Anandtech dashes cpus from ps3 and xbox360

[ringlord]

What a disaster. Anyone want to take bets on if the article ever comes back? :?

 

[Shifty Geezer]

God forbid!

 

[pso]

On the Gamespot forums a user by the name "Xdrive" posted a rebuttal supposedly coming from an IBM engineer about the Anandtech article:

Anyways, let's see what David Milford, who works at IBM has to say about the article, shall we, haters?

"
Anand is out to lunch on a lot of this. His previous article was pretty sketchy, but this one is pretty much overboard with inaccuracies and random leaps to conclusions.

For one, he bases a lot of his discussions on the fact that the PPE is "identical" to a Xenon core, which is far from the truth. In fact, internally the Xenon team and the Cell team were not allowed to communicate with eachother for legal reasons. The cores were developed by different teams, they just happened to have made the same design decisions for a couple fundamental aspects (like SMT and in-order execution).

He gets so much fundamental basic things wrong that I'm shocked it was published, usually Anand is very knowledgable.
Floating point multiplies are also not "1/3" as fast as on Xenon, I've no idea where he got that from. It sounds like some developer told him that and he took it as verbatim truth.

What I think is going on here is Anand talked to some (curiously anonymous) developers who took their code designed for more PC-like processors (like the G5/PowerPC 970) and ran it on Xenon/Cell and was amazed that it didn't perform that well.

Xenon and Cell are in-order processors. The order of the instructions is very important, because the processor doesn't dynamically re-order them on Xenon/Cell, unlike the PC processors. A straight port will give you sub-par performance, easily.

Once they learn more about the chip, and how to program for it, the performance will be much, much higher.

This article is from a PC-centric website likely talking to PC-centric developers who don't have much experience with in-order cores. It's pretty worthless."

It may be nothing or it could be something, but its interesting since the article got pulled.

 

[AlNom]

Thanks for that.

 

[Alpha_Spartan]

Ecmaster76: Eh, something was messed up with the content management system. PS3 article is pulled for now because Anand is worried about MS tracing his anonymous insider.

Kristopher

http://www.anandtech.com/talkarticle.aspx?i=2458

So this was all based on the ramblings of one developer?

Anand is full of shit. That article wasn't taken down because he was afraid of MS, it was taken down because it was chock full of inaccuracies. Anand is a PC apologist.

 

[one]

On the Gamespot forums a user by the name "Xdrive" posted a rebuttal supposedly coming from an IBM engineer about the Anandtech article:

Where can I read what this David Milford guy wrote firsthand? I'll appreciate the pointer.

For one, he bases a lot of his discussions on the fact that the PPE is "identical" to a Xenon core, which is far from the truth. In fact, internally the Xenon team and the Cell team were not allowed to communicate with eachother for legal reasons. The cores were developed by different teams, they just happened to have made the same design decisions for a couple fundamental aspects (like SMT and in-order execution).

I've always assumed like this just by the contract nd common sense, but the Ars Technica writer and likes have been insisted they are identical, not just virtually identical for the sake of comparison, so I want to make sure they are not.

 

[therealskywolf]

On the Gamespot forums a user by the name "Xdrive" posted a rebuttal supposedly coming from an IBM engineer about the Anandtech article:

Where can I read what this David Milford guy wrote firsthand? I'll appreciate the pointer.

For one, he bases a lot of his discussions on the fact that the PPE is "identical" to a Xenon core, which is far from the truth. In fact, internally the Xenon team and the Cell team were not allowed to communicate with eachother for legal reasons. The cores were developed by different teams, they just happened to have made the same design decisions for a couple fundamental aspects (like SMT and in-order execution).

I've always assumed like this just by the contract nd common sense, but the Ars Technica writer and likes have been insisted they are identical, not just virtually identical for the sake of comparison, so I want to make sure they are not.

He posts sometimes on the IGn Boards. He is pretty credible. Original post here http://boards.ign.com/PS3_General_Board_/b8267/91468602/p1 if you have insider.

 

[Shifty Geezer]

How much difference can there be between XeCPU and PPE? They use the same core as a derivative, the same API, and the same principles as other processors. Short of a few more registers here, a few less there, and a few unique commands, I'm not expecting a lot of diversity. Save in memory access perhaps. I'd be interested to what the real differences are.

 

[one]

They use the same core as a derivative

Do you have any sources on it? i.e. Waternoose is derived from this PPC core?

 

[DukenukemX]

If you follow the XBOX mod community, in order to emulate an XBOX or PS2 on a current PC would require much much more power than simply twice a 733 Celeron.

It simply can't be true.

If it's only twice as powerful there's no way it can emulate the XBOX.

The emulator CXBX emulates the Xbox on the PC at full speed using a Pentium 3 at 800 Mhz and a Geforce 3 or higher graphics card.

The PS2 isn't emulated yet at full speed. Mind you that Sony and Microsoft have all the documentation that is known about their console systems. Where the free emulators must reverse engineer in a way that is legal. Witch takes time. Currently no PS2 game is fully playable on PC. Let alone any real optimizations which is always done last.

So therefore saying that because Xbox 360 and PS3 will emulate doesn't hold any ground.

On the contrary, expecting better visuals from high end PC's will leave you disapointed (in the first couple of years of next gen anyway).

The RSX chip in the PS3 will be using a type of Turbo Cache. The 22.4 GB/sec of bandwidth will probably be shared with the CPU's in the 360.

Even the fastest graphics card will fall to it's knees if it isn't supplied with enough bandwidth. I'm thinking that even a Geforce 6800 would be faster compared to the GPU in PS3.

The Geforce 6800 Ultra has 35.2 GB/sec that isn't sharing it with anything else. Which makes PC graphic cards already faster then some of the new console systems that are coming out.

 

[Embedded Sea]

SPE's not having a lot of local memory is a legitimate concern...

Like Deano has said before on this board (which seems to have a limited memory), PS3 will be good for some things and bollocks for others.

It will not be very good for porting something like Half Life 2 without a major re-write. Trust me it will be like pulling out nose hairs. Too much AI, too much gameplay code (imagine large objects being flushed in and out like mad). Any advanced FPS type game will be easier to write for any western house on Xbox 360. PS3's advantages are easily outweighed by the difficulty in achieving those results.

 

[Shifty Geezer]

The emulator CXBX emulates the Xbox on the PC at full speed using a Pentium 3 at 800 Mhz and a Geforce 3 or higher graphics card.

So therefore saying that because Xbox 360 and PS3 will emulate doesn't hold any ground.

The reason a P3 and GeForce3 can emulate XB is because it's basically the same hardware! Same instructions for the CPU, and GPU, using the same interface (DirectX). The reason it's hard for XB360 to emulate XB is because the processor uses a totaly different API (language if you like) and the GPU works differently coming from a different manufacturer - totally different architectures.

The 22.4 GB/sec of bandwidth will probably be shared with the CPU's in the 360.
The Geforce 6800 Ultra has 35.2 GB/sec that isn't sharing it with anything else. Which makes PC graphic cards already faster then some of the new console systems that are coming out.

You're getting caught up in the numbers. ie. Xenos doesn't NEED as much bandwidth as it can use the eDRAM (32 GB/s to eDRAM, internal processing, frees up a lot of external requirement).

If bandwidth were horsepower, we know bigger engine doesn't alway mean a faster car. A smaller engine with less horsepower can outperform a big, heavy motor with a huge engine. The important consideration is TOTAL SYSTEM DESIGN, not individual component performance.

 

[Titanio]

Embedded - you're adding 2 + 2 and getting a billion. A rather humourous billion at that. At the very least, how difficult it is to harness is a matter of subjectivity. And no one in their right mind would take PC HL2 code and expect good results out of PS3. X360 may be easier to harness, but if top performance is not THAT good, I'd take my chances with Cell (assuming once the SPEs are tapped, they prove useful and fast for a few key areas as has been mooted).

edit - going back to "difficulty", look at UE3 - running on PS3 kits in less than 2 months. Sure, it's nowhere near using all the power of Cell, but that should excite rather than initimidate.

As for the Anand article - so what of it was actually correct? I don't know if we should just write it all off since it was pulled and because of Anand's own spin. Some points from it would seem to be reasonably true given the reaction of some devs here, and seemingly some of the ones that caused most controversy.

 

[one]

The emulator CXBX emulates the Xbox on the PC at full speed using a Pentium 3 at 800 Mhz and a Geforce 3 or higher graphics card.

I think CXBX just hooks and redirects APIs... In XBOX 360 you have to do binary translation.
EDIT: Shifty Geezer beat me!

 

[seismologist]

It's a shame the article got pulled. MS is the one who's been citing general purpose code performance in their CPU benchmark.

If you want to talk general purpose code performance then I think Anand's article was perfectly valid.

 

[mckmas8808]

If you want to talk general purpose code performance then I think Anand's article was perfectly valid.

There's no way that the article is valid. If it was so valid why'd he pull it? Everyone here can see through that article, so we should just write it off.

Check this out. This is why we should be dissappointed in Anand's article.

Sony's PS3 and Microsoft's Xbox 360 will not be as powerful as their technical specifications - and clever marketing - make out when they land in your living room, according to technical experts.
In an article on website Anandtech the author explains why the two next-gen consoles will not be as powerful in 'real terms' (i.e. when they actually have to crunch games) as they may seem from the specs touted by Sony and Microsoft.

Anandtech ultimately surmises that both consoles will be around as powerful as each other. On top of that bombshell, it's suggested that neither next-gen console will offer the advances in power-draining areas like game physics we were expecting.

According to the report the Xbox 360 will only be twice as powerful as the original 733MHz Xbox, while the seven Synergistic Processing Elements (SPEs) in the PS3 will hobble the console's peak performance to a similar power.
The article references anonymous developers with knowledge of PS3 and 360 dev kits as being "not pleased" with the performance of either system. However, it is pointed out the combination of superior processing power, smoother online gaming and wireless pads will make the whole next-gen console gaming experience far better than the current-gen machines.

And it's not all bad news - Anandtech finds that the Graphical Processing Units (GPUs) in both the PS3 and Xbox 360 are extremely impressive. Developers are apparently "quite pleased" with the power of the GPUs, although the disappointing processing power does mean that they cannot be exploited to their maximum potential.

It's worth remembering that these findings are not endorsed by Sony or Microsoft, but the Anandtech article makes interesting and enlightening reading for anyone interested in finding out what all those crazy numbers and letters on spec sheets mean. It even explains what a teraflop and a floating point calculation is, which is pretty special if you ask us.

You can check out Anandtech here and catch up with the article here.

This is why I'm pissed to see a poor excuse for an article to be written. Stupid websites like CVG and others will repeat it like its actually true. Very sad, this is very very sad.

 

[seismologist]

If you want to talk general purpose code performance then I think Anand's article was perfectly valid.

There's no way that the article is valid.

More than likely it was pulled due to outside pressure.

If it was so valid why'd he pull it? Everyone here can see through that article, so we should just write it off.

See through what? I dont think anyone would argue that the next gen console CPUs are superior to a top of the line P4 at running general purpose code.

He also bring up a lot of good points about theoretical performance vs. real world. Most of the complaints are based on discussions with developers (specifically PC developers) who are rightly disappointed to see poor performance cut in half when porting their game engine to so called susperior next gen consoles. So the ony solution is to rewrite their engine from scratch to fit the new architecture. It's going to be PC ports to the PS2 all over again. With games never reaching the theoretical performance claims.

With Sony it's pretty much a given that they'd have a proprietary solution but you can't blame them for wishing MS had stuck with a more traditional approach. Especially PC developers who were some of Xbox's strongest supporters thanks to the ease of porting.

My guess is that the Anand's "anonymous source" might be one of the UE3 people...

 

[gosh]

If you want to talk general purpose code performance then I think Anand's article was perfectly valid.

There's no way that the article is valid.

More than likely it was pulled due to outside pressure.

If it was so valid why'd he pull it? Everyone here can see through that article, so we should just write it off.

See through what? I dont think anyone would argue that the next gen console CPUs are superior to a top of the line P4 at running general purpose code.

He also bring up a lot of good points about theoretical performance vs. real world. Most of the complaints are based on discussions with developers (specifically PC developers) who are rightly disappointed to see poor performance cut in half when porting their game engine to so called susperior next gen consoles. So the ony solution is to rewrite their engine from scratch to fit the new architecture. It's going to be PC ports to the PS2 all over again. With games never reaching the theoretical performance claims.

With Sony it's pretty much a given that they'd have a proprietary solution but you can't blame them for wishing MS had stuck with a more traditional approach. Especially PC developers who were some of Xbox's strongest supporters thanks to the ease of porting.

My guess is that the Anand's "anonymous source" might be one of the UE3 people...

you pull an article if it gives out information which is under NDA not information of numbers and facts which everyone knows. Pulling the article only demonstrates they got thier information from "ONE developer" who basically...well how should i put it...glorified himself to anand by lying

 

[gosh]

If you want to talk general purpose code performance then I think Anand's article was perfectly valid.

There's no way that the article is valid.

More than likely it was pulled due to outside pressure.

If it was so valid why'd he pull it? Everyone here can see through that article, so we should just write it off.

See through what? I dont think anyone would argue that the next gen console CPUs are superior to a top of the line P4 at running general purpose code.

He also bring up a lot of good points about theoretical performance vs. real world. Most of the complaints are based on discussions with developers (specifically PC developers) who are rightly disappointed to see poor performance cut in half when porting their game engine to so called susperior next gen consoles. So the ony solution is to rewrite their engine from scratch to fit the new architecture. It's going to be PC ports to the PS2 all over again. With games never reaching the theoretical performance claims.

With Sony it's pretty much a given that they'd have a proprietary solution but you can't blame them for wishing MS had stuck with a more traditional approach. Especially PC developers who were some of Xbox's strongest supporters thanks to the ease of porting.

My guess is that the Anand's "anonymous source" might be one of the UE3 people...

you pull an article if it gives out information which is under NDA not information of numbers and facts which everyone knows. Pulling the article only demonstrates they got thier information from "ONE developer" who basically...well how should i put it...glorified himself to anand by lying

 

[mistwalk]

Hey...PPE transistors are about 55M.(includeing 512kb L2 cache)
G5's are about about 58M(including 512kb L2 cache).
I don't think that PPE will be slower than G5.
Maybe worse GP,but better FP
And I rememder EPIC vice president ever said that bunchs of SPEs are much similar to PPU architecture.
So you could regard Cell as PPE + a super PPU.
Then..... assuming one of XeCPU's tri-core is like PPE....
Well,three cores.......good enough.