-
We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.
7900 GTX
- Thread starter Ramax57
- Start date
tuteja1986
Diamond Member
plenty of hyperflamewar detail :? or real detail ?
It's supposedly the 7800GTX with 8 extra pipelines, clocked at 700mhz.
So if you take the 7800GT at 400mhz x 20 pipelines x 500mhz RAM, you get "4 Million SickBeast Marks©".
So if you take the standard GTX at 430mhz x 24 pipelines x 600mhz RAM, you get "6.19 Million SickBeast Marks©".
If you take the GTX 512MB at 550mhz x 24 pipelines x 850mhz RAM, you get "11.22 Million SickBeast Marks".
If you take the "new" 7900GTX at 700mhz x 32 pipelines x 850mhz RAM, you get "19.04 Million SickBeast Marks©".
Now, quickly compared with the ATI cards:
For the X1800XT, at 625mhz x 16 pipelines x 750mhz memory, you get "7.5 Million SickBeast Marks©".
For the X1900XT at 625mhz x 48 pipelines x 725mhz memory, you get "21.75 Million SickBeast Marks©".
So there you have it. Based on current specs, the 7900GTX will be 88% the theoretical speed of the X1900XT, if my SickBeast Mark Rating System© is at all accurate.
I hereby take no responsibility for my rating system and please take my results with a few grains of salt.
These numbers were generated using basic high school math.
I hope you all enjoyed them. :beer:
So if you take the 7800GT at 400mhz x 20 pipelines x 500mhz RAM, you get "4 Million SickBeast Marks©".
So if you take the standard GTX at 430mhz x 24 pipelines x 600mhz RAM, you get "6.19 Million SickBeast Marks©".
If you take the GTX 512MB at 550mhz x 24 pipelines x 850mhz RAM, you get "11.22 Million SickBeast Marks".
If you take the "new" 7900GTX at 700mhz x 32 pipelines x 850mhz RAM, you get "19.04 Million SickBeast Marks©".
Now, quickly compared with the ATI cards:
For the X1800XT, at 625mhz x 16 pipelines x 750mhz memory, you get "7.5 Million SickBeast Marks©".
For the X1900XT at 625mhz x 48 pipelines x 725mhz memory, you get "21.75 Million SickBeast Marks©".
So there you have it. Based on current specs, the 7900GTX will be 88% the theoretical speed of the X1900XT, if my SickBeast Mark Rating System© is at all accurate.
I hereby take no responsibility for my rating system and please take my results with a few grains of salt.
These numbers were generated using basic high school math.
I hope you all enjoyed them. :beer:
your messing with pixel pipelines and shader pipelines. They are not the same, and they do not provide the same performance advantages. If you compare pixel pipes on nvidia, compare pixel pipes on ati, and you will see how far higher nvidia will score. viceversa, ati would score much higher.
the whole point is how applications make use of these different architectures, and ultimately which performs better. If you ask me going from 16 shader pipes to 48 shader pipes, did not give 3 times the performance to the X1900 from the previous gen. nVidias balance between increasing pixel pipes and shader pipes slowly seems to be working better then keeping the same number of pixel pipes and increasing a lot shader pipes. Proof of this is that the X1900 is not so much better then the 7800 GTX 512.
the whole point is how applications make use of these different architectures, and ultimately which performs better. If you ask me going from 16 shader pipes to 48 shader pipes, did not give 3 times the performance to the X1900 from the previous gen. nVidias balance between increasing pixel pipes and shader pipes slowly seems to be working better then keeping the same number of pixel pipes and increasing a lot shader pipes. Proof of this is that the X1900 is not so much better then the 7800 GTX 512.
One problem with the system is the X1900 only has 16 pipelines. It has 48 pixel SHADERS, not pipelines. so it would really be 625mhzx16 pipelinesx725mhz memory and you get "7.2 million SickBeast Marks".
That would be the true score. However, we all know the X1900 is faster than the 7800GTX so that system doesn't really work too well. So you should create a system that takes pixel shaders into account too then we would be able to compare video cards with SickBeast Marks.
Actually, the 7800 has 16 pipelines as well, if you count a pixel pipeline as starting with the shader and ending with the ROP. But, yeah, SickBeastMarks could be improved by throwing in the TMUs and ROPs. Then we could work on FP16 blend ops and maybe MADDs/ADDs per clock! 😀
Ramax, it's possible (even likely, IMO) the 7900 will be generally faster than the 1900. At the very least we can expect the 7900 to be faster without AA and the 1900 to be faster with AA. I expect it to be faster without AA whether it's a 700MHz 24 pipe card or a 550MHz 32 pipe card, but there's almost no question it'll be faster if its a 700MHz 32 pipe card (which I consider a huge reach, but that's just a skeptical hunch). The question is whether its bandwidth, which probably won't be much higher than the current GTX-512, will be a bottleneck and so will drop it to the 1900's level when you use AA. And when you buy a $400+ video card these days, you probably expect to use AA.
Ramax, it's possible (even likely, IMO) the 7900 will be generally faster than the 1900. At the very least we can expect the 7900 to be faster without AA and the 1900 to be faster with AA. I expect it to be faster without AA whether it's a 700MHz 24 pipe card or a 550MHz 32 pipe card, but there's almost no question it'll be faster if its a 700MHz 32 pipe card (which I consider a huge reach, but that's just a skeptical hunch). The question is whether its bandwidth, which probably won't be much higher than the current GTX-512, will be a bottleneck and so will drop it to the 1900's level when you use AA. And when you buy a $400+ video card these days, you probably expect to use AA.
the 48 pipelines on the 1900XT are kind of....not all useful...right?
i mean, even with all those pipelines, which game will ever use them all and have considerably better performance then with 20, or 24, or EVEN 16?
i just want to know the whole meaning behing the pipelines, and what makes the card so much better because of them. i still trust in nVidia's 7900GTX will beat out the 1900XT overall, and be a winner, as always 😉
i mean, even with all those pipelines, which game will ever use them all and have considerably better performance then with 20, or 24, or EVEN 16?
i just want to know the whole meaning behing the pipelines, and what makes the card so much better because of them. i still trust in nVidia's 7900GTX will beat out the 1900XT overall, and be a winner, as always 😉
I think the only way the G71 can possibly out perform the 1900XXTTX is by skipping the pipelines all together and concentrating on consonants at the end of the model number. Lets face it boys GTX just no longer cuts it in a dual core dominated world. Double the cores you better double your model number suffix. A 7900 GTX has no chance to catch the newest from ATI. But a 7900 GTXTIPDE OC will clearly pummel ATI. So here's drinking to the GTXTI Platinum Diamond Special Edition Over the top clocker.
The blizzard is only going to make this worse🙁
The blizzard is only going to make this worse🙁
Heh, OatMan.
ChonChon, you're right that games haven't shown a X1900 to be three times as fast as a X1800, even tho it has three times the pixel shader units (not pipelines). In fact, I don't think even synthetic benchmarks have shown a X1900 to be three times as fast, tho I think some have shown well over a 2x improvement. Bottlenecks abound in a video card, so while the X1900 might have 3x the pixel shaders, games aren't yet totally pixel-shader-limited. The reason for that? Well, b/c the X1900 (and X1600, which also has an unusually high number of pixel shaders) isn't quite like previous video cards. Every video card before it has been much more "balanced" between shader operations and texture operations. Well, the GF7 series has beefier pixel shader units than the GF6 series, so you can say the 7800 was the first step in the X1900's direction.
But most games will continue to focus on running best on the most hardware, and for now that's more "balanced" GPUs. ATI probably needs to convince developers to crank the shader usage in order for the X1900 and X1600 to really use all that shader power. They need to shift the bottleneck toward the pixel shading. It remains to be seen if they can do it within the X1900's and X1600's useful life. The shader power of the GPUs in Xbox 360 and PS3 is encouraging, but I don't think we'll see the X1900 scale quite to the level of its pixel shaders (meaning, we won't see it 3x faster than the otherwise identical X1800).
So, yeah, not all useful in current games/benchmarks, but it's a better balance than the X1800, which seems to have too few pixel shaders relative to the 7800.
The meaning of a pipeline is changing relative to, say, a Voodoo 2. Let's just say it's still easiest to just run some benchmarks, but you now have to take into account pixel shader power in addition to texturing "power" and even rendering "power." In that sense, a 7900 may be better balanced than a X1900, at least for current games. We'll find out soon enough, though, and the X1900 is sitting pretty ATM.
ChonChon, you're right that games haven't shown a X1900 to be three times as fast as a X1800, even tho it has three times the pixel shader units (not pipelines). In fact, I don't think even synthetic benchmarks have shown a X1900 to be three times as fast, tho I think some have shown well over a 2x improvement. Bottlenecks abound in a video card, so while the X1900 might have 3x the pixel shaders, games aren't yet totally pixel-shader-limited. The reason for that? Well, b/c the X1900 (and X1600, which also has an unusually high number of pixel shaders) isn't quite like previous video cards. Every video card before it has been much more "balanced" between shader operations and texture operations. Well, the GF7 series has beefier pixel shader units than the GF6 series, so you can say the 7800 was the first step in the X1900's direction.
But most games will continue to focus on running best on the most hardware, and for now that's more "balanced" GPUs. ATI probably needs to convince developers to crank the shader usage in order for the X1900 and X1600 to really use all that shader power. They need to shift the bottleneck toward the pixel shading. It remains to be seen if they can do it within the X1900's and X1600's useful life. The shader power of the GPUs in Xbox 360 and PS3 is encouraging, but I don't think we'll see the X1900 scale quite to the level of its pixel shaders (meaning, we won't see it 3x faster than the otherwise identical X1800).
So, yeah, not all useful in current games/benchmarks, but it's a better balance than the X1800, which seems to have too few pixel shaders relative to the 7800.
The meaning of a pipeline is changing relative to, say, a Voodoo 2. Let's just say it's still easiest to just run some benchmarks, but you now have to take into account pixel shader power in addition to texturing "power" and even rendering "power." In that sense, a 7900 may be better balanced than a X1900, at least for current games. We'll find out soon enough, though, and the X1900 is sitting pretty ATM.
coldpower27
Golden Member
The 48 Pixel Shaders in R580 are all useful, it just that if you don't increase the TMU's to this amount as well, you can't get a 3x fold performance in many cases. You also need at least I would recommend 0.66x of the amount of pipelines in the amount of ROP's, so you can output the pixels.
If G71 is indeed 32Pixel Shaders/32 TMU's/24ROPs cloked at 700mHZ, coupled with GDDR1.1 at 1.8GHZ, it would be safe to say it should beat out the X1900 Series.
Graphics are a very parrallel thing, you can make use of 48 Pipelines if we had them, a Pipeline being a "1 Pixel Shader, 1 TMU, 0.66x or higher ROP Unit"
Pipelines made sense when we had a 1:1 Ratio of everything, now we are moving beyond that. With the X1900 introducing a 3:1 for Pixel Shader to TMU's for high end cards.
Note the emphasis will be on increase in Shader Power, memory bandwidth is unlikely to increase by a whole lot as we are pretty near the maximum for GDDR3.
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 25K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes + WCL Discussion Threads
- Started by Tigerick
- Replies: 25K
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 24K
-
-
