If that's the case Riddle, what not just stick with your PC? I'm less enthused with the Xbox BECAUSE of the similarities. Throwing in a lot more isn't going to help, don't think. It'll just make the gaming experience even closer--which would be a lot of useless overlap.
TXB on Xbox2 = 51.2GB/s bandwidth??
- Thread starter chaphack
- Start date
DeadmeatGA
Banned
...
XGPU2 having eDRAM is a different matter from R500 having one altogether. While R500 has to support a frame buffer size upto 1600x1200, XGPU2 has no requirement and can be optimized for one specific resolution, namely 640x480. Such optimization affords the use of 5.2 MB eDRAM that R500 can't have.
XGPU2 having eDRAM is a different matter from R500 having one altogether. While R500 has to support a frame buffer size upto 1600x1200, XGPU2 has no requirement and can be optimized for one specific resolution, namely 640x480. Such optimization affords the use of 5.2 MB eDRAM that R500 can't have.
my $0.02
memory bandwidth will not be the most critical issue in the next generation, the speed of pixel shaders will. having super fast memory won't matter if it's sitting idle waiting for results from the pixel program. based on what NV and ATI have done in the past I think they would rather spend their transistor budget on logic than embedded memory.
memory bandwidth will not be the most critical issue in the next generation, the speed of pixel shaders will. having super fast memory won't matter if it's sitting idle waiting for results from the pixel program. based on what NV and ATI have done in the past I think they would rather spend their transistor budget on logic than embedded memory.
DemoCoder said:
Thank you, thank you. I've debated posting something along these lines several times now and have not because I've learned that people only care about numerology and use the "bigger" nomenclature = "better" ideology.
There is alot of prediction in this thread that I find ridiculous (eg. Segemented RAM, Multichip, and Fragment Shading's supposed importance to name only a few), but who am I to contradict others?
Panajev2001a
Veteran
Nappe1 said:
PC GPUs also have, separate from system RAM, VERY fast ( compared to system RAM ) off-chip VRAM.
Consoles would likely have a UMA or hybrid UMA configuration.
Riddlewire
Regular
cthellis42 said:
Cost.
That's the reason most people like me bought Xboxes. For $199 we got a high powered game machine that didn't require $500 in upgrades once a year in order to play the top games for the system. Additionally, I've been a PC gamer all my life and I had never owned a console (except a C64). I wanted to get some of the other types and styles of games that aren't always featured on the PC (fighting games, etc.). With the Xbox, I got enough overlap to provide me with the big name FPS and action games, but I also got games that I never had access to before. It's also a nice change of pace to be able to just sit in the floor in front of my TV and play some HALO co-op with a friend, as opposed to the more solitary gaming environment on the PC. I'll never give up PC gaming, but my upgrades have now been extended to the 3 year range. Thankfully, all the games I'm looking forward to that I'd normally have to upgrade my PC for are also coming out for the Xbox. Unfortunately, I'll just have to endure the lousy gamepad control.
So basically, I'm just very selfish and cheap. I want the exact games that I want, with the exact control that I want, for as little money as possible.
Megadrive1988
Veteran
but isn't it likely that ATI's R500 will go beyond the PS/VS 3.0 that will be supported by R420 and NV40 ? And since XBox 2 graphics will almost certainly be based on R500 (or R550) would it not be reasonable to expect that XBox 2 has similar shader capabilities as the PC R5XX VPUs?
I am not saying that XBox 2 will "sux" if it only has VS/PS 3.0 - I am just saying that going by XBox 2's timeframe (2005-2006) it will probably have a graphics processor with modern shading capabilities, much like the Xbox of 2001 had abilities beyond the GPUs from 1999-2000.
I think its only reasonable....
as for segmented memory vs UMA. lets look at what the last few generations of consoles were:
Saturn: segmented
PlayStation: segmented
Nintendo 64: UMA
M2 (unreleased) UMA
Dreamcast: segmented
PlayStation2: segmented/UMA
GameCube: segmented/UMA
Xbox: UMA
it's not so hard to believe that XBox 2 could be a hybrid segmented/UMA
or even segmented, as previous consoles used this approach, and some have been VERY sucessful, while strictly UMA consoles have, ahem, not been #1
Panajev2001a said:
Yes I know, that's why I said DEEP COLOR DEPTH man because that's what I meant.
At 128 bits per pixel almost the entire 5MB space alotted by DMGA is filled by a 640 screenbuffer. Forget about AA, forget about HDTV resolutions.
I don't expect XB2 to make do with just 32bpp, and I don't expect it to rely on 64bpp exclusively (after all, Nvidia's not making the GPU this time 'round!
). It'll need deep color back buffer(s), that much is clear.Burn? I just pointed out the fact that floating-point buffers do an even lousier job of fitting in DMGAs silly-sized on-chip memory. I don't see how you interpret that as trying to "burn" you, if you always see people pointing out an omission of yours as them trying to "burn" you, I bet you're considered a bit of a cranky guy by your friends.
Take it easy, alright?They may or they may not, but I'm willing to wager if they do it'll be considerably larger than 5MB.
I wonder however, would a console manufacturer be willing to sacrifice multipass rendering for AA that does not require screen buffers X times larger than the degree of AA used?The GPU could filter the AA samples on-chip before writing them to the buffer, however that would screw with multipass rendering of course... With DX9+, maybe you wouldn't need that as much anymore? It would save quite a bit of memory, maybe it is a valid trade-off. At least as an option for games that won't multipass and need all the memory they can get.
Then again, if we're not gonna multipass we might as well write straight 8bpp buffers anyway so in that case 5MB would be sufficient after all.
This is a lot of ifs however, so the chance of this happening seem almost nil to me.So color at 8 bits per channel and no alpha at all? Yay, that'll drum up enthusiasm for XB2 amongst developers alright...
*G*
Would you expect a completely new architecture for just minor shader revisions?
Riddlewire said:
<sigh> I was REALLY hoping you wouldn't say this, as this particular comment bugs me to no end. I play on computer equipment that pre-dates the Xbox and you know what? I can still play EVERY PC game out there! OMIGOSH!
Upgrading has nothing to do with anything but one's personal lust for more graphical performance. It means absolutely squat on one's ability to play the game. Heck, a bit ago I was running UT2003 on a PII 266 with Voodoo 5500. Doom 3 and HL2 will be coming back able to scale down to seemingly ALL the GeForces and Radeons.
You don't "require" squat. You buy what you WANT. If random people are shelling out $500 a year on computer equipment, they're doing it because they can afford to and want to. (And I think that it's quite silly.) My own upgrades come when I have the money, and I buy strictly along price/performance/longevity lines, and am always pleased with them. You can upgrade cheap on 2-3 year old equipment and still come out fine, too. Will I get all the sparklies? No. But I've found those don't really affect the GAME astall. (Oh, and the rig I'm on now I paid less than an Xbox worth for.)
What I think is BEST is to keep the games as divergent as possible, as the industry copies itself and seeks as many shortcuts as possible if it leads to more profit. The Xbox actually has fewer "PC-centric" games than I thought it would, but it still has an awful lot. Adding a keyboard and mouse? That would just send the convergence miles ahead, which seems like a particularly awful idea. Should the Xbox end up just being a cheap, gaming-focused PC with a couple extras you normally wouldn't find there?
Silliness. You can play most games just fine without hideous investments on the PC-side, so why seek it out in consoles form?
Riddlewire
Regular
cthellis42 said:
Yes.
You clearly didn't read my post. Otherwise, you would've noticed that I answered this question with my last statement.
on a related note
if ATI Team Marlborough is working on R500 (and the xbox custom chip)
does this mean the former ArtX is back with Nintendo for N5?
since in an interview with ATI they did mention separate teams working on both
and isn't it a bit underwhelming, since R500 is expected in 2005 after all, and xbox2 is either end of 2005/early 2006, so the situation is not unlike xbox1 where NV2a is a bit better than the current available NV20 (at least in terms of features, if not speed)
shouldn't xbox2 cjip at least be based on R550 if not a close-to-R600 (like NV2a is a close-to-NV25)
http://www.eetimes.com/sys/news/OEG20030815S0012
-aneep-
if ATI Team Marlborough is working on R500 (and the xbox custom chip)
does this mean the former ArtX is back with Nintendo for N5?
since in an interview with ATI they did mention separate teams working on both
and isn't it a bit underwhelming, since R500 is expected in 2005 after all, and xbox2 is either end of 2005/early 2006, so the situation is not unlike xbox1 where NV2a is a bit better than the current available NV20 (at least in terms of features, if not speed)
shouldn't xbox2 cjip at least be based on R550 if not a close-to-R600 (like NV2a is a close-to-NV25)
http://www.eetimes.com/sys/news/OEG20030815S0012
-aneep-
DaveBaumann said:
Yes. Does Intel come up with a new instruction set (IA-32 1.0, 2.0, 3.0, 4.0?) for each new chip architecture. No. Once you have a turing complete language, there's not much else to add, except for macros.
We've gone through multiple 3D architectures just to increase performance. I think people need to get away from this idea that you will continually see new "features" add to pixel shader languages.
The whole point of having a general purpose language is so that you don't need to keep extending it.
I ask you, what do you think they can possibly add to PS3.0 that is significant and that isn't already there? PS3.0 already has just about anything any CPU has, except for the capability to read/write arbitrary memory (it's still stream based and stateless). Frame buffer reads have been considered, and then disallowed, for example.
I simply don't see what can be further added to PS3.0 that justifies a "4.0" moniker.
Any company implementing PS3.0 already has their hands full trying to make it run super fast with dynamic branching and partial derivatives.
Now, if you want to talk about something like a programmable primitive processor, that's a different issue, and is not VS3.0/PS3.0, but more like PPS 1.0
I'm a DirectX beta tester and there is nothing being discussed on the internal mailing lists/newsgroups about PS4.0.
aneep said:
Why "should" it be any of these things? Just because the current XGPU is sort of a NV22.5 doesn't mean XB2's GPU will follow a similar pattern.
Who says it's even going to be based on R5xx to begin with btw? That's as far as I can tell just wishful thinking at this moment!
*G*
maybe because the economics of consoles and pc are different?
i thought consoles should always outperforms pc at launch?
say xbox2 comes out early 2006, and by that time R500 was available for 6 months already, where's the allure ? (other than DX9/DX10 games at launch, and by that time i'd expect there are significant amount of DX9 games on pc already)
-aneep-
i thought consoles should always outperforms pc at launch?
say xbox2 comes out early 2006, and by that time R500 was available for 6 months already, where's the allure ? (other than DX9/DX10 games at launch, and by that time i'd expect there are significant amount of DX9 games on pc already)
-aneep-
