if they tested at the IMO silly ultra HQ setting, I can see b/w (or framebuffer? I forget Ail's framebuffer #s, and his site isn't cooperating).
But, man, that thing's kickin' some ass in Riddick, even over a 7900GT. I thought the GF7 series handed the RX1 series its ass across the board in that game?
Perhaps this had something to do with it.
It seems to be available around the world already, the 2 week waiting period can only be described as a way to better prepare the marketing blitz.
Yea, thats what i meant. April 17th is what i first thought but these cards are already on sale at some retailers!
http://www.pricespy.co.nz/search.php?q=8600GTS
In NZ we already got several being listed already, pre order, some are being sold.
Ne that This product will be in stock on Monday 16 April, 2007 according to one retailer in NZ.
April 16th dont sounds too bad than April 17th..
jimmyjames123
Regular
trinibwoy said:
Maybe because they feel that the framerates will not be very playable at these resolutions on these games with these cards when using AA?
However, I can't say why there is a bench between overclocked and not overclocked.This:
Coupled with this, taken from the latest FW 101.70 drivers (notice the order of each G80 card, and how the 8800 Ultra is below the 8800 GTS, with a higher id number -GTX has the lowest so far-):
Leads me to believe that the "8800 Ultra" may actually be an even cheaper version of the G80 processor, perhaps with only 64 sp's (but keeping the 320bit bus), or 96 sp's and a 256bit bus.
Of course, this is just a guess, since Ultra is a synonym of "cheaper", instead of "faster" lately in Nvidia's vocabulary (650i Ultra, NF4 Ultra, etc).
Coupled with this, taken from the latest FW 101.70 drivers (notice the order of each G80 card, and how the 8800 Ultra is below the 8800 GTS, with a higher id number -GTX has the lowest so far-):
Leads me to believe that the "8800 Ultra" may actually be an even cheaper version of the G80 processor, perhaps with only 64 sp's (but keeping the 320bit bus), or 96 sp's and a 256bit bus.
Of course, this is just a guess, since Ultra is a synonym of "cheaper", instead of "faster" lately in Nvidia's vocabulary (650i Ultra, NF4 Ultra, etc).
Similar threads
