upnorthsox
Veteran
No I don't understand, do you have a picture of this? AFAIK Toshiba didn't make a Blu-ray drive until 2009.
NGGP: NextGen Garbage Pile (aka: No one reads the topics or stays on topic) *spawn*
- Thread starter Barso
- Start date
- Status
No I don't understand, do you have a picture of this? AFAIK Toshiba didn't make a Blu-ray drive until 2009.
bgassassin
Regular
I was coming back in regards to this as I had some extra info. This does after all seem to be the case for Xbox 3's CPU. I'm not sure on the exact details, but from a FLOPs perspective it seems it may have as much as double the performance of "vanilla" Jaguar cores.
Interesting, really interesting.
Aeoniss could you point to the whole post? It goes against the data from VGleaks, it smells fishy.
"I need to dedicate some time to having a good look at the specs i never realised there was a page 2 and 3 of the vgleaks article regarding the gpu. And a cursory glance tells me this is a very highly customised chip.
People do understand this is not 12 vs 18 cu's right? Its 12 vs 14.
They also understand its read write/rate per per pixel is roughly twice that of the PS4 gpu?
The issue i have is, there isn't a whole heck of a lot of data about the PS4 gpu, so we have to assume it's a standard chip. We also dont know enough about gcn engines."
It's coming from an indie dev on the forum. He knows his stuff and he's made some games, he's from New Zealand. Insofar as I'm aware he isn't signed on with any big development company.
He's only interpreting what information has been leaked to us so far, so what all of us who aren't tied by NDA's know.
I just don't know what information he's interpreting to reach the aforementioned conclusion.
Edit: He just provided context
"
PS4
Dual Shader Engines:
- 1.6 billion triangles/s, 1.6 billion vertices/s
720
Triangle rate
2 tri/clock * 800 MHz = 1.6 Gtri/sec
Vertex rate
2 vert/clock * 800 MHz = 1.6 Gvert/sec
You cannot push more vertices through than the 720.
Both the DB and the CB have substantial caches on die, and all depth and color operations are performed locally in the caches. Access to these caches is faster than access to ESRAM. For this reason, the peak GPU pixel rate can be larger than what raw memory throughput would indicate. The caches are not large enough, however, to fit entire render targets. Therefore, rendering that is localized to a particular area of the screen is more efficient than scattered rendering.
Must be said, 99% of what you render at any one time is in the one spot.
Rain would be the obvious exclusion to that.
Colour buffer
The CB is optimized for 64-bit-per-pixel types, so there is no local performance advantage in using smaller color formats, although there may still be a substantial bandwidth savings.
PS4
32 color ops/cycle
Now, again, my interpretation of this would be that the ROP is 64 bit per pixel on the 720 and only 32 bits per pixel on the PS4."
People do understand this is not 12 vs 18 cu's right? Its 12 vs 14.
They also understand its read write/rate per per pixel is roughly twice that of the PS4 gpu?
The issue i have is, there isn't a whole heck of a lot of data about the PS4 gpu, so we have to assume it's a standard chip. We also dont know enough about gcn engines."
It's coming from an indie dev on the forum. He knows his stuff and he's made some games, he's from New Zealand. Insofar as I'm aware he isn't signed on with any big development company.
He's only interpreting what information has been leaked to us so far, so what all of us who aren't tied by NDA's know.
I just don't know what information he's interpreting to reach the aforementioned conclusion.
Edit: He just provided context
"
PS4
Dual Shader Engines:
- 1.6 billion triangles/s, 1.6 billion vertices/s
720
Triangle rate
2 tri/clock * 800 MHz = 1.6 Gtri/sec
Vertex rate
2 vert/clock * 800 MHz = 1.6 Gvert/sec
You cannot push more vertices through than the 720.
Both the DB and the CB have substantial caches on die, and all depth and color operations are performed locally in the caches. Access to these caches is faster than access to ESRAM. For this reason, the peak GPU pixel rate can be larger than what raw memory throughput would indicate. The caches are not large enough, however, to fit entire render targets. Therefore, rendering that is localized to a particular area of the screen is more efficient than scattered rendering.
Must be said, 99% of what you render at any one time is in the one spot.
Rain would be the obvious exclusion to that.
Colour buffer
The CB is optimized for 64-bit-per-pixel types, so there is no local performance advantage in using smaller color formats, although there may still be a substantial bandwidth savings.
PS4
32 color ops/cycle
Now, again, my interpretation of this would be that the ROP is 64 bit per pixel on the 720 and only 32 bits per pixel on the PS4."
Averagejoe
Regular
No matter what people try to pretend is not 12CU vs 14 CU,the PS4 has 4 CU on the compute side that i am sure sony did not put there just because they wanted to fill the die space with something.
Every times something is shut down another magical theory appears on how Durango will be on par or even ahead is tiresome.
Is he trying to say that Durango has half the ROPs yet the same bandwidth as Orbis (going with the additive split pools meme: DDR3 + EDRAM ~= 170GB/s), hence "twice" the "read/write rate per pixel?" It still doesn't make sense to me, but what do I know.
"Our new car model has manual windows so that you can escape when you drive into a large body of water!"
No matter what people try to pretend is not 12CU vs 14 CU,the PS4 has 4 CU on the compute side that i am sure sony did not put there just because they wanted to fill the die space with something.
Every times something is shut down another magical theory appears on how Durango will be on par or even ahead is tiresome.
Assume Sony built a closed box system optimized for a GPU with 14CUs, 8 jaguar cores, and 2GB of GDDR5. The feedback from developers is then that Durango has 8GB of memory, a "custom" jaguar CPU with roughly twice the compute power, and a custom GPU which is at least comparable to Orbis'. What would be the easiest way to close the performance gap?
Wouldn't upping the memory to 4GB and add CUs to the GPU be one of the most obvious answers? (The fact that the extra 4 CUs weren't planned in the initial system design phases could explain why they're said to only provide limited benefit in the rendering pipeline)
Averagejoe
Regular
Some times it surprise me because some people actually want to put Durango as the mother of all efficient systems ever created,one that rival the 680GTX GPU based on silly theories about efficiency.
But some how Orbis the more straight forward design,with faster ram and unified isn't as efficient,it has ROP over kill,and the 4 CU it has for compute would not do little difference if any.
Arguments like this are tiresome,duango hardware has been heavily tied to secret sauce ever since it was hint that it wasn't as powerful as orbis.
Averagejoe
Regular
Actually that is kind f different,but the manual windows car will cost you less than the power windows version,the car will take you to the same places the power window one will.
Some times it surprise me because some people actually want to put Durango as the mother of all efficient systems ever created,one that rival the 680GTX GPU based on silly theories about efficiency.
But some how Orbis the more straight forward design,with faster ram and unified isn't as efficient,it has ROP over kill,and the 4 CU it has for compute would not do little difference if any.
Arguments like this are tiresome,duango hardware has been heavily tied to secret sauce ever since it was hint that it wasn't as powerful as orbis.
There's two reasons for this IMO.
1. MS already did this last gen with a system that seemingly on paper was overmatched and then, as i think Shifty put it, "merrily kicked PS3 in the nether regions on a number of multiplatform titles." I.E. there's a track record here of efficiency, performance, and "bang for your buck" from the MS engineering team so they deserve some benefit of doubt
2. There has been a steady stream of leaks about custom (and not so custom) pieces of this system that clearly allude to the fact that MS has put some significant thought into getting the highest possible performance with the budget they've been given. There has been no such information that Orbis is anything but exactly what's on paper. By the way, thats not a bad thing, its just the case right now.
Also, this is a technical forum and i think subconsciously the crowd here wants to uncover the fabled "secret sauce" and a bold, cutting edge design, under the covers that provides more performance than could have been expected from the given transistor budget
Averagejoe
Regular
That wasn't case from what i read,the 4CU were part of the plan the 4GB of memory came from pressure from developers,the 4 CU apart did not.
Either way from what i read you have 12 CU on Durango vs 14 CU on orbis.
A CPU with 102 gflops on Orbis and one with double the flops in Durango (if true,it could be the new secret sauce) so 204 Flops on the CPU side for Durango.
That still leaves 4CU with what 400Gflops on the compute side that still are there to be use.
Ok, but doesn't your analysis here also totally ignore other hardware MS is surrounding the GPU with? Can any of those extra parts (eSRAM, display planes, DME's) make up for the 2 CU difference or 200 Gflops gap simply by being significantly more efficient?
Lucid_Dreamer
Veteran
I still can't pass DD+, TrueHD and DTS-HD MA to my A/V receiver for decoding on my 60GB PS3. You can't do it on the 80GB or the 40GB PS3s either. I can do it on the Slim and Super Slim PS3s, though. I wonder why that is. According to statements about, that didn't happen.
- 60GB to 80GB removed one hardware feature (GS) and added another (20GB of storage).
- Later models dropped EE as well and added bigger HDD
- bitstreaming was added later with even larger HDDs, later.
and this is before taking into consideration that the OS *may* (if plans haven't changed- referring to yukon leak) operate on it's own exclusive CPU+GPU arrangement, as well as all the audio work being piped through the SoC/dedicated DSP.
Saving 5-10% of the system resources and adding 5-10% more puts the two systems in roughly the same ballpark before you even start talking about efficiency.
Averagejoe
Regular
MS did not do this with the xbox 360.
People apparently forget that the 360 was ready before the PS3,the xbox 360 had a stronger GPU and more advance than the weak and starved PS3 one.
Cell was the real difference in the battle,because it helped the GPU get closer to its peak,by offloading task from it,in this case scenario not only Orbis has the more powerful GPU,it also has extra help for physics,animations and several other things once again offloading from the GPU,the 4CU.
The whole 2TF performance claim by Sony was a joke,but MS started that war when they claimed the 360 was a 1 TF machine on may 12 2005 on the MTV unveil,the PS3 spec where unknown by then.
http://www.digitaltrends.com/features/xbox-360-unveiled/
What MS is doing is not making systems that is ultra efficient,is just patching that system that from the memory side is bottle neck,if they don't use ESRAM the console will suffer allot,this look to me like what they did with DVD vs Blu-ray,that they just rely on extreme compression hoping that the Data will not step into a second DVD,whith the results we already know,at first it was ok,but as the generation went on more and more games need it a 2 and even 3 disc.
They built a console that was a contemporary of the PS3, performed as well as (and outperformed in some cases) the PS3, and was cheaper to manufacture over its lifecycle. At the time, most people including Sony believed the PS3 was far more powerful. Thats the precedent i'm talking about to which MS deserves a nod this go 'round.
To anyone, what is the difference between 100 GLOPS on the CPU and 100 GFLOPS on a GPU? are they 'equal' in terms of accessibility, flexibility, 'power', etc?
- Status