Geko vs. EE vs. Pentium3-733

[JF_Aidan_Pryde]

I wasn't into consoles until recently and must have missed the CPU debate. How does it stack up? Also in related CPU architecture land, what is the main reason slower clocked RISC processors can perform on par with much higher clocked Intel CPUs?

Thanks.

 

[Phil]

I think a comparison would be rather difficult, as each of the CPUs have different tasks they need to complete.

If it's the EE architecture you're interested in, you might want to read the following two articles from ArsTechnica:

Sound and Vision: A Technical Overview of the Emotion Engine

The Playstation2 vs. the PC: a System-level Comparison of Two 3D Platforms

cheers,
Phil

 

[London Geezer]

I think a comparison would be rather difficult, as each of the CPUs have different tasks they need to complete.

If it's the EE architecture you're interested in, you might want to read the following to articles from ArsTechnica:

Sound and Vision: A Technical Overview of the Emotion Engine

The Playstation2 vs. the PC: a System-level Comparison of Two 3D Platforms

cheers,
Phil

yep, anyone could say that the Emotion Engine is much more powerful than the other 2 (and it is) but it has to do a lot more things than the Xbox and GC's CPUs (3D calculations, T&L, at a certain extent sound etc) as well so there goes your performance headstart...

 

[V3]

Also in related CPU architecture land, what is the main reason slower clocked RISC processors can perform on par with much higher clocked Intel CPUs?

I don't think it does anymore. I think by the time of Pentium Pro, Intel pretty much caught up to RISC processors.

But on RISC Vs CISC when it was first raised it has something to do with CISC spending transistors on instructions that can be done with several other basic instructions. Compiler is also easier to write for RISC. CISC has the advantage of having smaller code size, which was somewhat important when memory size was limited, not the case today though.

RISC, CISC discussion is not relevant today, today's processor sits happily somewhere in the middle.

I think a comparison would be rather difficult, as each of the CPUs have different tasks they need to complete.

Yeah T&L was needed because CPUs was inadequate, if Intel however decided to stick several vector processors in their Pentiums, you probably don't need T&L.

I am still puzzle why Intel don't make a GPU, and stuff it together with their CPUs, it could make a cheap mid range solution.

 

[JF_Aidan_Pryde]

I've actuallyl read the first Ars article before, it's very good. I can see that he EE's advantage is in its parallel nature. But why does a much lower clocked G4s today perform on par to much higher clocked Pentium4s? (or does it not?)

 

[London Geezer]

I've actuallyl read the first Ars article before, it's very good. I can see that he EE's advantage is in its parallel nature. But why does a much lower clocked G4s today perform on par to much higher clocked Pentium4s? (or does it not?)

well it does for certain things.................... but its doesnt for others..... i guess its like the GFFX doing some things better than the R300 and a lot more things worse

 

[Ozymandis]

On this same note, I'm curious as to how many polys/sec the latest Intel and AMD CPUs can do, and how that compares to the Playstation2.

Anywhere I can find such information?

 

[London Geezer]

On this same note, I'm curious as to how many polys/sec the latest Intel and AMD CPUs can do, and how that compares to the Playstation2.

Anywhere I can find such information?

well of course the Emotion Engine would stomp all over any Intel or AMD CPU available at the moment in terms of Polygon performance..... but when u look at the bigger picture, with the EE+GS and a (u said latest right) Pentium4 (cant remember the latest clock at the moment but should be around 3GHz) + GeforceFX 5800 Ultra then things start getting a bit different..........

still its pretty amazing that the Emotion Enigne is clocked at 300MHz and the GS at 150MHz while the others are like 10x faster and still it can output more polygons. Hell it outperformed PC boards pretty much until the Geforce4 Ti came out (some 3 years later). only talking polygons here....

 

[Ozymandis]

...but when u look at the bigger picture, with the EE+GS and a (u said latest right) Pentium4 (cant remember the latest clock at the moment but should be around 3GHz) + GeforceFX 5800 Ultra then things start getting a bit different..........

Yeah, but you look at the price of a 3.06 ghz P4 plus GeForceFX versus a 200 dollar Ps2. That's where consoles in general stomp all over PCs- price/performance

That's irrelevant here, but I just thought I'd point that out 8)

 

[Gubbi]

You're right, it's irrelevant. The P4 is a general purpose chip. Go compare the P4 and the EE does on SPEC2K.

I think that a 3GHz P4 will beat the EE in T&L quite handily. With a peak performance of 12GFLOPS (as compared to 6.4) and the memory subsystem to boot. -But again irrelevant.

Cheers
Gubbi

 

[London Geezer]

Yep, like comparing apples to cats...

 

[Ozymandis]

You're right, it's irrelevant. The P4 is a general purpose chip. Go compare the P4 and the EE does on SPEC2K.
Cheers
Gubbi

Well, it's not totally irrelevant to me. The only reason why I'd buy such a fast CPU is for gaming. It's not like you need more than a Pentium 2 for web-browsing and such. And if I can get similiar graphics on a 200 dollar system as opposed to a 2000 dollar one, it's definitely not irrelevant.

It IS irrelevant to this discussion, of course

I think that a 3GHz P4 will beat the EE in T&L quite handily. With a peak performance of 12GFLOPS (as compared to 6.4) and the memory subsystem to boot. -But again irrelevant.

I wish I could find some benchmarks of software T&L on that P4. Say a 3dmark 2001 T&L benchmark, just to get some idea.[/i]

 

[Tagrineth]

I am still puzzle why Intel don't make a GPU, and stuff it together with their CPUs, it could make a cheap mid range solution.

Timna. Wasn't economically feasible. Look it up.

On topic:

The reason other CPU's tend to be faster per-clock than Pentiums isn't a RISC versus CISC question... it's pure ISA.

The P3/Celeron hybrid in Xbox is a 32-bit x86 processor... x86 is pretty horrific as far as CPU ISA's go.

IBM's Gekko in GCN is also "32-bit" but it can work two 32-bit ops per cycle. Also it isn't x86 - it's PowerPC, which is pretty much known to be more efficient than x86 in general. Especially in games, believe it or not.

Basically, Gekko > XCPU, hands down. The only things XCPU has going for it are ace compilers and SSE instructions... other than that Gekko has the upper hand.

Sony's Emotion Engine is totally different altogether, and still gets the job done, while doing a hell of a lot more work... as has already been aluded to, GCN and Xbox have TCL processors on their graphics chips, which means Gekko/XCPU can focus on... well, everything else.

The Emotion Engine OTOH has to do all the TCL itself. It has two vector processors for this, but it's still a hell of a chore. If EE was allowed to do full game code... well, it would never hit even 50% use, I suspect... except in tech demos with controls

 

[Ozymandis]

Basically, Gekko > XCPU, hands down.

Well, the XCPU does have a significant clockspeed advantage. Even if the Gekko is faster per-clock, I wonder if it can make up that almost-50% difference in mhz?

I've heard that it cannot.

 

[randycat99]

I think that a 3GHz P4 will beat the EE in T&L quite handily. With a peak performance of 12GFLOPS (as compared to 6.4) and the memory subsystem to boot.

Note- that's 10x the clockspeed just to gain 2x the GFLOPs. I'd say the EE hangs in there quite admirably. Now compare die sizes. Can you imagine how easy it would be to scale a prototype EE 2x (maybe not even the entire core, but just throw in 2 more vector units and an a$$load of local cache) and still not match the die size of a P4? You might get equal GFLOPs at only 1/10 the P4 clockrate on this EE.

Also the memory subsystems aren't as far spread as you may think, either. You got the FSB of the P4 working at 133 Mhz in quad-data-rate and 64 bits wide while the EE sits on a 300 Mhz bus in single-data-rate and 128 bits wide. It's very close just by looking at the bulletpoint numbers alone. Efficiency in actual use, OTOH, who knows...it just works. Consider how many memory ticks (using a clock analogy) a 300 Mhz EE has available for every CPU tick, compared to how many memory ticks a 3 Ghz P4 has available for every CPU tick, and you realize the EE is in a far more favorable situation. It's darn lucky the P4 has the cache sizes it has (though that is surely by design, not luck, of course).

 

[Jabjabs]

Good point Randycat99, the EE is great for graphics for it's speed and cost. If a 3Ghz pentium 4 can do 12 Gflops that means in can render about 40 million pps if it doesn't have to do T&L work.

The CPU's are not comparable however because they work the graphics systems work in all the diffrent systems.

The EE generally does most of the work in the machine in term of graphics before the GS does the final render. The Gekko has to handle the geometry but very little else in terms of visuals. And the xbox handles even less due to the GPU's vertex shaders, I mean the XCPU can bearly produce 10 million pps and many games already push at least twice that number.

 

[clem64]

IBM's Gekko in GCN is also "32-bit" but it can work two 32-bit ops per cycle. Also it isn't x86 - it's PowerPC, which is pretty much known to be more efficient than x86 in general.

How is PowerPC more efficient?

Especially in games, believe it or not.

Why especially for games? Just curious.

My understanding of the difference in performance between the two chips is that, the theoretical numbers are very much in favor of the XCPU but that the architecture in which Gekko operates (bigger cache, bigger FSB, lower latency RAM) makes the real world difference negligible.

 

[Tagrineth]

Good point Randycat99, the EE is great for graphics for it's speed and cost. If a 3Ghz pentium 4 can do 12 Gflops that means in can render about 40 million pps if it doesn't have to do T&L work.

The CPU's are not comparable however because they work the graphics systems work in all the diffrent systems.

The EE generally does most of the work in the machine in term of graphics before the GS does the final render. The Gekko has to handle the geometry but very little else in terms of visuals. And the xbox handles even less due to the GPU's vertex shaders, I mean the XCPU can bearly produce 10 million pps and many games already push at least twice that number.

Well, first off, you're totally wrong. EE does T&L. Gekko and XCPU do not except in very specific cases. GameCube and Xbox offload the geometry to the graphics chip (Flipper and XGPU, respectively).

Second...

How is PowerPC more efficient?

Research.

x86 is one of the most poorly-designed ISA's ever; pretty much every x86 CPU so far has tried desperately to 'patch over' all the inherent pitfalls. Even the ISA's creator, Intel, is working to get rid of it, slowly but surely.

 

[V3]

Timna. Wasn't economically feasible. Look it up.

Maybe not before, but what about now and in the future ? They can make console like PC, that would be quite a performer.

 

[zurich]

Gekko and XCPU do not except in very specific cases. GameCube and Xbox offload the geometry to the graphics chip (Flipper and XGPU, respectively).

Considering that Flipper's T&L is fixed function, I'd think that quite a lot of transform and lighting ops are done on Gekko. I remember reading an interview with the IBM guys that make the chip, and they commented that every time they played Luigi's Mansion they could spot the lighting effects that were being done on the host CPU.