PSP's CPU not 333Mhz, only 222 ....suxx

[Teasy]

How does this relate to the performance figures given for PSP? In terms of polys and Pixels per second is a 222MHz system providing 66% of these specs, or is a 222MHz system = 100% these specs and a boost to 333 MHz will give a 50% performance boost?

Definitely the first one, no way would Sony be conservative in the specs department.

 

[Megadrive1988]

Look What Smeagol Finds

from November 18

333 Mhz is a lie (Score:5, Informative)
by Anonymous Coward on Thursday November 18, @01:38PM (#10856599)
Sony continues to brag about the 333 Mhz processor in the system, but it's a lie. While the processor may technically be *capable* of running that fast, it's in fact been throttled to 222 Mhz. Developers have the option of running it even slower, but there is no way to run it faster.

http://games.slashdot.org/article.pl?sid=04/11/17/2247229&tid=207

 

[Megadrive1988]

also, more recently:

http://hardware.gamespot.com/Story-ST-x-1480-x-x-x&body_pagenum=3

Software developers tell us that 222MHz is the current magic number for the CPU.

http://games.slashdot.org/article.pl?sid=05/03/06/1438248&tid=207&tid=212

I find it interesting that the processor speed isn't listed. From what I've heard, even though it runs at 333MHz it has been software locked at 222MHz to help battery life. Can anyone confirm this?

my question is, why does the 222 MHz clock seem like a surprise?

 

[ANova]

The PSP's processor has a feature akin to Intel's speedstep which downclocks the speed depending on its use to save battery power. I'd imagine that 222 MHz is just the default speed to keep battery life at a premium while 333 MHz is surely an option for those that need it.

What they are saying is that a later revision may have an updated core using the 65nm process to reduce power consumption and thus allow for a 333 MHz default speed while maintaining or gaining battery life.

 

[cybamerc]

This is despicable. Not only is it borderline false advertising to state that the system runs at up to 333 MHz but it also put early adopters at a very disadvantageous position if Sony decides at some point to allow developers to take full advantage of the system. Battery life is poor enough as it is.

Sony deserves nothing but contempt.

 

[Shifty Geezer]

They never said it runs at 333 MHz. Sony said the processor speed was scalable. I think every complaint of Sony 'lying' about it's PlayStation systems' specs is totally misplaced. Seems to me people misinterpret the facts provided.

As for early adopters being at a disadvantage, the report suggests it's a matter of battery conservation - not performance limitations. Future PSP releases (Sony have talked of a redesign for example) can be sold as new and improved, same as GBASP, and provide better battery performance on 333 MHz games, while a more powerful battery pack/fuel cell pack for older PSPs will enable more than 2 hours 333 MHz performance. But it's unlikely there will be a situation where early tech won't run a product.

Early adopters who complain later iterations of a technology is better/cheaper should be slapped silly! Of course it's better - it's come out later and technology progresses!

 

[passerby]

I like to do this:

Early Adopters.

(Unfortunately my turn may come soon, as I battle the temptation to import a PS3 launch unit... )

For the record, at least for Japanese marketing and PR, HW specs were hardly ever mentioned - almost never. All they talk about is the screen and 'playstation quality graphics'.

 

[one]

When GBA2 arrives, a new firmware update will be available for the PSP to enable true PS2-level graphics

 

[Shogmaster]

The PSP's processor has a feature akin to Intel's speedstep which downclocks the speed depending on its use to save battery power. I'd imagine that 222 MHz is just the default speed to keep battery life at a premium while 333 MHz is surely an option for those that need it.

If that's true, why is there no option to raise the bus to 166Mhz nor the GPU to operate @ 166Mhz?

What they are saying is that a later revision may have an updated core using the 65nm process to reduce power consumption and thus allow for a 333 MHz default speed while maintaining or gaining battery life.

Either way, since these early revs are locked to 111Mhz bus and GPU clock and 222Mhz CPU speed, effectively, what they do later doesn't matter much unless these are mere firmware locks.

BTW, I guess my source was pretty much right all along.

 

[V3]

The PSP's processor has a feature akin to Intel's speedstep which downclocks the speed depending on its use to save battery power. I'd imagine that 222 MHz is just the default speed to keep battery life at a premium while 333 MHz is surely an option for those that need it.

If that's true, why is there no option to raise the bus to 166Mhz nor the GPU to operate @ 166Mhz?

There is. The whole thing is on one chip.

 

[Shogmaster]

The PSP's processor has a feature akin to Intel's speedstep which downclocks the speed depending on its use to save battery power. I'd imagine that 222 MHz is just the default speed to keep battery life at a premium while 333 MHz is surely an option for those that need it.

If that's true, why is there no option to raise the bus to 166Mhz nor the GPU to operate @ 166Mhz?

There is. The whole thing is on one chip.

According to Faf and Pana, it's restricted by the Sony library and cannot be done.

 

[aaaaa00]

 

[V3]

The PSP's processor has a feature akin to Intel's speedstep which downclocks the speed depending on its use to save battery power. I'd imagine that 222 MHz is just the default speed to keep battery life at a premium while 333 MHz is surely an option for those that need it.

If that's true, why is there no option to raise the bus to 166Mhz nor the GPU to operate @ 166Mhz?

There is. The whole thing is on one chip.

According to Faf and Pana, it's restricted by the Sony library and cannot be done.

To developers it cannot, but Sony themself shouldn't be a problem. If they choose to do so at later date.

 

[Panajev2001a]

Shog and aaaaa00,

apparently (the word comes from people attending GDC and other developers) there is a typo in those slides and the GPU can be raised to 166 MHz (I assume the bus speed can be changed to: throttled up to 166 MHz): of course there is still the library lock for now.

 

[Shogmaster]

Shog and aaaaa00,

apparently (the word comes from people attending GDC and other developers) there is a typo in those slides and the GPU can be raised to 166 MHz (I assume the bus speed can be changed to: throttled up to 166 MHz).

What, the CPU and the bus/GPU can run asynchronous? I thought it had to run @ 2X multiplier?

 

[Panajev2001a]

Shog and aaaaa00,

apparently (the word comes from people attending GDC and other developers) there is a typo in those slides and the GPU can be raised to 166 MHz (I assume the bus speed can be changed to: throttled up to 166 MHz).

What, the CPU and the bus/GPU can run asynchronous? I thought it had to run @ 2X multiplier?

They can run asynchronous it seems as the minimum speed of the CPU is 1 MHz and the minimum bus speed is 37 MHz according to those slides.

However, 333 MHz fits the 2x multiplier "rule" if the bus speed/GPU is 166 MHz.

Likely the typo was the bus specification: 37 MHz - 111 MHz should have been 37 MHz - 166 MHz.

 

[Simon F]

Anyone got a link to what's contained in the FPU and VFPU?

<from slide> 4x4 Matrix/vector multiply, 22 cycles

That strikes me as a very strange figure. A full 4x4Mat *4Element vector multiply needs 16 FP Muls and 12 FP adds. How do they get to 22 cycles? My guess is that they have two FPU units, and each has a throughput of a MUL in two clocks and add in one clock.

In any case, it implies it's doing ~1.3 MFLOPS/MHz so at ~200Mhz=>260MFLOPs. That doesn't seem like stellar performance for a handheld device.

 

[Panajev2001a]

Anyone got a link to what's contained in the FPU and VFPU?

<from slide> 4x4 Matrix/vector multiply, 22 cycles

That strikes me as a very strange figure. A full 4x4Mat *4Element vector multiply needs 16 FP Muls and 12 FP adds. How do they get to 22 cycles? My guess is that they have two FPU units, and each has a throughput of a MUL in two clocks and add in one clock.

In any case, it implies it's doing ~1.3 MFLOPS/MHz so at ~200Mhz=>260MFLOPs. That doesn't seem like stellar performance for a handheld device.

Simon, that would mean that 4x4 Matrix * 4x1 Vector does not push the VFPU to the max (there migth be dedicated dto product instructions that yield that figure): for 333 MHz operation the CPU was and still is specced at 2.6 GFLOPS.

 

[Fafalada]

That strikes me as a very strange figure.

It's not strange, it's latency. The throughput is 16cycles, just like you would expect, and you can always execute invidual DOT4/MADDs single cycle if you prefer.

The thing has a very impressive feature set though, it's not just about how fast it is (and it's quite fast, especially once you go beyond simple stuff like madds).

 

[-tkf-]

I'd imagine that 222 MHz is just the default speed to keep battery life at a premium while 333 MHz is surely an option for those that need it.

I think it's more an option for Sony, in a sense they have buildt in a 33% upgrade for "free".

And with the insane bitching over battery life i can understand why Sony decided to put a hard lock on speed. To bad they didn't trust the developers to do it themselves...