CELL PROCESSOR "locked SPE" Question.....

Status
Not open for further replies.
(other than, with every game being patched, dedicating this SPE to GWAA for mich improved IQ on a lot fo games).

Not even that. GWAA runs as well as it does without creating additional lag because a relatively small portion of 4-5 (forgot the exact number) SPEs can be dedicated to it, bringing back the overhead in the processing pipeline to a low enough level that it doesn't cost you additional frames of lag. Adding just one SPE to that which would be dedicated to this purpose would be comparitively useless. Instead, I think there are currently NO games out there that couldn't handle the extra load on their existing SPEs just fine. At best it would take a little bit of additional scheduling and testing. At worst, they may need to do a little more work to 'de-optimise' their current AA solution?

Anyway, that discussion belongs to the AA thread. But the point stands, the answer remains NO to all questions (in fact, the answer remains NO to an increasing number of questions ;) ).
 
I believe unless I am mistaken that Sony's top engineers highly desired to launch PS3 at 65nm to reach frequency goals and help keep thermals in check but realisticly 65nm was more than two years away in 2005 and even though 90nm was there there was no way to avoid the competition launching early like it did...

They certainly had targeted 65nm, but that's not mutually exclusive with having wanted to launch 'early.' The fact is that getting to 65nm simply took significantly longer than anyone expected when the plans were initially laid, and 90nm itself turned out to be a problematic node. (Not just for STI, but everyone) Cell was conceived prior to the 'thermal wall' that the industry collectively hit with Prescott and 90nm, and so reality imposed some new constraints on the ideas of years prior.

And as Arwin mentioned in the EE+GS thread, mass fabrication of diodes to BD spec were woefully behind schedule.
 
Not even that. GWAA runs as well as it does without creating additional lag because a relatively small portion of 4-5 (forgot the exact number) SPEs can be dedicated to it, bringing back the overhead in the processing pipeline to a low enough level that it doesn't cost you additional frames of lag. Adding just one SPE to that which would be dedicated to this purpose would be comparitively useless. Instead, I think there are currently NO games out there that couldn't handle the extra load on their existing SPEs just fine. At best it would take a little bit of additional scheduling and testing. At worst, they may need to do a little more work to 'de-optimise' their current AA solution?

Anyway, that discussion belongs to the AA thread. But the point stands, the answer remains NO to all questions (in fact, the answer remains NO to an increasing number of questions ;) ).

What is GWAA?
 
I apologize up front for contributing to this thread, but while the 8th SPU is definitely an ex-SPU, the 7th SPU is not.

I wonder how busy GameOS keeps that SPU? I expect it's involved in doing things like DTS-MA encoding or what-have-you, but I don't know how much oomph is left in that SPU for whatever purpose Sony might have in mind.

They can't give back the dead SPU, but perhaps they could donate some of the reserved one.?

Yech. I feel dirty, now.
 
I apologize up front for contributing to this thread, but while the 8th SPU is definitely an ex-SPU, the 7th SPU is not.

I wonder how busy GameOS keeps that SPU? I expect it's involved in doing things like DTS-MA encoding or what-have-you, but I don't know how much oomph is left in that SPU for whatever purpose Sony might have in mind.

They can't give back the dead SPU, but perhaps they could donate some of the reserved one.?

Yech. I feel dirty, now.

The problem I see with that is that you already have a cpu and 6 spus, that seems like plenty as it is, you really need to find the making of killzone 2 subtitled video its 43 some minutes long and very detailed and the claim that they make is 60% spu usage but that was over a year ago, by february of 2011 we will see what GG was able to do, its alot sooner than waiting for Uncharted 3.

Other than that if you are just a power monger (I'm guilty of being one) lets invent a time machine to delay the PS3 to 2009, that way Sony would have had a Power XCell with 2CPUs and 16 SPUs!! and a 8xBRdrive ;)
 
It is not possible to use the 8th SPU, but what about increase Ghz of the SPUs?

You can enable it in the new PS3 systems only. This systems then get some form of performance boost while not changing the old non-capable systems...of course devs have to develop with the old specs in mind...
 
I'm a complete tech noob, pretty much...

But I can almost guarantee on the little knowledge that I do have,
That bringing in another SPE is more than the 2% you've estimated...especially since parallel is in the mix here.

Again....me...noob,

But I bet it's in the neighborhood of 10 to 15 percent...give or take 5.

The rest makes sense though.

No...just...No. :p

Even if it could be enabled. Even if it could then be used. Unless your entire system is SPU bound and everything else is sitting around idling while waiting you are not going to see an increase of even 10%.

Your performance bottleneck in any given game is going to be constantly shifting around depending on what is actually being stressed at any given time.

As far as unlocking,

Couldn't it be done "situationaly"....say, if NAUGHTY DOG wanted the extra boost to get UNCHARTED 3 at 60 frames (again hypothetically), couldn't they create code to run a check to see if all 8 SPE's are working, and if they are, then Uncharted 3 will run at 60 frames....if not, then the game reverts back to 30 frames. Something of a "toggle on...toggle off...sorta.

Again, no... Using SPUs effectively is more involved that using a normal CPU core which in itself isn't exactly easy when you have multiple cores.

SPUs have to be coded for specifically in order to make use of them. So if (again we're talking about something that will never ever happen) Sony were to enable the 8th core, to use it you'd have to code specifically for it.

You'd then need 2 branches of code. 1 for your regular PS3's and another for the PS3+1. Considering the size of developement budgets and length of developement as it is currently for PS3, I don't see any devs jumping around for joy if they suddenly felt the need to implement 2 seperate code paths... Especially when any gains would most likely be fairly negligible.

Regards,
SB
 
Status
Not open for further replies.
Back
Top