Originally posted by: ArchAngel777
Doubt we will see anything under $400. They will probably come out with an XT and XTX version of it first, then follow up with something around April/May in the midrange. All speculation at this point, but I would highly doubt we see any midrange parts for quite a while after the initial flagship release.
I am pretty excited to see this new card of ATi's. I was origional planning an entire system overhaul, however, if they come out with a decent mid-range card sometime before June, I will just spend $300 or so and save the rest for a future upgrade.
Originally Posted by Morgoth the Dark Enemy from beyond3d
This is going bad...it`s worse than the whole 512-bit stuff. It wasn`t like that with the G80, it seems that ppl expect ATi to make a card that would improve even starship Enterprise`s systems AND wipe your ass at the same time.
Originally posted by: Cookie Monster
Firstly, what the heck is madd4 shaders? people who dont know these things shouldnt act like they do. You people realise G80 use scalar shaders which are madd+mul capable. The current prospect is that R600 is going to be vec4+scalar (like the Xenos) or vec4, or even vec2+vec3. Althought theorectically 64 vec4 shaders clocked at ~750mhz equal 128 scalar shaders clocked at 1350mhz, it has its advantages and disadvantages. If its vec4, it can do more work per cycle, (4 compared to 1) but the ultisation of the shaders are far less then scalar ones. This is why nVIDIA chose scalar shaders because you are garaunteeing 100% ultisation which the vec4 shaders can only dream of matching (increasing the core past 1ghz is going to be mission impossible). In a real world performance sense, 100% ultisation sounds more promising than less ultisation but more work per cycle.
This is why ATi is rumoured to hate the G80 so much. Although current drivers maybe a mess, the architecture itself is very impressive and the bar that nVIDIA has raised is VERY high. So right now, the R600 isnt going to rape anything in that sense. Sites that said about ATi having 32 ROPs is just BS, because a) having more ROPs dont equal in gaining performance and b) it would be a waste of transistor budget. The only good thing is that its a good marketing material for the ATi PR. Not to mention 2GB of GDDR4? Unless you want AMD to lose money by selling these rumoured cards at affordable prices, i think theyve made the whole thing up. Half of the specs listed by was already floating around in the internet by the INQ and word of mouth.
Literally, dont get your hopes high because it will get to you once the real benchmarks hit. Im thinking its either 5~15% faster in some 3d apps while slower in others compared to the G80. AVIVO will be there i bet, and the "new" crossfire. I think the transistor count is so high (720 million rumoured) because of the ring bus design (which also casues the chip to have much more pins, e.g R600 die shot had 2240 pins), its unified shader architecture using vec4+scalar shaders, and the 512bit memory interface.
GDDR3? i guess GDDR4 isnt a good option due to availability as ATi hoped for. However hearing about their 2nd revision working is great and working at 800mhz is pretty amazing. But werent they aiming for 750mhz initially? 800mhz sounds abit too much for such a complex chip at 80nm. G80 on the other hand has much more headroom, with the GTX managing avg of 650~ from cherry picked cores (which means 1.5ghz shader clock).
Originally Posted by Morgoth the Dark Enemy from beyond3d
This is going bad...it`s worse than the whole 512-bit stuff. It wasn`t like that with the G80, it seems that ppl expect ATi to make a card that would improve even starship Enterprise`s systems AND wipe your ass at the same time.
Originally posted by: Matt2
I was just gonna post about scalar shaders being 100% utilized.
I think it's also worth mentioning that we might see Nvidia drop the driver bomb on R600 on Jan. 20th.
I'm personally holding my breath for Nvidia to release a driver that unlocks the full potential of this card. Think of what Nvidia did with the original Detonator drivers.
My reasoning for thinking that? There is no reason why an 8800GTS overclocked should come close to matching 8800GTX performance with a 320bit bus and less shaders.
Driver bomb? They should start by releasing a driver that doesn't have issues with basic rasterization.I think it's also worth mentioning that we might see Nvidia drop the driver bomb on R600 on Jan. 20th.
Originally posted by: BFG10K
Driver bomb? They should start by releasing a driver that doesn't have issues with basic rasterization.I think it's also worth mentioning that we might see Nvidia drop the driver bomb on R600 on Jan. 20th.
Bah! I guess I will just pick up an X1950XT in Janury than, really wanted a mainstream DX10 though :/Originally posted by: TBSN
Originally posted by: ArchAngel777
Doubt we will see anything under $400. They will probably come out with an XT and XTX version of it first, then follow up with something around April/May in the midrange. All speculation at this point, but I would highly doubt we see any midrange parts for quite a while after the initial flagship release.
I am pretty excited to see this new card of ATi's. I was origional planning an entire system overhaul, however, if they come out with a decent mid-range card sometime before June, I will just spend $300 or so and save the rest for a future upgrade.
So the ATI D600 XT and XTX would be roughly equivalent (in terms of performance and price difference to each other) to the Nvidia GTS and GTX? I think I may do the same as you and save my money until they come out, and see whatever comes out on top, (in terms of value) G80 or R600...
Originally posted by: Matt2
Sorry ronn, I know it was another post about Nvidia's drivers, but BFG and Pugnante egged me on!
If that is really what they're doing then Nvidia has the worst business practice out of any GPU vendor. To make their early adopters pay a premium to do nothing but deal with troubleshooting and frustration just so they can one-up the competition several months down the road isn't something I think enthusiasts appreciate.I think they're keeping you on the edge of your seat BFG.
Originally posted by: munky
The only driver bomb I expect to magically boost performance for the g80 is an fx5800-style hack/cheat bomb. And I dont think Nvidia want that fiasco to happen all over again. If you look at the Firingsquad g80 overclocking article you can see that in most tests a 8800gts @647mhz still falls behind a stock 8800gtx, sometimes by a large margin. I dont buy into any theories that Nv is holding back performance with the g80. If anything, the most likely response to the r600 I can expect from Nv is a higher clocked version of the 8800gtx using cherry-picked cores.
Originally posted by: Cookie Monster
Firstly, what the heck is madd4 shaders? people who dont know these things shouldnt act like they do. You people realise G80 use scalar shaders which are madd+mul capable. The current prospect is that R600 is going to be vec4+scalar (like the Xenos) or vec4, or even vec2+vec3. Althought theorectically 64 vec4 shaders clocked at ~750mhz equal 128 scalar shaders clocked at 1350mhz, it has its advantages and disadvantages. If its vec4, it can do more work per cycle, (4 compared to 1) but the ultisation of the shaders are far less then scalar ones. This is why nVIDIA chose scalar shaders because you are garaunteeing 100% ultisation which the vec4 shaders can only dream of matching (increasing the core past 1ghz is going to be mission impossible). In a real world performance sense, 100% ultisation sounds more promising than less ultisation but more work per cycle.
This is why ATi is rumoured to hate the G80 so much. Although current drivers maybe a mess, the architecture itself is very impressive and the bar that nVIDIA has raised is VERY high. So right now, the R600 isnt going to rape anything in that sense. Sites that said about ATi having 32 ROPs is just BS, because a) having more ROPs dont equal in gaining performance and b) it would be a waste of transistor budget. The only good thing is that its a good marketing material for the ATi PR. Not to mention 2GB of GDDR4? Unless you want AMD to lose money by selling these rumoured cards at affordable prices, i think theyve made the whole thing up. Half of the specs listed by was already floating around in the internet by the INQ and word of mouth.
Literally, dont get your hopes high because it will get to you once the real benchmarks hit. Im thinking its either 5~15% faster in some 3d apps while slower in others compared to the G80. AVIVO will be there i bet, and the "new" crossfire. I think the transistor count is so high (720 million rumoured) because of the ring bus design (which also casues the chip to have much more pins, e.g R600 die shot had 2240 pins), its unified shader architecture using vec4+scalar shaders, and the 512bit memory interface.
GDDR3? i guess GDDR4 isnt a good option due to availability as ATi hoped for. However hearing about their 2nd revision working is great and working at 800mhz is pretty amazing. But werent they aiming for 750mhz initially? 800mhz sounds abit too much for such a complex chip at 80nm. G80 on the other hand has much more headroom, with the GTX managing avg of 650~ from cherry picked cores (which means 1.5ghz shader clock).
Originally Posted by Morgoth the Dark Enemy from beyond3d
This is going bad...it`s worse than the whole 512-bit stuff. It wasn`t like that with the G80, it seems that ppl expect ATi to make a card that would improve even starship Enterprise`s systems AND wipe your ass at the same time.
Originally posted by: Cookie Monster
:Q
Firstly the X1800/X1900 used vec3+scalar.. So did the R300..
Im not talking crap, because things you say right now which are false can make the other less informed to really believe thats the case.
You have some good points though, although some of them dont makes sense.
Originally posted by: Cookie Monster
Originally posted by: munky
The only driver bomb I expect to magically boost performance for the g80 is an fx5800-style hack/cheat bomb. And I dont think Nvidia want that fiasco to happen all over again. If you look at the Firingsquad g80 overclocking article you can see that in most tests a 8800gts @647mhz still falls behind a stock 8800gtx, sometimes by a large margin. I dont buy into any theories that Nv is holding back performance with the g80. If anything, the most likely response to the r600 I can expect from Nv is a higher clocked version of the 8800gtx using cherry-picked cores.
You seem to enjoy bringing up "fx5800-style hack/cheat bomb". What makes you think so? These performance drivers are all possibilities because the G80 drivers hasnt given one yet. IF you look at most generations, it took couple of driver versions to gain further 10~30% performance e.g the 7800GTX, and the X1800XT. How about the GF3? However what we need is isnt a performance driver, but rather a special "fix" driver to get rid of most the bugs.
nVIDIA might be having trouble also becauses it a new architecture beyond what they have experienced before.
Back to the R600. Looks like R600 is ready to go. Source: beyond3d. January 20th doesnt look that bad as a launch date
Originally posted by: josh6079
If that is really what they're doing then Nvidia has the worst business practice out of any GPU vendor. To make their early adopters pay a premium to do nothing but deal with troubleshooting and frustration just so they can one-up the competition several months down the road isn't something I think enthusiasts appreciate.I think they're keeping you on the edge of your seat BFG.
Originally posted by: schneiderguy
anyone know what the MSRP of the highest end R600 card is going to be? hopefully ATI wont follow in nvidia's footsteps of raising their top end card to $650 :|
Originally posted by: Cookie Monster
You got yours at 650/1000? wow :Q
Exactly.If you look at the driver history of of the last few generations, there were steady, gradual improvements in performance, a trend I expect to continue with the g80. However, with the g80 some people seem to expect a major performance boost from unlocking some secret power that lay dormant all this time, and in the most recent time Nvidia tried to accomplished such a feat was by cheating in benchmarks via clip planes and shader replacement, and lowering IQ in areas such as texture filtering/sampling. I find the whole notion of Nvidia intentionally holding back the g80 peroformance utterly ridiculous. If the g80 is limited in performance, it's from immature drivers. But that's the way things go with all newly designed gpus, and can hardly be considered a secret weapon driver bomb waiting in the wings to be used at the right time.
I expect the r600 to go through the same process of increasing performance with maturing drivers. I do not expect either card to receive major performance boosts from a certain driver bomb.