Gabe Newell on NextGen systems

[swaaye]

What exactly is your POV?

I guess my biggest question would be why they went with 3 crippled cores (very simplistic cores) instead of 1 or 2 fully featured cores that don't require multiple times the effort to extract great performance?

What exactly is their POV? LOL. I realize that PC CPUs are for general purpose tasks, but does a console not benefit from most fo the same performance tweaks?

It just doesn't seem like a decision made to benefit anyone other than the marketers. I guess we'll see soon enough. And I'm sure that developers will get some serious performance out of these CPUs anyway. It's just too bad that it had to become so much more difficult.

 

[darkblu]

What exactly is your POV?

I guess my biggest question would be why they went with 3 crippled cores (very simplistic cores) instead of 1 or 2 fully featured cores that don't require multiple times the effort to extract great performance?

What exactly is their POV? LOL. I realize that PC CPUs are for general purpose tasks, but does a console not benefit from most fo the same performance tweaks?

note exactly. desktop cpus _have_no_choice_ but to handle whatever code's been thrown at them - be that badly written code or code _not_targetted_ at them at all (e.g. amd running p4-optimised code). for them not having OOOE is not really an option these days.

console makers, OTH, can deside what they want those transistors to be used for.

It just doesn't seem like a decision made to benefit anyone other than the marketers. I guess we'll see soon enough. And I'm sure that developers will get some serious performance out of these CPUs anyway. It's just too bad that it had to become so much more difficult.

you're talking as sombody who'd actually have to code for those. is that the case?

 

[Fox5]

I thought the actual benched performance of cell was around 40Gflops. Which would be pretty impressive, since I don't think any x86 chips come anywhere close. What's the current top x86 single core performer? I'd imagine somewhere between 10Gflop to 20Gflop.

I see your numbers have already been corrected, but...

How would it have been pretty impressive? A CPU that has been designed from ground to up to offer optimum floating point performance, with 8 separate "cores", would have reached only 2x - 4x single core x86 performance?

A cpu that doesn't cost any more to produce is delivering 2-4x the performance. If this is a performance area that really matters then I'd consider that very impressive.

note exactly. desktop cpus _have_no_choice_ but to handle whatever code's been thrown at them - be that badly written code or code _not_targetted_ at them at all (e.g. amd running p4-optimised code). for them not having OOOE is not really an option these days.

Do VIA's cpus have OOOE?

 

[Carl B]

A cpu that doesn't cost any more to produce is delivering 2-4x the performance. If this is a performance area that really matters then I'd consider that very impressive.

It does cost more to produce.

Do VIA's cpus have OOOE?

Indeed they do.

 

[jvd]

A cpu that doesn't cost any more to produce is delivering 2-4x the performance. If this is a performance area that really matters then I'd consider that very impressive.

Do you have hte proof to back this up ?

As far as I can see a dual core athlon 64 chip is half the size of a cell chip on 90nm . I don't see the cell costing the same .

Not to mention we don't know how well a cell chip will perform doing normal pc tasks . So while it may be suited for graphic intensive apps , it can still be 2-4 times slower in normal pc tasks than a dual core athlon 64 or dual core p4 .

 

[mckmas8808]

But doesn't CELL really seem like it's best suited for decoding video streams and other stuff you would find in DVD players and TV's?

It just doesn't seem like the best design for video games...i mean it might kick ass, new consoles always do, but it still doesn't seem optimized/designed for a console, but more of an A/V processor.

I have to really really disagree with that. Ken didn't help design this chip with Toshiba's and IBM's help just to be a A/V processor. Believe me they designed the chip with the PS3 on their minds.

 

[Qroach]

...and why should we believe you Mckmas8808?

 

[Nite_Hawk]

A cpu that doesn't cost any more to produce is delivering 2-4x the performance. If this is a performance area that really matters then I'd consider that very impressive.

Do you have hte proof to back this up ?

As far as I can see a dual core athlon 64 chip is half the size of a cell chip on 90nm . I don't see the cell costing the same .

Not to mention we don't know how well a cell chip will perform doing normal pc tasks . So while it may be suited for graphic intensive apps , it can still be 2-4 times slower in normal pc tasks than a dual core athlon 64 or dual core p4 .

Using the same logic we could say that a turtle can swim faster than a cheetah. It's probably true, but it's rather pointless, especially when the race's primary objective is to run through a jungle.

Typical PC tasks do not matter precisely because they are not slow to begin with. Increasing the processor speed won't make word apparently faster, and it won't observably speed up your web browsing despite what Intel wants you to think. There is little to no reason to bother increasing the processor speed for those kinds of applications. Instead, the driving force behind PC CPU upgrades in the last 10 years have been for multimedia applications and number crunching (either large scale server/db work or scientific data analysis). Coincidently, these are exactly the kinds of tasks that Cell is good at.

Nite_Hawk

 

[Shifty Geezer]

These are precisely the tasks Cell was designed for! As KK said, existing PCs do multimedia work built from an ancestry that had nothing to do with MM. In essence it's been tacked on. He felt a clean-sheet design was needed to address these new processing requirements from the roots of their needs, such as feeding the processing elements with high-bandwidth data. I agree wholeheartedly. PCs are tied to legacy support, and the CPU has to cope with different ways of doing the same things, running 5+ year old software, handling all sorts of functions. A jack of all trades may be good at most things, but won't excel at any.

The only obvious downside to Cell is that fact people have to learn a new way of doing things, but it's not that bad. Every time a change comes along people grumble, but they adapt and in a couple of years multithreaded in-order code will be second nature to most devs. Either that or there'll be some killer compilers and tools!!

 

[phat]

I find it kinda ironic that the CELL benchmarks clearly show huge GFLOPS differences between different types of problems, but people here put so much faith in GFLOPS without even knowing how it rates when it comes to game problems.

That's going to be the case for any CPU (include Xcpu) that gains its FLOPS advantage by being optimized for vectorized, parallelized rephrasings of problems.

 

[jvd]

Typical PC tasks do not matter precisely because they are not slow to begin with. Increasing the processor speed won't make word apparently faster, and it won't observably speed up your web browsing despite what Intel wants you to think. There is little to no reason to bother increasing the processor speed for those kinds of applications. Instead, the driving force behind PC CPU upgrades in the last 10 years have been for multimedia applications and number crunching (either large scale server/db work or scientific data analysis). Coincidently, these are exactly the kinds of tasks that Cell is good at.

I don't know when the last time was that i sat around and just used word. I'm normaly running a few dozen im convos , runing my browser , running a mp3 file and playing a mmorpg at the same time tabing between them . Faster cpus are needed for this and so are multi core cpus .

The athlon 64 is designed to be a general processer . To run the tasks assigned to it as quickly as possible . To look at one area in this case flops and claim another cpu is faster with out looking at the whole picture is a stupid comparission .

 

[aaaaa00]

Increasing the processor speed won't make word apparently faster, and it won't observably speed up your web browsing despite what Intel wants you to think.

Actually it will. Have you tried to browse any Flash-heavy website on a old machine lately? Ugh. All those stupid alpha-blends and animations take forever to paint.

 

[scooby_dooby]

These are precisely the tasks Cell was designed for! As KK said, existing PCs do multimedia work built from an ancestry that had nothing to do with MM. In essence it's been tacked on. He felt a clean-sheet design was needed to address these new processing requirements from the roots of their needs, such as feeding the processing elements with high-bandwidth data. I agree wholeheartedly.

IMO that's nothign but marketing spin.

Look at Sony man, they are in trouble! They're TV's aren't selling, PSP is not selling, the company as a whole is bleeding. They lost $500million last quarter, and just released another $200million loss this 1/4.

They have no room to invest 3 billion in a piece of hardware meant only for ps3, and it's totally unrealistic to think they would do so. PS3 is a delivery mechanism for CELL, to hopefully open the door to other uses such as HD decoding, TV's chips, DVD chips etc. It's the same as blu-ray, it wasn't really included to create better games, but to try and open the door for blu-ray as the worldwide standard.

End goal: increase profits for Sony Corp. through royalties on both blu-ray and cell when they both become industry standards, as well as reduce production costs/increae profits on sony's own HD players/TV's etc etc

Sony is in this wierd situation where they are a hardware maker, but their chief profit comes froma game consoles. So you end up with the console having technologies that are intended, or cross-engineered for other purposes and goals.

It's nice, as a PS3 fan to think they are providing all this stuff for games, but look at the company, look at what they are trying tp push, look at the losses they are taking and you see this isn't be the case.

 

[jvd]

Sony is in this wierd situation where they are a hardware maker, but their chief profit comes froma game consoles. So you end up with the console having technologies that are intended, or cross-engineered for other purposes and goals.

Thats not fully a bad thing however .

Cell would be good in a tivo situation and also as a decoder for a bluray stand alone player .

 

[Fox5]

These are precisely the tasks Cell was designed for! As KK said, existing PCs do multimedia work built from an ancestry that had nothing to do with MM. In essence it's been tacked on. He felt a clean-sheet design was needed to address these new processing requirements from the roots of their needs, such as feeding the processing elements with high-bandwidth data. I agree wholeheartedly. PCs are tied to legacy support, and the CPU has to cope with different ways of doing the same things, running 5+ year old software, handling all sorts of functions. A jack of all trades may be good at most things, but won't excel at any.

The only obvious downside to Cell is that fact people have to learn a new way of doing things, but it's not that bad. Every time a change comes along people grumble, but they adapt and in a couple of years multithreaded in-order code will be second nature to most devs. Either that or there'll be some killer compilers and tools!!

The G5 isn't tied to the x86 legacy, and has more impressive theoretical numbers, yet it doesn't do so well in real world tests. Who's to say that if you suddenly added tons more FLOPs execution units onto an Athlon that it would suddenly perform much better?

BTW, don't gpus, especially as they become more programmable, largely satisfy the PC's need for FLOPs? Modern gpus can even decode and encode video...
And stupid flash animations could probably be accelerated by them too.

Sony is in this wierd situation where they are a hardware maker, but their chief profit comes froma game consoles. So you end up with the console having technologies that are intended, or cross-engineered for other purposes and goals.

Really? I didn't see PSP using an OLED screen.(don't think it uses ATRAC either)

 

[darkblu]

Really? I didn't see PSP using an OLED screen.(don't think it uses ATRAC either)

psp screen is sharp's make. but the thing uses atrac. it also uses memory sticks.

 

[Jawed]

PS3 is a delivery mechanism for CELL, to hopefully open the door to other uses such as HD decoding, TV's chips, DVD chips etc. It's the same as blu-ray, it wasn't really included to create better games, but to try and open the door for blu-ray as the worldwide standard.

You seem to be buying into the hype. Cell is far too powerful (in FLOPS!) for all these applications.

I really can't think of a single consumer application (non-gaming) that merits even a 1xSPE Cell.

Sony is going to need a whole new class of application, e.g. using the computing power of Cell to create virtual TV programmes with computer generated hyper-real news presenters melding some kind of multi-media RSS-type feeds into a seamless whole indistinguishable from actual TV news programmes.

Until then Cell is just way too costly, bulky and power-hungry to merit inclusion in flat-screen TVs or BR players or multi-screen home-networked multi-media server systems.

Jawed

 

[Carl B]

Look at Sony man, they are in trouble! They're TV's aren't selling, PSP is not selling, the company as a whole is bleeding. They lost $500million last quarter, and just released another $200million loss this 1/4.

This creeps up every now and then, but I have to emphasize Sony is *not* in trouble in the sense of losses/profit squeeze. They are in trouble due to lower margins in their core electronics business relative to competitors like Samsung - but that's a different situation.

The fourth quarter is a traditionally 'lossy' quarter for Sony, but the year on the whole was one of their most profitable ever.

Here's an extremely truncated article I found on the subject to quickly put their last years performance in perspective financially:

Article

 

[scooby_dooby]

According to gamespot this is sony's 2nd 1/4 not the 4th, and the 1st quarter they lossed nearly 3 times as much as this quarter.

That article doesn't make sense, they claim sony had a net loss yet at the same time their profits rose 85%, maybe you can clarify that.

jawed - i'm not buying into the hype. common sense just tells me that a company that is struggling to make profits wouldn't invest ~3 billion dollars in a technology that was intended only for use in their consoles, especially when they could've just gone with a simple 3 core IO CPU like MS did.

and also that the strength of CELL truly seem to be as a massive A/V decoder/encoder and not as a console CPU.

Then again they did a very similar thing with the EE, so trying to guess at Sony's motives/rationale is kinda hard.

 

[Carl B]

According to gamespot this is sony's 2nd 1/4 not the 4th, and the 1st quarter they lossed nearly 3 times as much as this quarter.

That article doesn't make sense, they claim sony had a net loss yet at the same time their profits rose 85%, maybe you can clarify that.

jawed - i'm not buying into the hype. common sense just tells me that a company that is struggling to make profits wouldn't invest ~3 billion dollars in a technology that was intended only for use in their consoles, especially when they could've just gone with a simple 3 core IO CPU like MS did.

and also that the strength of CELL truly seem to be as a massive A/V decoder/encoder and not as a console CPU.

Then again they did a very similar thing with the EE, so trying to guess at Sony's motives/rationale is kinda hard.

Ah ha! Sorry, I forgot it was time for a new earnings report - TOTALLY my fault and I apologize for that.

Anyway the loss is for the quarter in that previous article, and the gain is profits for the entire year (all four quarters combined, and compared to the previous years performance.)

As far as Cell and it's investment, cosnider it 'only' ~$600 million for the R&D itself, and the rest as investments in fabbing capacity.

Yeah though, Sony is reporting a loss for the second quarter in a row here; bad news. Still, they'll be profitable for the year again, even if seemingly less so than last year. It's certainly not what you want, but it's not the vultures of bankruptcy just yet either.