http://www.guru3d.com/newsitem.php?id=5361
New G92 card is said to be 3 times more powerful then the 8800GTX =/
New G92 card is said to be 3 times more powerful then the 8800GTX =/
cool, i have £1500 ready for that
tech has begun to move to quick.
fucking slow it down!
fucking slow it down!
isn't that multi-core chip? good that I didn't buy new gfx now and I think it will cost about as much as current high-end cards...
3930K | H100i | RIVF | 16GB DDR3 | GTX 480 | AX750 | 800D | 512GB SSD | 3TB HDD | Xonar DX | W8
ill take two plus a 5000w power supply and maybe some stock in an electric company
Last edited by wachtler83 (2007-05-30 09:09:30)
The fact that the FLOP rate is 3x higher does not mean the card will be 3 times faster, although it could be even quicker than that due to a number of other extended features. It is also rumoured to draw less power than the 8800.
Without knowing more about the architecture it is almost impossible to predict how the card will perform, even then there's no real substitute for proper performance testing.
I didn't expect a new Nvidia GPU series out so soon.
Without knowing more about the architecture it is almost impossible to predict how the card will perform, even then there's no real substitute for proper performance testing.
I didn't expect a new Nvidia GPU series out so soon.
More flops = Output.Bertster7 wrote:
The fact that the FLOP rate is 3x higher does not mean the card will be 3 times faster, although it could be even quicker than that due to a number of other extended features. It is also rumoured to draw less power than the 8800.
Without knowing more about the architecture it is almost impossible to predict how the card will perform, even then there's no real substitute for proper performance testing.
I didn't expect a new Nvidia GPU series out so soon.
No. More FLOPs = more FLoating point OPerations. What about integer operations? What about CPI? There are lots of other factors that are concerned with the architecture. The instruction set itself will play an important role as will the efficiency of a the pipeline (which will be an enhancement on the existing stream processing technology, combining vertex and pixel shaders, to perform T&L and rasterization concurrently). There are many important factors, you can't tell much from one of them on its own.cyborg_ninja-117 wrote:
More flops = Output.Bertster7 wrote:
The fact that the FLOP rate is 3x higher does not mean the card will be 3 times faster, although it could be even quicker than that due to a number of other extended features. It is also rumoured to draw less power than the 8800.
Without knowing more about the architecture it is almost impossible to predict how the card will perform, even then there's no real substitute for proper performance testing.
I didn't expect a new Nvidia GPU series out so soon.
Last edited by Bertster7 (2007-05-30 09:35:23)
it is very likely to be dual core chip. and combined with 50% increase in performance/core it would be theoretically 3x faster than 8800gtxBertster7 wrote:
No. More FLOPs = more Floating Point OPerations. What about integer operations? What about CPI? There are lots of other factors that are concerned with the architecture. The instruction set itself will play an important role as will the efficiency of a the pipeline (which will be an enhancement on the existing stream processing technology, combining vertex and pixel shaders, to perform T&L and rasterization concurrently). There are many important factors, you can't tell much from one of them on its own.cyborg_ninja-117 wrote:
More flops = Output.Bertster7 wrote:
The fact that the FLOP rate is 3x higher does not mean the card will be 3 times faster, although it could be even quicker than that due to a number of other extended features. It is also rumoured to draw less power than the 8800.
Without knowing more about the architecture it is almost impossible to predict how the card will perform, even then there's no real substitute for proper performance testing.
I didn't expect a new Nvidia GPU series out so soon.
3930K | H100i | RIVF | 16GB DDR3 | GTX 480 | AX750 | 800D | 512GB SSD | 3TB HDD | Xonar DX | W8
I'm not saying it won't be. I'm just saying don't start judging it on the FLOP rate. It's not a good indication. It's like judging it by clock rate - inaccurate. these things can only be used in comparison to over chips using the same architecture and we don't know what the architecture is yet. This is just more ill informed people jumping to conclusions before any real information is out.[69th_GFH]GC_PaNzerFIN wrote:
it is very likely to be dual core chip. and combined with 50% increase in performance/core it would be theoretically 3x faster than 8800gtxBertster7 wrote:
No. More FLOPs = more Floating Point OPerations. What about integer operations? What about CPI? There are lots of other factors that are concerned with the architecture. The instruction set itself will play an important role as will the efficiency of a the pipeline (which will be an enhancement on the existing stream processing technology, combining vertex and pixel shaders, to perform T&L and rasterization concurrently). There are many important factors, you can't tell much from one of them on its own.cyborg_ninja-117 wrote:
More flops = Output.
Remember the R600 that was going to blow the 8800 out of the water? The 2900XT appeared and is nothing super special - despite having a higher FLOP rate than the 8800GTX. Which gives a good indication that FLOP rates are not a good thing to judge performance by.
And I bet this thing is going to have the same price tag as a BMW 5 series
...My dad just ordered a 8800gts 640mb for my birthday(16! Yay!) on Monday... (I payed for half, just so people don't think I'm spoiled)
Technology is moving wayyy too fast, I don't like it, since I don't have an infinite amount of money at my disposal.
Technology is moving wayyy too fast, I don't like it, since I don't have an infinite amount of money at my disposal.
You don't like the fact that since technology is moving so fast, better components are becoming cheaper more quickly? How does that make things more expensive, unless you need to buy the best of everything straight away, which you don't - since software doesn't keep pace with these new hardware developments.Jello.01 wrote:
...My dad just ordered a 8800gts 640mb for my birthday(16! Yay!) on Monday... (I payed for half, just so people don't think I'm spoiled)
Technology is moving wayyy too fast, I don't like it, since I don't have an infinite amount of money at my disposal.
The faster technology improves, the sooner good components that will run current and near future software well become affordable. It's a good thing for the consumer, not a bad thing.
yea. have you seen those benchmarks about 2900xt shader power? it's 2x faster than 8800gtx! but because it sucks in pixel power it gets pwned by 8800 seriesBertster7 wrote:
I'm not saying it won't be. I'm just saying don't start judging it on the FLOP rate. It's not a good indication. It's like judging it by clock rate - inaccurate. these things can only be used in comparison to over chips using the same architecture and we don't know what the architecture is yet. This is just more ill informed people jumping to conclusions before any real information is out.[69th_GFH]GC_PaNzerFIN wrote:
it is very likely to be dual core chip. and combined with 50% increase in performance/core it would be theoretically 3x faster than 8800gtxBertster7 wrote:
No. More FLOPs = more Floating Point OPerations. What about integer operations? What about CPI? There are lots of other factors that are concerned with the architecture. The instruction set itself will play an important role as will the efficiency of a the pipeline (which will be an enhancement on the existing stream processing technology, combining vertex and pixel shaders, to perform T&L and rasterization concurrently). There are many important factors, you can't tell much from one of them on its own.
Remember the R600 that was going to blow the 8800 out of the water? The 2900XT appeared and is nothing super special - despite having a higher FLOP rate than the 8800GTX. Which gives a good indication that FLOP rates are not a good thing to judge performance by.
3930K | H100i | RIVF | 16GB DDR3 | GTX 480 | AX750 | 800D | 512GB SSD | 3TB HDD | Xonar DX | W8
A 4GHz P4 is significantly slower than a 2GHz C2D.cyborg_ninja-117 wrote:
http://www.guru3d.com/newsitem.php?id=5361
New G92 card is said to be 3 times more powerful then the 8800GTX =/
But yes, that does look promising.
http://www.fudzilla.com/index.php?optio … ;Itemid=34
http://www.legitreviews.com/news/3543/
PCIe 2.0?
Since we're using rasterized graphics, pixel power is very important. Good performance in one field is unacceptable, it is overall performance designers want. The 2900XT does not deliver.[69th_GFH]GC_PaNzerFIN wrote:
yea. have you seen those benchmarks about 2900xt shader power? it's 2x faster than 8800gtx! but because it sucks in pixel power it gets pwned by 8800 seriesBertster7 wrote:
I'm not saying it won't be. I'm just saying don't start judging it on the FLOP rate. It's not a good indication. It's like judging it by clock rate - inaccurate. these things can only be used in comparison to over chips using the same architecture and we don't know what the architecture is yet. This is just more ill informed people jumping to conclusions before any real information is out.[69th_GFH]GC_PaNzerFIN wrote:
it is very likely to be dual core chip. and combined with 50% increase in performance/core it would be theoretically 3x faster than 8800gtx
Remember the R600 that was going to blow the 8800 out of the water? The 2900XT appeared and is nothing super special - despite having a higher FLOP rate than the 8800GTX. Which gives a good indication that FLOP rates are not a good thing to judge performance by.
Intel (the worlds biggest manufacturer and supplier of GPUs) have plans to shift away from current graphics rendering techniques and try to move towards real time raytraced graphics, which I'd be keen to see, if they can pull it off.
is it really that much?damn,that's more than my computer costs all together.surgeon_bond wrote:
cool, i have £1500 ready for that
There is no way it'll cost that much. A 3rd of that perhaps.aj0404 wrote:
is it really that much?damn,that's more than my computer costs all together.surgeon_bond wrote:
cool, i have £1500 ready for that
Whatever the price tag will be, I'd like to see something similar to the 8800GTS 640/320-solutionBertster7 wrote:
There is no way it'll cost that much. A 3rd of that perhaps.aj0404 wrote:
is it really that much?damn,that's more than my computer costs all together.surgeon_bond wrote:
cool, i have £1500 ready for that
3D-Graphics are a pipeline of FLOPS (for tessellation) and more FLOPS = more frames per sec. (FPS) = faster card, in theory.
But the Graphics-Card has a CPU dependency; and has a code-dependency. The game-code (game-engine) is CPU related and structured in a linear fashion (usually). The GPU-pipeline will stall if it is waiting on an instruction being executed by the CPU.
It doesn't matter if the card is faster (more FLOPS) if:
1.) You have a slow/slower CPU.
2.) The coding of the video-driver is not state of art.
3.) The coding in DX9 (and its relation to the game-engine) doesn't support the advances included in the new/newer card design.
4.) The game itself can't take advantage of increases in resolution (res. modes).
5.) The meshes and textures in game are of lower resolution than the card is capable of supporting.
6.) Parallel processing (taking advantage of dual/quad core technology) has to be implemented in the: video-driver, in DX9 or DX10, and/or in the game engine itself; to make sure the 3D-graphics pipeline is not waiting on the CPU - so that FLOPS = FPS = the actual performance threshold is realized.
3x FLOPS will equal more FPS, but not 3x FPS, the increases in speed will be a percentage increase, and that (% increase in performance) will be dependent upon all those other things mentioned.
But the Graphics-Card has a CPU dependency; and has a code-dependency. The game-code (game-engine) is CPU related and structured in a linear fashion (usually). The GPU-pipeline will stall if it is waiting on an instruction being executed by the CPU.
It doesn't matter if the card is faster (more FLOPS) if:
1.) You have a slow/slower CPU.
2.) The coding of the video-driver is not state of art.
3.) The coding in DX9 (and its relation to the game-engine) doesn't support the advances included in the new/newer card design.
4.) The game itself can't take advantage of increases in resolution (res. modes).
5.) The meshes and textures in game are of lower resolution than the card is capable of supporting.
6.) Parallel processing (taking advantage of dual/quad core technology) has to be implemented in the: video-driver, in DX9 or DX10, and/or in the game engine itself; to make sure the 3D-graphics pipeline is not waiting on the CPU - so that FLOPS = FPS = the actual performance threshold is realized.
3x FLOPS will equal more FPS, but not 3x FPS, the increases in speed will be a percentage increase, and that (% increase in performance) will be dependent upon all those other things mentioned.
Last edited by topal63 (2007-05-30 10:18:08)
I'm sure you will. The GTSs have been very successful. I'm sure Nvidia plan to build on that success by doing more of the same.Stormscythe wrote:
Whatever the price tag will be, I'd like to see something similar to the 8800GTS 640/320-solutionBertster7 wrote:
There is no way it'll cost that much. A 3rd of that perhaps.aj0404 wrote:
is it really that much?damn,that's more than my computer costs all together.
yeah, AMD/ATI is in big trouble cuz of 8800GTS. and when G90 is released amd will be even more in trouble.Bertster7 wrote:
I'm sure you will. The GTSs have been very successful. I'm sure Nvidia plan to build on that success by doing more of the same.Stormscythe wrote:
Whatever the price tag will be, I'd like to see something similar to the 8800GTS 640/320-solutionBertster7 wrote:
There is no way it'll cost that much. A 3rd of that perhaps.
3930K | H100i | RIVF | 16GB DDR3 | GTX 480 | AX750 | 800D | 512GB SSD | 3TB HDD | Xonar DX | W8
Nvidia make a lot of brash claims about GPU FLOP rates. Remember when quad SLI came out? 6TFLOPS, apparently. Maybe so, but that power is not usable due to the way the cards are set up.topal63 wrote:
3D-Graphics are a pipeline of FLOPS (for tessellation) and more FLOPS = more frames per sec. (FPS) = faster card, in theory.
But the Graphics-Card has a CPU dependency; and has a code-dependency. The game-code (game-engine) is CPU related and structured in a linear fashion (usually). The GPU-pipeline will stall if it is waiting on an instruction being executed by the CPU.
It doesn't matter if the card is faster (more FLOPS) if:
1.) You have a slow/slower CPU.
2.) The coding of the video-driver is not state of art.
3.) The coding in DX9 (and its relation to the game-engine) doesn't support the advances included in the new/newer card design.
4.) The game itself can't take advantage of increases in resolution (res. modes).
5.) The meshes and textures in game are of lower resolution than the card is capable of supporting.
6.) Parallel processing (taking advantage of dual/quad core technology) has to be implemented in the: video-driver, in DX9 or DX10, and/or in the game engine itself; to make sure the 3D-graphics pipeline is not waiting on the CPU - so that FLOPS = FPS = the actual performance threshold is realized.
3x FLOPS will equal more FPS, but not 3x FPS, the increases in speed will be a percentage increase, and that (% increase in performance) will be dependent upon all those other things mentioned.
So is it fair to say that G92 cards will be 6x slower than a quad SLI setup? Of course not. Many other factors will play equally important roles. The GPU pipeline is extremely important in actually harnessing the power of the GPU efficiently - but the unified shader architecture seems to do a very good job of most of that stuff, so it's good to hear they are extending that.
If it costs more than 1200$ im out.
SLI's main advantage is not FPS at lower or normal resolutions - here there is little or no improvement in FPS. Currently, SLI is really all about AA and/or Higher Resolution FPS (with AA enabled). But once again there is a dependency to Code issue. Greater resolution = more pixels per view on triangle faces = more FLOPS needed = a FRAME occasionally being hung (or feeling like the action is stuttering) while waiting on Code being processed by the CPU. All game titles (now) are either OpenGL or DX9; and they are very linear (not much parallel code execution happening in current game titles). An SLI rig needs a good CPU to make it all run smoothly at higher resolutions (or with 16x AA enabled).Bertster7 wrote:
Nvidia make a lot of brash claims about GPU FLOP rates. Remember when quad SLI came out? 6TFLOPS, apparently. Maybe so, but that power is not usable due to the way the cards are set up.topal63 wrote:
3D-Graphics are a pipeline of FLOPS (for tessellation) and more FLOPS = more frames per sec. (FPS) = faster card, in theory.
But the Graphics-Card has a CPU dependency; and has a code-dependency. The game-code (game-engine) is CPU related and structured in a linear fashion (usually). The GPU-pipeline will stall if it is waiting on an instruction being executed by the CPU.
It doesn't matter if the card is faster (more FLOPS) if:
1.) You have a slow/slower CPU.
2.) The coding of the video-driver is not state of art.
3.) The coding in DX9 (and its relation to the game-engine) doesn't support the advances included in the new/newer card design.
4.) The game itself can't take advantage of increases in resolution (res. modes).
5.) The meshes and textures in game are of lower resolution than the card is capable of supporting.
6.) Parallel processing (taking advantage of dual/quad core technology) has to be implemented in the: video-driver, in DX9 or DX10, and/or in the game engine itself; to make sure the 3D-graphics pipeline is not waiting on the CPU - so that FLOPS = FPS = the actual performance threshold is realized.
3x FLOPS will equal more FPS, but not 3x FPS, the increases in speed will be a percentage increase, and that (% increase in performance) will be dependent upon all those other things mentioned.
So is it fair to say that G92 cards will be 6x slower than a quad SLI setup? Of course not. Many other factors will play equally important roles. The GPU pipeline is extremely important in actually harnessing the power of the GPU efficiently - but the unified shader architecture seems to do a very good job of most of that stuff, so it's good to hear they are extending that.
Current titles don't really take advantage of the unified shader architecture and its threshold capabilities. But next year with a bunch of DX10 releases being out or due-out - this will be rather a fun time for me.
And the best part is that, for those not interesting in getting DX10 hardware or Windows Vista right away, the DX10 hardware software advances will drive current DX9 & OpenGL hardware costs way down.
Last edited by topal63 (2007-05-30 11:45:58)