How is having a incredibly powerfull VPU with a simple Rasterizer that can do filter operations insanely fast coupled with the Rasterizer having insane floating point performance so you can run any type of shader function you want bad? This is the world in which we are going into. One where effects such as Pixel shaders aren't locked into hardware, but where there is enough power on the front end to run or do any effect you could ever want to do.
Kutaragi Ken:Want a teraflop? You have to buy a rack from us
- Thread starter DeadmeatGA
- Start date
How is having a incredibly powerfull VPU with a simple Rasterizer that can do filter operations insanely fast coupled with the Rasterizer having insane floating point performance so you can run any type of shader function you want bad? This is the world in which we are going into. One where effects such as Pixel shaders aren't locked into hardware, but where there is enough power on the front end to run or do any effect you could ever want to do.
Yes, like i said, a strangely major oversight for KK if he is using the same old blue gene slide for Cell. 
As for your incredibly power VPU thingie, we shall see. Thats the way 3D game might be going, but it be interesting to see who does the graphics better. ie. SCEI aint the only ones looking to advance realtime 3D. Remember people, Sony != World.
PS3 1TFLOPS >< GFLOPS <>>< HOPS.
As for your incredibly power VPU thingie, we shall see. Thats the way 3D game might be going, but it be interesting to see who does the graphics better. ie. SCEI aint the only ones looking to advance realtime 3D. Remember people, Sony != World.
PS3 1TFLOPS >< GFLOPS <>>< HOPS.
Hmm...2 chips(64 gigaflops) would be about the power I'd expect from ps3's cpu, if there wasn't all this talk about 256 gigaflops or 1 teraflop. (ignoring everything I've heard, I'd say ps3's cpu would do between 64-128 gigaflops) Then with the graphics chip, I'd say it could get up to half a teraflop in power.
How can you do it better.
A insanely fast Rasterizer doing filter operations that can SUSTAIN 128GFLOPS for Shader operations would do the job the same as one of relatively the same power. How will one be "doing graphics better?"
There are going to be no IQ problems for any of the next gen systems either, hardware based functions will be relatively the same for next gen GPU's. These are your standard things like AA and compression.
This thread is a joke. I don't even have to comment, just save this pure comedy to a file and wait a year for my bud to be "correct once again." heh.
Put this in perspective, you're reading an article by The Register quoting a Kutaragi presentation that has a BlueGene slide in it... that's being hailed as frickin' inspired text by DM and Chappers (both with the anti-Cell/Sony sigs - kinda cute). Hell, hot damn, if that doesn't put a stamp of approval on it, what would?
Put this in perspective, you're reading an article by The Register quoting a Kutaragi presentation that has a BlueGene slide in it... that's being hailed as frickin' inspired text by DM and Chappers (both with the anti-Cell/Sony sigs - kinda cute). Hell, hot damn, if that doesn't put a stamp of approval on it, what would?
Panajev2001a
Veteran
Let's use Deadmeat(tm) Fuzzy Math(r), just from a different, but still valid angle.
The name cell is our dummy variable ( no pun intended, it is a mathematical term ) so we can substitute it to whatever we want, right ?
64 cells = 2 TFLOPS is our equation...
Let's set cell equal to APU:
64 APUs = 2 TFLOPS
Suzuoki's patent said 32 APUs = 1 TFLOPS.
Let's multiply Suzuoki's equation by 2 on both sides:
32 APUs * 2 = 1 TFLOPS * 2
This becomes:
64 APUs = 2 TFLOPS
Q.E.D.
The name cell is our dummy variable ( no pun intended, it is a mathematical term ) so we can substitute it to whatever we want, right ?
64 cells = 2 TFLOPS is our equation...
Let's set cell equal to APU:
64 APUs = 2 TFLOPS
Suzuoki's patent said 32 APUs = 1 TFLOPS.
Let's multiply Suzuoki's equation by 2 on both sides:
32 APUs * 2 = 1 TFLOPS * 2
This becomes:
64 APUs = 2 TFLOPS
Q.E.D.
Paul said:
You didn't even read the rest of the reply, or; ignored it. Typical.
You obviously didnt understand mine for the better. Better, brudder better. The key word. Not to forget Sony != World. Unless, realtime 3D will reach its limits next gen(
), i have doubts that IQ will be equal on all counts. Typical.
Noone understands you, because you don't know practical English. You'll get out of High School soon enough though... or maybe not.
Anyway, let's try and read what you wrote here.
If 3D is going towards the GPU having massive floating point power to run shader programs then why are you saying who will do the graphics better? Don't you mean... Who will have the higher Floating point power on their GPU, Sony or MS?
Like I said, you didn't read the REST of the reply, I know reading isn't something you like doing(at all it seems) so just try and keep up with the Men around here.
This is what I said.
That's right, how can one do the graphics better when both would have around the same floating point power in my scenario?
A insanely fast Rasterizer doing filter operations that can SUSTAIN 128GFLOPS for Shader operations would do the job the same as one of relatively the same power. How will one be "doing graphics better?"
You didn't read this? You didn't, you reacted; not unlike most times.[/quote]
Panajev2001a
Veteran
AzBat said:
BlueGene and the architecture we see in Suzuoki's patent and IBM patents that cover the PE and the APU and a SMP configuration made of diffetrent PEs are quite different.
The Version Deadmeat talks about uses two CPUs for each node: a worker core and a I/O processing core that ran the OS and controlled the message passing between nodes.
The other BlueGene processor ( the one IBM likes the most ) used to have 1-to-8 Thread Units per node and in that node we find an FPU and an arbiter which also regulates access to that FPU from the Thread Units in that node.
Each node comes with SRAM based cache and 0.5 MB of e-DRAM.
32 of those nodes form one core and you can pack multiple cores on the same chip and multiple chips on the same board.
Ideas from BlueGene Cellular architecture influenced CELL, but IBM, Toshiba and SCE started from practically scratch with CELL and the 5 IBM guys who directed the early work on CELL's Architecture and ISA did not take BlueGene and added FPUs to it.
The point of CELL is that all processors share the same APU ISA to be able to have Apulets go from CELL device to CELL device, migrating over local busses, LANs across networks.
CELL CPUs not using APUs are not CELL CPUs.
The problem is, will ~128GFLOPS be enough to do all that nifty software rendering dooda, and still present a smooth next gen graphics? I wont know, but 128GFLOPS is like just ~6X more than cards today...Will that give you, your full FFX CG realtime, 60fps, AA/AF, 1080i....Do you know?
Hmm...i recall Pana saying that, even with 1TFLOPS, still be better to do stuff like AA/IQ in hardware..
Finally, last i heard, FLOPS ratings are quite the quirky number game...
Hmm...i recall Pana saying that, even with 1TFLOPS, still be better to do stuff like AA/IQ in hardware..
Finally, last i heard, FLOPS ratings are quite the quirky number game...
Chap listen.
The 128GFLOPS thing ISNT literal, it is a made up situation.
I'm comparing two GPU's of 128GFLOPS peak floating point power, made up GPU's.
Basic functions like AA are always done in hardware, this I even said. It's the shaders or "effects" that are done with the Floating point power, why lock things in hardware on the chip when you can program any effect you want with massive front end floating point power.
The 128GFLOPS thing ISNT literal, it is a made up situation.
I'm comparing two GPU's of 128GFLOPS peak floating point power, made up GPU's.
Basic functions like AA are always done in hardware, this I even said. It's the shaders or "effects" that are done with the Floating point power, why lock things in hardware on the chip when you can program any effect you want with massive front end floating point power.
Panajev2001a said:
I understand that Sony Cell and IBM Cell are not the same, but you still haven't gave me any comparisons on how they different. Just remember I didn't understand any of the technical mumbo-jumbo you just posted. I just want to be able to figure out what FLOPS, cores and APUs the Sony version will have and how it relates to these slides. Is it as DM says and that Sony won't release a 2 Tera FLOP capable part for PS3 or is it that Sony's PS3 is more capable than even IBM's Blue Gene? I'm totally confused.
Tommy McClain
Paul,
firstly, how well qualified are you with 3D hardware systems. be honest, do you really understand the full going-ons, or just playing the basic cut + paste job. it might sound rude(sorry), but being a n00b that im, i dont want be wasting time and not learning anything.
How did you come to the conclusion that performance be the same...? Sony Optimism?
firstly, how well qualified are you with 3D hardware systems. be honest, do you really understand the full going-ons, or just playing the basic cut + paste job. it might sound rude(sorry), but being a n00b that im, i dont want be wasting time and not learning anything.
yes. but how do you know they are going to be the same? even today, we have all those quirky MSAA/FSAA/Quicunx/OGASS/something whacawhaca. Even then, AA performs better on Ati R300 cards vs the same gen rivals NV30.. :?
How did you come to the conclusion that performance be the same...? Sony Optimism?
Because it might not be powerful enough?
Similar threads
