spiroyster wrote:

vladstamate wrote:

...the GPU that Sony was about to use before...

The suspense is killing me... what was it?

It was supposed to be a Sony + Toshiba based GPU. I started working at Sony on PS3 right before people realized that GPU won't cut it, so I got to work on it a bit. It was their own design. And it made sense since Sony + Toshiba were already doing the Cell together (the ST and STI). Nvidia was more expensive but it brought a good, performant, GPU. With very few HW bugs.

vladstamate wrote:

spiroyster wrote:

vladstamate wrote:

...the GPU that Sony was about to use before...

The suspense is killing me... what was it?

It was supposed to be a Sony + Toshiba based GPU. I started working at Sony on PS3 right before people realized that GPU won't cut it, so I got to work on it a bit. It was their own design. And it made sense since Sony + Toshiba were already doing the Cell together (the ST and STI). Nvidia was more expensive but it brought a good, performant, GPU. With very few HW bugs.

The PS3 narrowly escaped being a total clusterfuck of a hardware I guess.

appiah4 wrote:

The PS3 narrowly escaped being a total clusterfuck of a hardware I guess.

Hey Cell was/is great! 😀

Think about it. It had 8 separate SPUs each running at 3.2Ghz doing (sometimes) dual issue instruction execution. On top of its 2 core main PowerPC CPU. On the SPUs you could do a matrix-vector multiplication in a handful of cycles. Did I mention it was at 3.2Ghz? And the instructions were all in local cache for the SPUs. With careful crafted code you could beat the GPUs vertex processing. Memory bandwidth was an issue though.

vladstamate wrote:

appiah4 wrote:

The PS3 narrowly escaped being a total clusterfuck of a hardware I guess.

Hey Cell was/is great! 😀

Think about it. It had 8 separate SPUs each running at 3.2Ghz doing (sometimes) dual issue instruction execution. On top of its 2 core main PowerPC CPU. On the SPUs you could do a matrix-vector multiplication in a handful of cycles. Did I mention it was at 3.2Ghz? And the instructions were all in local cache for the SPUs. With careful crafted code you could beat the GPUs vertex processing. Memory bandwidth was an issue though.

Don't listen to him vlad, he knows not of which he speaks. Ironically PS3's were so powerful they dented PC's reign in folding@home and SETI projects at the time and there were some beowulfs set up with them. So yeah, the phrase 'clusterfuck' is more appropriate than he realises. o.0

Thanks for sharing, I'm a PS3 fan, and thus a fan of your work... a few years ago the PS3 brutally knocked the Dreamcast of the top spot as my favourite console. PS2 took a lot of my time, but didn't overall impress me as it graphically just felt like an overclocked, more mem Dreamcast (and early PS2 regurgitated a lot of Dreamcast games with simply slightly more content). The PS3 however blew me away, the jump from PS2 -> PS3 was I felt a greater jump than previous generational leaps (perhaps this was the gfx/gpu scene overall around then?). The xbox 360 did well to keep up with the PS3 visually, and for its efficiency with limited CPU power in relation should be commended... but sticking the processing power of the PS3 in that box, for that price in 2006...well, I multiply doth my cap to you. 😀

Scali wrote:

Yea, the history of OpenGL is interesting... also within Microsoft itself. I recommend you read this blog:
http://web.archive.org/web/20170612154308/htt … on-of-direct3d/

Very interesting, thanks. I'll return the favour with this interesting video I watched few months ago about the 3dfx guys. If you have 2.5 hours to spare at any point that is o.0
https://www.youtube.com/watch?v=3MghYhf-GhU

vladstamate wrote:

appiah4 wrote:

The PS3 narrowly escaped being a total clusterfuck of a hardware I guess.

Hey Cell was/is great! 😀

Think about it. It had 8 separate SPUs each running at 3.2Ghz doing (sometimes) dual issue instruction execution. On top of its 2 core main PowerPC CPU. On the SPUs you could do a matrix-vector multiplication in a handful of cycles. Did I mention it was at 3.2Ghz? And the instructions were all in local cache for the SPUs. With careful crafted code you could beat the GPUs vertex processing. Memory bandwidth was an issue though.

It was certainly technically very interesting, but was also a very bad choice for a device at a time when developers to that device were already trying to adjust to a spike in development time and costs..

spiroyster wrote:

Don't listen to him vlad, he knows not of which he speaks. Ironically PS3's were so powerful they dented PC's reign in folding@home and SETI projects at the time and there were some beowulfs set up with them. So yeah, the phrase 'clusterfuck' is more appropriate than he realises. o.0

Judging by this post I should fit right in this thread. I'm sure millions of PS3 owners found relief and consolation in the fact that their gaming hardware could cure cancer and find aliens but not manage to output native HD resolution games during the first half of its lifetime.