Intel Conroe: From a Console's Perspective

[Edge]

Conroe is Intel's next generation architecture just like Cell is the next generation of IBM/Sony PPC Architecture. How would the performances compare if say Xbox 360 had a next generation Intel Conroe (single or triple core) vs a PS3 Next generation PPC Cell Processor. Would it make much of a difference over PPC Tri-core processor? Anyone have some knowledge of this new intel architecture?

How does this matter considering Conroe is coming out a full year after the Xbox 360 release?

Other than that, Conroe looks awesome, as I'm sure you will get higher instructions/per second throughput than any other achitecture out there today. I would give 360's CPU the nod on floating point though.

 

[Megadrive1988]

I don't expect Intel CPUs to get interesting for potential console use (apple ?) until Intel has implemented its Platform 2015 into an actual CPU or CPU-GPU.

 

[Inane_Dork]

A Conroe-like CPU would make for a really easy-to-start-developing console with less room to grow into. It's a trade-off.

 

[pjbliverpool]

A dual core conroe would wipe the floor with Xenon in just about anything. Does anyone really believe otherwise?

 

[Laa-Yosh]

AFAIK Xenon's VMX units are quite a bit better than Conroe at mass processing FP data. And its integer performance may be higher, too. Then there's the memory bandwith to the north bridge; even though Xenos may consume a lot of the main RAM bandwith, there's still a separate bus that the two chips can utilize for various things like procedural geometry and textures.

 

[Guilty Bystander]

A dual core conroe would wipe the floor with Xenon in just about anything. Does anyone really believe otherwise?

Well the Xenon can do 90.000DMIPS and 115,2GFlops.
Which means the Xenon is almost 3 times better at Integer and 10 times better at Floating Point than an Amd 64 FX60+.
Seeing the Conroe is about 40% better at best.
What makes you think the Conroe is able to touch the Xenon at all?

Please remember both Cell and Xenon are made to be able to run very heavy AI, heavy physics (comparable to those done by the Ageia PPU) and then some and still run a graphics pounding game with all that at 30-60fps.
Xenon and especially Cell can crush any PC CPU out there with ease.

Just remember it took the PC market till 2005 to finaly being able to match the Emotion Engine (PS2 CPU) in power.

 

[Edge]

A dual core conroe would wipe the floor with Xenon in just about anything. Does anyone really believe otherwise?

I agree. It's going to do more work per cycle than 360's cores, and with it's larger cache, it's going to have more localized data. Plus compiler technology is far more advanced for x86 than the PPC architecture.

 

[Robert.L]

Well the Xenon can do 90.000DMIPS and 115,2GFlops.
Which means the Xenon is almost 3 times better at Integer and 10 times better at Floating Point than an Amd 64 FX60+.
Seeing the Conroe is about 40% better at best.
What makes you think the Conroe is able to touch the Xenon at all?

Please remember both Cell and Xenon are made to be able to run very heavy AI, heavy physics (comparable to those done by the Ageia PPU) and then some and still run a graphics pounding game with all that at 30-60fps.
Xenon and especially Cell can crush any PC CPU out there with ease.

Just remember it took the PC market till 2005 to finaly being able to match the Emotion Engine (PS2 CPU) in power.

Please tell me your joking....

 

[j^aws]

Well the Xenon can do 90.000DMIPS and 115,2GFlops.

Source for the mips?

For the record, that 115.2 Gflops peak is PR BS because it's higher than what it's theoretically possible. The recent leaked doc states 77-90 GFlops theoretical peak.

 

[Guilty Bystander]

Please tell me your joking....

Well I'm not cause the Emotion Engine had to calculate everything for the Graphics Synthesizer and calculate everything else and it still managed to do God of War, ICO, Shadow of the Colossus etc.
Emotion Engine not powerfull and I'm joking?
I think not!

An Amd 64 4000+ can do about 6GFlops while the Emotion Engine can do 6,2GFlops realtime.

Source for the mips?

Can't give you a link of it but I read it through this forum.
Was an article stating each PPC in the Xenon could do 30.000DMIPS.

For the record, that 115.2 Gflops peak is PR BS because it's higher than what it's theoretically possible. The recent leaked doc states 77-90 GFlops theoretical peak.

For the record, the 115,2GFlops is theoretical just like the 40GFlops of the P4 955EE or 25GFlops of an Amd 64 FX60+.
The 77-90GFlops however is a realtime PEAK.
Which means it was done realtime.
Just like the 139-155GFlops done by the Cell.

An Amd 64 FX60+ however can only do about 10GFlops realtime PEAK and the P4 955EE can do about 12GFlops realtime PEAK.
Which can be measured with SiSandrasoft2006.



I don't care which way anyone wants to measure but there's no way in hell 77-90GFlops vs 10 or 12GFlops can be thought as less.
The exact same reason why Xenon can cope with heavy physics and everything (meaning gameplay conditions and 5.1 sound done by Xenon) at playable framerates while any PC CPU can't even sustain 1fps as shown by the CPU tests of 3Dmark2006.

 

[Ike Turner]

Just remember it took the PC market till 2005 to finaly being able to match the Emotion Engine (PS2 CPU) in power.

Ok..

 

[Guilty Bystander]

I agree. It's going to do more work per cycle than 360's cores, and with it's larger cache, it's going to have more localized data.

What does it do more?
It does less MIPS/DMIPS (Integer) and less Flops (Floating Point).

Does a larger cache mean better performance?
Tell about a P4 3,8GHz 2MB winning from the Amd 64 3800+ 512KB or a P4 3,0GHz 1MB from an Amd 64 3000+ then please.
That's even neglecting the fact the P4's L-2 cache has a 256bit interface and runs higher frequenties (an Amd 64 L-2 cache only has a 128bit interface and lower frequenties).

It could have more localized data but does it?
Theoretical means nothing the fact is that Intel Conroe will never be pushed anything more than 50% of what it's capable of while the developers will try to get each drop of power out of the Xenon.

Maybe the Conroe is better at some things like word processing or running Windows XP/Vista but who cares?
Xenon is better at running heavy AI, heavy physics, gamecode and everything else that make a Xbox 360 game or application.
That's where it's made for and where it does best and nothing else matters.
For Cell however it's a whole different story because it will venture into more than just the PS3.
But as with the Xenon the story for the Cell remains the same.
It does great (better than any CPU money can buy) where it's made for (running heavy workstation, server, clusterserver, supercomputer applications smoothly with a single Cell and running PS3 games) and where it does best and nothing else matters.

Plus compiler technology is far more advanced for x86 than the PPC architecture.

 

[Guilty Bystander]

Ok..

Well I'm not cause the Emotion Engine had to calculate everything for the Graphics Synthesizer and calculate everything else and it still managed to do God of War, ICO, Shadow of the Colossus etc

I wanna see a P4 2,8GHz or Amd 64 2800+ do that with a rasterizer processor because that's all the Graphics Synthesizer does is draw Pixels.

 

[Guilty Bystander]

Plus compiler technology is far more advanced for x86 than the PPC architecture.

That's the only thing the PC market has got going for it.
Using a 20 year old architecture with constantly hold up to date compilers.
So it's simple and there are no suprises but my god it's slow and dated.

It would be better for the PC market if they just gave up that dinosaur architecture.
Otherwise Cell like revolutionairy leaps can never be made.

 

[NANOTEC]

Just remember it took the PC market till 2005 to finaly being able to match the Emotion Engine (PS2 CPU) in power.

The only reason why PC cpus were not able to do a lot of FP math compared to EE at the time is due to the fact most programs for Windows don't require lots of FP arithmetic. The only reason why the EE was able to do 6 GFLOPS at the time was the fact the GS didn't do any of the TnL so the EE had to be designed to take on the extra load. Also there were other PC cpus during the time of EE that had high GFLOPS performance. In fact the mobile P3 in Xbox was capable of 3 GFLOPS and that wasn't even the fastest PC cpu at the time.

 

[Robert.L]

You really haven't got a clue .......


And Mr Jaws is right xecpu's 115 gflops number is bunch of bull MS counted shuffling as "flops" witch makes it around 115Gflops , theoretically xecpu does 77 gflops


and it looks like Sony inflated ppe's flop's number also

 

[Robert.L]

damn typo's ..please some one tell me if there is an edit option somewhere

 

[j^aws]

An Amd 64 4000+ can do about 6GFlops

I'd like to see a source or your derivation for this.

Can't give you a link of it but I read it through this forum.
Was an article stating each PPC in the Xenon could do 30.000DMIPS.

Well if you can't source it then please show a derivation or refrain from throwing that number around because I've never seen it before on this forum.

For the record, the 115,2GFlops is theoretical

It's not theoretical. The number cannot be achieved even with a theoretical derivation. They just summed up all instructions without taking into consideration what you could actually issue in a given cycle. It's theoretically impossible. So it's not theoretical but actually BS.

just like the 40GFlops of the P4 955EE or 25GFlops of an Amd 64 FX60+.

I'd like to see links or derivations please.

The 77-90GFlops however is a realtime PEAK.

That's the actual theoretical peak quoted. If by realtime peak you mean real world peak, then no it's not. The above IS a theoretical maximum as stated in the docs. Nothing to do with realworld numbers which would be even lower.

Which means it was done realtime.

No, it's theoretical.

Just like the 139-155GFlops done by the Cell.

Cells theoretical peak is 205 Gflops and not 218 Gflops. And the above figures look like they are from some IBM benchmarks (synthetic tests), at least they're not theoretical!

I don't care which way anyone wants to measure but there's no way in hell 77-90GFlops vs 10 or 12GFlops can be thought as less.

If you're gonna compare, at least know what your comparing. Mixing PR BS peaks, with theoretical peaks, with synthetic peaks, with realworld peaks! That's 4 different types of peaks and makes your basis for comparison null and void...

The exact same reason why Xenon can cope with heavy physics and everything (meaning gameplay conditions and 5.1 sound done by Xenon) at playable framerates while any PC CPU can't even sustain 1fps as shown by the CPU tests of 3Dmark2006.

You'd have to run the same bench on Xenon to compare that FPS.

Anyway, Xenon should do better at certain tasks but OOOe CPUs also have their strengths...

 

[swaaye]

Theoretical numbers are so damn useless it drives me insane to see so many on console forums constantly quote them. OMG. It's like calling up Linksys and asking them if their routers are good. Great unbiased opinion there! (my neighbor did just that too lol. Poor guy is a wee bit limited)

I found Ars Technica's write up by Hannibal to be very candid and informative, and realistic. If you really want to have an idea of what the hell Xenon really is/does, go read that. It's not exactly easy reading, but the guy does have a clue.
http://arstechnica.com/articles/paedia/cpu.ars

Xenon is an interesting chip with huge potential performance. Whether devs can get it there or not is what's up in the air. What's interesting is how far it's actually de-evolved compared to even the P3 in Xbox. It tells how wild chip designers are getting to get mad performance for cheap.

BTW, I giggle over all the red reputations in this console area. LOL.

 

[SugarCoat]

They may own the rights to that particular implementation of those chips and have the ability to fab it wherever they please (including getting it shrunk or whatever), but I doubt they could do any suing if IBM or ATI decided to use modified versions of those elsewhere -- who knows though.

I think the whole "owning the IP" is a bit overplayed -- they likely didn't need to own the entire IP (which would be sort of impossible on the CPU side, as its still a power core, which is very much in IBMs control still) to get the benefits they needed (having more control over it -- getting it fabbed and shrunk wherever and whenever they please).

I'm sure MS got themselves a good deal (surely better than the deals they had with Xbox1), but I have my doubts that IBM/ATI really gave up anything. It just isn't really necessary for either side.

i'll try to clear this up on the gpu side. During Xbox production, they had a deal with Nvidia where Nvidia produced and supplied the chips per their request as well as paying Nvidia a royalty fee for each console produced. The console was selling under expectations and MS felt they could say no to what they requested, Nvidia got pissed, they settled, Nvidia won. Fast forward to Xbox360 and what we have is a deal where Microsoft completely controls the production and orders of Xenos chips. ATI still recieves a royalty for each Xbox 360 sold as well as holding design rights. Microsoft basically had ATI make them a mold from which they get their stock of endless Xenos, the only thing different is that ATI makes a little bit less because Microsoft is controlling all production of the chips, its that simple. At any time MS can produce more, produce less, halt production or shrink the chip to reduce production costs and that would be the end of it. That is the right that they own.