Sony delays PS3 to 2006, concentrates on PSP !?

[jvd]

Chap's little hidden agenda is so obvious, it almost hurts to see so many people falling for it.
Fear leads to Anger. Anger leads to Hate. Hate leads to Suffering.

You know the fear , anger , hate and suffering is what i think sony fans are affraid of .

If anyone other than chap and i guess myself posted this , no one would question its source. If this true it will hurt sony. No matter how u look at it . It will hurt sony. If everything was going perfect with the cell chips they wouldn't need to push it back or even entertain the idea of pushing it back.

Ms coming up with a system on par or surpassing the ps3 performance will require the same effort sony has put into it. Which i'm sure ms will be more than willing to put in. Tflops aren't important. Overall system performance is important .

Also who ever gave a performance estimate of 20billion polygons for a 1 tflop chip please show me some proof or equations of how u came up with that .

 

[Panajev2001a]

And what makes you so sure that both the consoles would be the same technology wise? If anything PS3 will blow Xbox2 away if they are released at the same time, Intel won't have anything that can touch a 1tflops Cell, and im doubting Nvidia would be able to match the GS3 spec wise for a while.

Technology wise I just don't feel there will be a big difference. The PS 3 may have a more elegant design compared to a legacy x86 architechture of a X-Box 2, but the advantages gained by Sony having a highly customized design will hardly be that great. Microsoft just feeds off of the PC industry which all the time upholds Moores Law regardless of architechture. Sony isn't going to blow away Moores Law.

There will be a difference...

First, Sony going the custom route and having spent all these years with IBM and Toshiba on a brand new architecture not as restrained by backward compatibility as IA-32 is has a quite nice performance advantage over Intel processors available at the same time as Intel processors...

Intel processors run very varied workloads and try to be as good as they can in all of them... Your Pentium 4 has to run Windows fast and also your games and also your Word Processor, your compiler, etc...

The PlayStation 3 Cell CPU has a more clear performance target and typical workload and this gives it an advantage.

It is not like Cell will not be good in general computing, it won't just be phenomenal in it, certainly not as good as it will run multi-threaded applications, applications that are based on streaming huge amounts of data and applications which deal with vecotr processing.

Moore's Law is too slow for Intel to start for MS in this year ( or even last year ) a new processor that matches Cell and it is also IA-32 backward compatible ( and with more than decent performance in it... )...

A current 3.06 GHz Pentium 4 yelds 12 GFLOPS... 1 TFLOPS = 83x

Even if we talked about 120 GFLOPS this would be 10x the performance increase... if you double the performance even every 6 Months you have 2 years * 12 / 6 = 4 product cycles and only 8x increase in FP performance.

Also Microsoft should have announced such an humongous task to Intel and paid lots of dollars for it a good while ago... Intel is busy with the IA-64 line, their Strong ARM branch and the R&D of their next generation IA-32 core ( after Prescott )...

Neither Intel nor IBM could design such a CPU in such a short ammount of time...

another advantage in owning the IP of the CPU and manufacture it yourself is being able to lower the manufacturing costs faster: new manufacturing process ready ? Perfect, a team is assigned to shrink the die and tadaaah after a certain while ( not THAT long as this research is basically done almost in parallel to the research on the new manufatcuring process ) we have a smaller CPU and we have cut the cost per CPU...

When you have contracts with 3rd parties to build the chips for you, you are in a less advantagious position... for example, you cannot re-negotiate the price until the current large batch of processors as been shippd and sold to you... even if Intel is now 3 manufacturing processes ahead, if you could not sell what you were getting and asked Intel to slow down the rate they sent you CPUs at, you could not easily ( unless you are willing to pay huge premiums ) renegotiate a new contract and a new price...


An advantage of going the custom route is to pack the exact technologies YOU need, not what the market itself brings to you... it might cost a bit, but it can be worth it

I do not recal if Microsoft makes as much money per Xbox as Sony does per PlayStation 2

Microsoft will go with "standard" IA-32 ( maybe x86-64 ) and custom nVIDIA GPU because most of the 3D graphics processing will not be done on the host CPU... the Xbox 2 will NOT need a CPU as powerful as Cell...

I am not saying that over-all the Xbox 2 will not be comparable performance wise from the PlayStation 3, but this is not a bad thing, not even from the business point of view... think about GCN and Xbox... who went the more custom route ? GCN... Whose's HW was still comparable yet much more expensive to manufacture ? The Xbox...

The real bad scenario would be if a company were to use custom technology and paid a lot of money to another company to design this custom HW and to manufacture it for them... this would be the worst of both worlds and even worse...

 

[Panajev2001a]

Ms coming up with a system on par or surpassing the ps3 performance will require the same effort sony has put into it. Which i'm sure ms will be more than willing to put in.

Wrong !

Aside from some other issues which would be affecting this statement ( amnufacturing, etc... ) the simple fact that Sony has had an at least 1+ year longer R&D cycle makes your statement validity very questionable

 

[Paul]

If everything was going perfect with the cell chips they wouldn't need to push it back or even entertain the idea of pushing it back.

And when exactly did Sony even "entertain" the idea of pushing it back? This whole shit about Cell being delayed came from the Designchain article, which we know turned out to be BS. Cell is currently and still is planned to go into production 2004, this is from Toshiba themselves.

Ms coming up with a system on par or surpassing the ps3 performance will require the same effort sony has put into it.

Which they have showed no signs of doing, Xbox2's power is going to be dependant on what Nvidia/Intel can produce, msoft has money, however not unlimited time.

Tflops aren't important.

So I guess those big bad 40tflops supercomputers are really garbage than? And a pentium 4 clocked at 3ghz can easily take one down.

Overall system performance is important .

And TFLOPS don't relate to overall system power how?

Also who ever gave a performance estimate of 20billion polygons for a 1 tflop chip please show me some proof or equations of how u came up with that .

By whoever you mean me, come out and say it.

Second, I said 15-20 not 20.

My calcuations are very dirty, I warn you, they aren't using exact numbers at all. I know some of the numbers will be off, so don't anyone come and say "but ee does 6.2gflops" or any of that, because I know.

At 6gflops EE is capable of 66 million polygons per second correct?

Well now, let's see the estimated poly performance of a 1200GFLOPS PS3, adding the CPU+GPU together.

1200 / 6 = 200

66 million * 200 = 13B polygons.

Now, from the fact that my calculations were dirty, add another 1.5B to that figure.

You can also add another 2B or so, as there will undoubtably be other ways Sony will squeese polygon performance out of ps3 other than pure FLOPS power alone.

 

[Panajev2001a]

Overall system performance is important .

I agree and that is why I like e-DRAM sitting on a fat 1,024 bits pipe and with fat 1,024 bits internal busses for the processor, several 128 KB Local Storage memories and 25.6 GB for external RAM

 

[Panajev2001a]

Even better... assume that each APU can do a basic transform ( with perspective divide... ) in let's say 14 cycles ( VU1 can do it in 5 ) to be on the safe side...

4 GHz / 14 = ~285 MVertices/s

We have 8 * 4 = 32 APUs in the Broadband Engine...

285 * 32 = 9.120 Billion Vertices/s

You want to be even safer ?

32 cycles ( VU1 can do it in 5 ) to be on the safe side... for the transform ( we could even fit a good bit of lighting in 32 cycles IMHO )...

4 GHz / 32 = 125 MVertices/s

We have 8 * 4 = 32 APUs in the Broadband Engine...

285 * 32 = 4 Billion Vertices/s

If our goal were only 1 Billion Vertices/s the Cell architecture described in the patent ( 32 APUs, 4 GHz [32 GFLOPS per APU and each APU has 4 FP Units] ) would be taking...

128 cycles per Vertex...

4 GHz / 32 = 31.25 MVertices/s

We have 8 * 4 = 32 APUs in the Broadband Engine...

31.25 * 32 = 1 Billion Vertices/s

 

[Josiah]

focusing on pure vertex throughput is being narrowsighted, you have to ask yourself: which platform will have superior shader abilities? will PS3 even have such abilities? which platform will be capable of superior lighting? animation? spatial/temporal antialiasing? etc etc, vertex TnL is only one part of the graphics pipeline and not even the biggest part. throwing more polygons into a scene gives diminishing returns as you move forward, and with the complexity we're already seeing in games things like per-pixel lighting and realistic animation are going to matter alot more.

 

[Brimstone]

And what makes you so sure that both the consoles would be the same technology wise? If anything PS3 will blow Xbox2 away if they are released at the same time, Intel won't have anything that can touch a 1tflops Cell, and im doubting Nvidia would be able to match the GS3 spec wise for a while.

Technology wise I just don't feel there will be a big difference. The PS 3 may have a more elegant design compared to a legacy x86 architechture of a X-Box 2, but the advantages gained by Sony having a highly customized design will hardly be that great. Microsoft just feeds off of the PC industry which all the time upholds Moores Law regardless of architechture. Sony isn't going to blow away Moores Law.

There will be a difference...

First, Sony going the custom route and having spent all these years with IBM and Toshiba on a brand new architecture not as restrained by backward compatibility as IA-32 is has a quite nice performance advantage over Intel processors available at the same time as Intel processors...

Intel processors run very varied workloads and try to be as good as they can in all of them... Your Pentium 4 has to run Windows fast and also your games and also your Word Processor, your compiler, etc...

The PlayStation 3 Cell CPU has a more clear performance target and typical workload and this gives it an advantage.

It is not like Cell will not be good in general computing, it won't just be phenomenal in it, certainly not as good as it will run multi-threaded applications, applications that are based on streaming huge amounts of data and applications which deal with vecotr processing.

Moore's Law is too slow for Intel to start for MS in this year ( or even last year ) a new processor that matches Cell and it is also IA-32 backward compatible ( and with more than decent performance in it... )...

A current 3.06 GHz Pentium 4 yelds 12 GFLOPS... 1 TFLOPS = 83x

Even if we talked about 120 GFLOPS this would be 10x the performance increase... if you double the performance even every 6 Months you have 2 years * 12 / 6 = 4 product cycles and only 8x increase in FP performance.

Also Microsoft should have announced such an humongous task to Intel and paid lots of dollars for it a good while ago... Intel is busy with the IA-64 line, their Strong ARM branch and the R&D of their next generation IA-32 core ( after Prescott )...

Neither Intel nor IBM could design such a CPU in such a short ammount of time...

another advantage in owning the IP of the CPU and manufacture it yourself is being able to lower the manufacturing costs faster: new manufacturing process ready ? Perfect, a team is assigned to shrink the die and tadaaah after a certain while ( not THAT long as this research is basically done almost in parallel to the research on the new manufatcuring process ) we have a smaller CPU and we have cut the cost per CPU...

When you have contracts with 3rd parties to build the chips for you, you are in a less advantagious position... for example, you cannot re-negotiate the price until the current large batch of processors as been shippd and sold to you... even if Intel is now 3 manufacturing processes ahead, if you could not sell what you were getting and asked Intel to slow down the rate they sent you CPUs at, you could not easily ( unless you are willing to pay huge premiums ) renegotiate a new contract and a new price...


An advantage of going the custom route is to pack the exact technologies YOU need, not what the market itself brings to you... it might cost a bit, but it can be worth it

I do not recal if Microsoft makes as much money per Xbox as Sony does per PlayStation 2

Microsoft will go with "standard" IA-32 ( maybe x86-64 ) and custom nVIDIA GPU because most of the 3D graphics processing will not be done on the host CPU... the Xbox 2 will NOT need a CPU as powerful as Cell...

I am not saying that over-all the Xbox 2 will not be comparable performance wise from the PlayStation 3, but this is not a bad thing, not even from the business point of view... think about GCN and Xbox... who went the more custom route ? GCN... Whose's HW was still comparable yet much more expensive to manufacture ? The Xbox...

The real bad scenario would be if a company were to use custom technology and paid a lot of money to another company to design this custom HW and to manufacture it for them... this would be the worst of both worlds and even worse...

My gripe with Cell right now is that it's based on faith. We haven't seen any type of benchmarks from working silicon. It's hard to really debate faith. You can't disprove faith. That doesn't mean Cell doesn't exist and it's a really good chip. The Intel x86 path is much easier to predict so any company can put up numbers that will trounce over what Intel will have in the future. The challenge is to execute and beat Intel in a given time frame. AMD beat Intel for a short time so Intel can certainly be beat on its own turf.

No doubt that Sony will have a extremley effecient platform in the cost/power ratio.

What I'm curious about is the console platform Intel was working on before Microsoft came to them about their plans for the X-Box. Did this have a custom architechture? Has a group inside Intel continued to work on this?

 

[jvd]

Ms coming up with a system on par or surpassing the ps3 performance will require the same effort sony has put into it. Which i'm sure ms will be more than willing to put in.

Wrong !

Aside from some other issues which would be affecting this statement ( amnufacturing, etc... ) the simple fact that Sony has had an at least 1+ year longer R&D cycle makes your statement validity very questionable

So your telling me you know when the r&d for both systems started ? For all we know the r&d for the xbox 2 started a year before the xbox was done. Not only that but its not ms that is putting the r&d into the parts its other companys. It could be ati or power vr or nvidia. Who knows how long they have been working on the chip that will be used in the xbox2. Doesn't matter if they were planed to be used in the xbox 2 from 3 years ago or tommorow ms decides to use them. They have been worked on for awhile now .

There is no reason why another company can not make a video game system on the same level as the ps3 and released around the same time. Of course sony will have some advantages going a custom chip. But ms will have advantages going after i.p .

 

[Paul]

And what is there to R&D jvd? Msoft doesn't make their own chips, so everything Xbox2 is going to become depends soley on the people making the chips, which Msoft is not.

So it's going to depend on Nvidia/Intel, and do you really think they will put huge resources into years development of a videogame console when they have their PC market to attend to? Intel and Nvidia that is.

 

[jvd]

Overall system performance is important .

I agree and that is why I like e-DRAM sitting on a fat 1,024 bits pipe and with fat 1,024 bits internal busses for the processor, several 128 KB Local Storage memories and 25.6 GB for external RAM

How about transforming polygons when using lights ? Bah this is all pointless. Lets debate cold fusion while we are at it . Anyway wish u guys were here in italy with me so i could kick your butts in soccer .

 

[Paul]

Lmao, I'll be in Barbados for two weeks leaving June 30th.. Italy.. Can't compete with almighty Barbados!

I think me, Pana and Vince can take on any trio on this board at soccer, bring it!

 

[Panajev2001a]

jvd... well I am not in perfect shape, but if you want I should be back in Italy for a couple of weeks this summer

I should be staying in the Torino area, maybe travelling by rain to visit friends if they are not too far...

Be afraaaaid, be VEEEEERY afraid...





Brimstone,

I know and that is why I do not think Intel can match the processor Sony has in mind for the PlayStation 3 FLOPS wise...

January 1999 they had a 6.2 GFLOPS ( at that conference it was rated 250 MHz so the FLOPS rating was a bit less [was it 5.5 GFLOPS or 5.2 GFLOPS... I forgot ], but it got a speed bump shortly after ) CPU...

I do not think Intel was selling or had in the labs similarly working silicon that could achieve that performance...

This time Sony is being much more aggressive on the manufacturing process side ( 65 nm process with Toshiba will see a fab mass producing mid 2004 and then we will have the new Nagasaki fab which is also in construction... then we have a 45 nm manufacturing process Sony and Toshiba are working on and the 65 nm lines are being set-up to be quickly upgradeable to 45 nm technology ) and has added IBM to its partnership as well as Toshiba and RAMBUS ( they licensed the promising Yellowstone and Redwood technologies from RAMBUS ).

Cell is also receiving more funding compared to the Emotion Engine and the Graphics Synthesizer not only from SCEI, but also from Sony corp.

How about transforming polygons when using lights ?

You quoted something and made comment regarding lights...

I was talking, in the part you quoted, about features that help the CPU to achieve nearer to maximum performance figures...

Like e-DRAM and its massive bandwidth helps rasterizers like the Graphics Synthesizer and Flipper to achieve a higher efficiency...

AFAIK, higher bandwidth and lower latency seen by the execution units is a good thing

For all we know the r&d for the xbox 2 started a year before the xbox was done.

By who ? nVIDIA ? :lol like they would have had time

If you look at Xbox dvelopment timetable you are now saying that while they were still in quite deep R&D phase for Xbox 1 they were doing R&D for Xbox 2...

This is incorrect also... how do I know ?

Let me say that a little bird, named Ed Fries was talking on G4TV ( a gaming network ) a while ago ( ICONS, dedicated to the Xbox project ) and he basically stated they were ultra busy with Xbox , more than busy enough and yes they started R&D for Xbox 2 already, but that didn't happen until after the launch of Xbox 1 and no partners have been yet announced ( of course they are hididing it in their balance sheets... well we will see when they release their FY report... Sony is quite OBVIOUSLY doing R&D for PlayStation 3 and PSP as you can see from their balance sheets ; )...

focusing on pure vertex throughput is being narrowsighted, you have to ask yourself: which platform will have superior shader abilities? will PS3 even have such abilities? which platform will be capable of superior lighting? animation? spatial/temporal antialiasing?

Josiah.. use your rEYES

Yes I am suggesting PlayStation 3 could use a micro-polygon based rendering approach...

That should cover AA and motion blur ( temporal and spacial AA are covered then )...

And by the way, the APUs are fully flexible units and the GPU ( the Visualizer ) also has APUs together with the Pixel Engines...

The APUs are like enhanced and faster VUs... I think they could do either Vertex or Pixel Shading...

It is not like I am listing T&L throuputs of a plethora of T&L units... the APUs can do basically anything YOU want them to do

Think about evolved VUs ( Emotion Engine's VUs were > DX9 BTW )...

 

[Panajev2001a]

How about transforming polygons when using lights ?

BTW, I was using 128 cycles for each Vertex... that is QUITE a lot

 

[PC-Engine]

You guys think Xbox2 won't be capable of 1 Billion polys/sec? You think there's going to be a drastic difference onscreen between PS3, Xbox2, and GCN2?

BTW the PS2's ingame polys/sec rating isn't 66 M

 

[jvd]

Heh. I never said it was ms that was doing the r&d for the xbox 2. They are not. They are using chips made by other companys. Nvidia has been making chips for a very long time . The nv30 was easily in deisgn phase for 2or 3 years. But the ground work for the chip was being done for years prior as seen in other projects like the tnt , geforce , geforce 2 and geforce 3. To say that ms wont be able to find things to compete with sony is plain wrong. Its very ignorant to even suggest that . Esp at this point in the game.

 

[chaphack]

delay too long and ambitious games like SH, MGS, GTA, FF, SC, RE etc might jump ship to the newer better platforms like XB2. Again, MS != Sega/Nintendo. The 3rd party support MS gained at their first try is terrific.

so what if PS3 have better cpu, not like that matters with the PS2. Bettered at IQ/textures by DC, overtaken by a cheaper GC and overpowered by MS XBOX.

 

[Panajev2001a]

To say that ms wont be able to find things to compete with sony is plain wrong. Its very ignorant to even suggest that

I have never made such a general statement as the one you put here...

If Cell delivers and it is reasonably close to their promises, talking CPU vs CPU here, Microsoft will not be able to find in 2 years a CPU that has the same performance ( unless loosing $3,000+ on each console you sell it is acceptable.. and even then money can buy so much engineer time... ) nor will they have to...

Xbox 2 will focus on an ultra strong GPU like Xbox 1 did and nVIDIA is molving at a good pace to reach their goals...

Intel is not moving at the pace of matching Cell feature by feature... because they have other problems to worry about,: they have other plans for their CPUs and a CPU with lots of e-DRAM and tons of little APUs is not in their plans... and hasn't been so far...

If Intel had to start today or even last year that would be a little on the late side...

It was not aproblem for Intel of having the Emotion Engine pushing 6.2 GFLOPS in early 1999... they have a different timetable...

Microsoft has lots of money, but money can buy just so much engineer time...

At the end I expect all three consoles to have their strenghts and weaknesses ( one has a better CPU, the other has a better GPU and so on... ) and to be comparable to each other in terms of overall performance and I have stated this before and I am still of the same idea now...

There is just so much "ground work" you can do while creatign an architecture in regard to a new one you are going to release at a later time, in regard to the NEXT step that is...

If you notice even in the GPU business the rate the vendors release new KEY points in their architecture is not as close as you think...

During most of the heavvy work on the CineFX architecture there was not much time to do more than throw ideas on the table regarding what is after CineFX... a good chunk of that work ( which is done a bit more by the upper management than the engineers working on the silicon ) will be done once 99% of the work on the chip is doen and we have working samples and we are finally close to shipping unless we have bad issues like NV30 and the NEED for the NV35 to come out As Soon As Possible...
During this time you see several refreshes coming out, but the senior engineers are already working on the next generation chip ( of course since they normally have more than 1 team they can work a bit faster than what I am describing here ) and this number of engineers will grow and grow...

In the case of two teams ( two major teams )...

Example ( I am not making a history report, just presenting a plausible example sort of grounded in reality ):
R200 East team starts...

R300 West coast team starts ( let's assume they started working on the R300 when the R200 team started working on the R200... R300 would be their first chip )...

R200 is shipped: core of top senior engineers start working on the R400

R300 is close to shipping

R200 team is almost fully moved onto the R400

R300 is shipped: core of top senior engineers start working on the R500

R200 team is working fully on the R400

[...]

 

[Josiah]

I know and that is why I do not think Intel can match the processor Sony has in mind for the PlayStation 3 FLOPS wise...

January 1999 they had a 6.2 GFLOPS ( at that conference it was rated 250 MHz so the FLOPS rating was a bit less [was it 5.5 GFLOPS or 5.2 GFLOPS... I forgot ], but it got a speed bump shortly after ) CPU...

XBox (1) has a CPU theoretically capable of what, about 3 gigaflops? So PS2 games should look more than twice as good right? The truth is XBox is producing visuals technically superior. There's obviously more to it than that.

For all we know the r&d for the xbox 2 started a year before the xbox was done.

By who ? nVIDIA ? :lol like they would have had time

Most likely XBox2 will simply be a derivative of NVIDIAs latest, as XBox was. Really, all that's needed is a re-tool of an upcoming chip, they will feed off the R&D they are constantly doing anyway.

focusing on pure vertex throughput is being narrowsighted, you have to ask yourself: which platform will have superior shader abilities? will PS3 even have such abilities? which platform will be capable of superior lighting? animation? spatial/temporal antialiasing?

Josiah.. use your rEYES

Yes I am suggesting PlayStation 3 could use a micro-polygon based rendering approach...

That should cover AA and motion blur ( temporal and spacial AA are covered then )...

And by the way, the APUs are fully flexible units and the GPU ( the Visualizer ) also has APUs together with the Pixel Engines...

The APUs are like enhanced and faster VUs... I think they could do either Vertex or Pixel Shading...

It is not like I am listing T&L throuputs of a plethora of T&L units... the APUs can do basically anything YOU want them to do

Think about evolved VUs ( Emotion Engine's VUs were > DX9 BTW )...

>DX9? If that's true, then why can't a PS2 compare lighting/image quality wise with a DX9 part? Honestly, if you tried to do a game like Doom3 (DirectX 8 level) you would likely get 0.1 FPS out of the PS2....Is PS3 PS2 all over again?

 

[Saem]

Intel is not moving at the pace of matching Cell feature by feature... because they have other problems to worry about,: they have other plans for their CPUs and a CPU with lots of e-DRAM and tons of little APUs is not in their plans... and hasn't been so far...

I don't know about the EDRAM, but various rumours state that Intel is looking into multi-core chips.

I think there might be other things to consider, Intel's processors are VERY fast considering they're single threaded, when used appropriately hyper-threading adds a significant amount of performance. This added with further enhancements and a multi-core design could yield a lot of easily realisable performance with great tools on the software side.

Then there is the fact that as you stated in various posts on many occassions, the GPUs will be ridiculiously powerful -- when compared to todays stuff.

Holistically, I think the X-Box 2 will not be signifcantly disadvantage. Many architectures have tried to compete against x86 in the price/performance arena in simillar applications -- consoles are very high performance desktopish these days-- and they got whooped. Not to say Cell will suck, but there will be stiff competition. At the same time it should be noted that Intel places little emphasis on FP performance --in a historicaly sense-- as they should since the Windows work load doesn't demand it often.