View Full Version : Xbox2 cpu performance?
Does anyone know what the computational power of Xbox2 cpu is, based on the leaked specs?
version
20-Dec-2004, 03:05
Does anyone know what the computational power of Xbox2 cpu is, based on the leaked specs?
168 GIPS
84 GFlops
Does anyone know what the computational power of Xbox2 cpu is, based on the leaked specs?
168 GIPS
84 GFlops
Compared to Athlon 64?
version
20-Dec-2004, 03:17
about 30 Gflops(?)
How did you calculate that figure of 84Gflops? Doesn't sound that great, only 84Gflops, not 100's of Gflops.
Megadrive1988
20-Dec-2004, 05:52
Xenon CPU is not likely to have massive performance like PS3 CPU.
Xenon CPU is estimated in the dozens of Gflops.
PS3 CPU is expected to be in the hundreds of Gflops.
Is it only in the dozens of flops ? I thought they were saying it was in the low 100 gflops ?
Anyway its best to see what both can actually sustain .
akira888
20-Dec-2004, 06:50
3.5 Billion clock cycles per second (or 3.5Ghz)
3 cores
1 Altivec unit (or whatever IBM calls it) per core
4 way SIMD on the Altivec unit
2 operations per FPU (FMADD)
-------
84 GFLOPs
Megadrive1988
20-Dec-2004, 07:13
Is it only in the dozens of flops ? I thought they were saying it was in the low 100 gflops ?
Anyway its best to see what both can actually sustain .
the Xbox 2 CPU, the only thing Ive ever read is that it'll be in the 84 to 90 Gflop range. not low hundreds of Gflops. and true, it's all about what a CPU can sustain,
That Xbox2 CPU sounds like an absolute beast. I'll be shocked if the PS3's Cell CPU just totally outguns it. Cell is completely unproven, PowerPC sure is proven and pretty refined I'd imagine.
Expectation, PS3 cell processor: four Gigaflops (Based on slide which shows "home server" with 4 cells, "home server"= apparently the deluxe version of PS3)
Which is probably why Sony turned to Nvidia to design a powerful gpu.
rabidrabbit
20-Dec-2004, 07:34
Four Gigaflops??? Isn't that less than PS2 (isn't PS2 6.4 Gigaflops??) and xbox!!! :shock: !! SONY Suxxas!!!
Is it only in the dozens of flops ? I thought they were saying it was in the low 100 gflops ?
Anyway its best to see what both can actually sustain .
the Xbox 2 CPU, the only thing Ive ever read is that it'll be in the 84 to 90 Gflop range. not low hundreds of Gflops. and true, it's all about what a CPU can sustain,
When i said low hundreds of g flop range i didn't mean 200gflops or something . I mean like 120gflops
Expectation, PS3 cell processor: four Gigaflops (Based on slide which shows "home server" with 4 cells, "home server"= apparently the deluxe version of PS3)
Which is probably why Sony turned to Nvidia to design a powerful gpu.
I hope your not wasting anyone's time with that 4Gflops crap. You do know that the current EE in the PS2 ~ 6Gflops, right?
I did some estimates here,
http://www.beyond3d.com/forum/viewtopic.php?p=430908#430908
In Gflops for 1 Cell and tri-core Xe CPU,
PS3 (CPU = 294 + GPU= 480) 2.4x > Xe (CPU = 84 + GPU = 240)
Anyway what does it matter right? It's all about the games? If your're a graphics whore then why don't you wait for all the consoles to be released then by the most powerful! ;)
Well, with the scaling calcs I did earlier with leaked specs etc,
Xe CPU(tri-core) = 84 GFlops @ 3.5GHz, 65nm (IBM)
Xe GPU = 240 GFlops @ 500Mhz, 90nm (TSMC)
PS3 CPU(1 CELL) = 294 GFlops @ 4.6GHz, 65nm (STI)
PS3 GPU = 480 GFlops @ 500MHz, 65nm (STI)
Where did u get gpu leaked specs from ?
Do you believe nvidia has the advantage in flops ? I highly doubt this is the case .
IN the pc sector the newly designed nv40 is trading blows with a modified r3x0 chip (the r420)
Not only that but u assume that the xe gpu is going to be on 90nm ? Why do you believe this to be the case ?
Well, with the scaling calcs I did earlier with leaked specs etc,
Xe CPU(tri-core) = 84 GFlops @ 3.5GHz, 65nm (IBM)
Xe GPU = 240 GFlops @ 500Mhz, 90nm (TSMC)
PS3 CPU(1 CELL) = 294 GFlops @ 4.6GHz, 65nm (STI)
PS3 GPU = 480 GFlops @ 500MHz, 65nm (STI)
Where did u get gpu leaked specs from ?
Do you believe nvidia has the advantage in flops ? I highly doubt this is the case .
IN the pc sector the newly designed nv40 is trading blows with a modified r3x0 chip (the r420)
Not only that but u assume that the xe gpu is going to be on 90nm ? Why do you believe this to be the case ?
All my assumptions are in that linked thread so I do not want to rehash them. If you follow the link back, you'll find more info.
As for the Xe GPU, it's from the original leaked diagram.
IIRC, there are 48 ALUs and each ALU can do 8 vector ops and 2 scalar ops simultaneously per clock cycle.
1 ALU = 10 flops per clock cycle.
48 ALUs = 480 flops per clock cycle.
Xe GPU @ 500Mhz = 0.5x480 = 240 GFlops.
I will have to look in that thread for the info i guess.
But anyway why do u assume that ati is going to decrease its clock speed ?
If the rumoured R500 Xenon GPU will have 48 unified ALUs from the leaked specs and is suspected to be manufactured on 90nm process, then I would expect the nVidia GPU to have 64 ALUs on a 65nm process and with a good dose of eDRAM. ...And possibly 2x32 pipeline configurations...though I couldn't tell you the benefits of 1x64, 2x32, 4x16 for what they would want to achieve with CELL, any guesses?
Your basicly just playing pocket pool
I will have to look in that thread for the info i guess.
But anyway why do u assume that ati is going to decrease its clock speed ?
If the rumoured R500 Xenon GPU will have 48 unified ALUs from the leaked specs and is suspected to be manufactured on 90nm process, then I would expect the nVidia GPU to have 64 ALUs on a 65nm process and with a good dose of eDRAM. ...And possibly 2x32 pipeline configurations...though I couldn't tell you the benefits of 1x64, 2x32, 4x16 for what they would want to achieve with CELL, any guesses?
Your basicly just playing pocket pool
Huh, I never said anything about ATI reducing clockspeed! :? Like I said, my assumptions are in that thread.
I will have to look in that thread for the info i guess.
But anyway why do u assume that ati is going to decrease its clock speed ?
If the rumoured R500 Xenon GPU will have 48 unified ALUs from the leaked specs and is suspected to be manufactured on 90nm process, then I would expect the nVidia GPU to have 64 ALUs on a 65nm process and with a good dose of eDRAM. ...And possibly 2x32 pipeline configurations...though I couldn't tell you the benefits of 1x64, 2x32, 4x16 for what they would want to achieve with CELL, any guesses?
Your basicly just playing pocket pool
Huh, I never said anything about ATI reducing clockspeed! :? Like I said, my assumptions are in that thread.
YOu have the xenon gpu listed at 500mhz that is a decrease in clock speeds .
X800xt pe is 520 . The x850 is what 560 ? Why would they go down ?
Guden Oden
20-Dec-2004, 09:12
Isn't this topic rather redundant? We're just regurgitating things we already know. No, actually, we're regurgitating things we've already SPECULATED! :D
We don't really KNOW anything, including if nextbox CPU will be 3.5GHz or Cell 4.6GHz, how many altivec units or PUs there will be, etc. And we definitely don't know how fast the GPUs will be clocked or how their internals will be arranged! :D
Well, here it is:
http://images.visualwebcaster.com/18267/25730/slide13.jpg
aaronspink
20-Dec-2004, 10:22
3.5 Billion clock cycles per second (or 3.5Ghz)
3 cores
1 Altivec unit (or whatever IBM calls it) per core
4 way SIMD on the Altivec unit
2 operations per FPU (FMADD)
-------
84 GFLOPs
Actually it probably should be:
Per Core:
Main FMAC = 2 F/C @ 3.5 = 7
1 4 wide SIMD FMAC = 8 F/C @ 3.5 = 28
Per Core total = 35 GFlops
Per Die
35 * 3 = 105 GFLOPS.
Though the SIMD performance could be higher if they include support for tree/combining operations.
aaronspink
20-Dec-2004, 10:26
In Gflops for 1 Cell and tri-core Xe CPU,
PS3 (CPU = 294 + GPU= 480) 2.4x > Xe (CPU = 84 + GPU = 240)
Anyway what does it matter right? It's all about the games? If your're a graphics whore then why don't you wait for all the consoles to be released then by the most powerful! ;)
I think you're probably overestimating the Nvidia GPU a little bit. I wouldn't be supprise if they were both design to be fittable on 90nm but primarily target for production die sizes on 65nm. If that is the case, I don't see the Nvidia GPU have 2x the shader resources.
Aaron Spink
speaking for myself inc.
cybamerc
20-Dec-2004, 10:35
Four Gigaflops??? Isn't that less than PS2 (isn't PS2 6.4 Gigaflops??) and xbox!!! :shock: !! SONY Suxxas!!!
Keep in mind that this is the guy who posts "news" from Spong.
passerby
20-Dec-2004, 11:01
Isn't this topic rather redundant? We're just regurgitating things we already know. No, actually, we're regurgitating things we've already SPECULATED! :D
Let me yell that for you again!
YOu have the xenon gpu listed at 500mhz that is a decrease in clock speeds .
X800xt pe is 520 . The x850 is what 560 ? Why would they go down ?
He's not the one listing the XeVPU at 500MHz, the leaked Xenon diagram showed a VPU clocked @500MHz+, so he just speculated using the infos he had, trying to get a rough estimation, nothing precise or anything. The XeVPU could, indeed be clocked @520, or 550MHz, but that would add another speculation, from him, to the equation, so he just used the numbers available.
By the way, the clockspeed of the R3XX derivatives is not really relevant, since the R500 is, supposed to be, a totally new architecture.
Expectation, PS3 cell processor: four Gigaflops (Based on slide which shows "home server" with 4 cells, "home server"= apparently the deluxe version of PS3)
Which is probably why Sony turned to Nvidia to design a powerful gpu.
Just ask yourself, why would they go through the troubles, make partnerships, spend billions to create and then use a Cell processor if it's weaker, by a large margin, than their actual EE?
I don't have to ask myself anything. It's you that has to face reality. The slide exists. Deal with it.
I think bbot is right.
It only makes sense, that SONY would join with IBM and TOSHIBA (not to mention nVidia)....and waste billions of dollars and years of development to make a .65nm chip, that will be slower then the PS2.
IT ALL MAKES SENSE NOW! :roll:
rabidrabbit
20-Dec-2004, 12:03
In the slide the 1 GFlop is a "Multi Cell Home Server", not a PS3.
As far as I understand it, a "Home Server" doesn't need as many GFlops as a PS3.
A "Home Server" maybe is basically just a HDD, used mainly for storage, with limited processing power. The PS3 that is connected to the "Home Server" does most of the processing needed in media applications and streaming data from the "Home Server".
If that slide is kept as a basis for speculation (though it maybe should not, as it is quite old already), a PS3 might fall between the "Blade" and the "Rack" (32 Gigaflop - 1 Teraflop)
It's no surprise really that some unmentioned posters don't want to see that slide beyond the "Multi Cell - 1 Gigaflops" point :lol:
People , we have a next gen dev ( DeanoC ) which say that the XeCPU is a clean sheet
Why do some people have a problem that XeCPU is a clean design, designed and built from scrarch for this one job. Blank sheet of paper, lots of engineers. As I have said many many times its not remotely related to a G5, its share a similar ISA but thats it and by the sounds of it so does Cell.
MS and IBM picked a set of features and built a processor around them
http://www.beyond3d.com/forum/viewtopic.php?p=430998&highlight=#430998
So why do you keep with this threads :?:
So why do you keep with this threads :?:
Expectation, PS3 cell processor: four Gigaflops (Based on slide which shows "home server" with 4 cells, "home server"= apparently the deluxe version of PS3)
Which is probably why Sony turned to Nvidia to design a powerful gpu.
I don't have to ask myself anything. It's you that has to face reality. The slide exists. Deal with it.
For laughs?
Sony to re-release PS2 in 2008 as PS3++, SOUNDS ABOUT RIGHT TO ME!
Expectation, PS3 cell processor: four Gigaflops (Based on slide which shows "home server" with 4 cells, "home server"= apparently the deluxe version of PS3)
I'm trying to scream but I can't breath, can anybody hear me? HAHAHAHAHAHAHAHAHA
Someone put this in their profile because I wanna see the reaction when Broadband Engine is well north of 500GFLOPS or dare I say a TFLOPS class microprocessor?
If PS3 CPU will be 2x faster than Xenon CPU...it would be an achievement!
I don't know about some here, but I certianly won't expect a 1 Tflop processor in PS3. They aren't going to be giving everyone a super computer for $400 american.
I don't know about some here, but I certianly won't expect a 1 Tflop processor in PS3. They aren't going to be giving everyone a super computer for $400 american.
I have a hard time believing it as well, but you never know. Guess we'll find out in due time. Same goes for Xbox2.
Tommy McClain
Mulciber
20-Dec-2004, 19:33
i think it takes like 256 2.8ghz p4s to reach 1.5 TFLOPS. 1 ps3 should do that easy
Joshua Luna
20-Dec-2004, 19:45
i think it takes like 256 2.8ghz p4s to reach 1.5 TFLOPS. 1 ps3 should do that easy
(Emphasis mine)
Why?
And what information gives us an indication that it should do 1.5 TFLOPs "easily" and what do you expect as a max?
Mulciber
20-Dec-2004, 19:49
i think it takes like 256 2.8ghz p4s to reach 1.5 TFLOPS. 1 ps3 should do that easy
(Emphasis mine)
Why?
And what information gives us an indication that it should do 1.5 TFLOPs "easily" and what do you expect as a max?
i was attempting to be EXTREMELY heavy on the sarcasm :(
Joshua Luna
20-Dec-2004, 19:54
Oh, hehehe. Sorry! :oops:
I have seen so many, "CELL will do 1 TFLOPs!!!" posts that I thought that you were serious. I was thinking maybe you knew something no one else did :)
Pointing out the obvious here.
But if Sony and MS split the workload between GPU and CPU like they did last time, then the PS3 CPU will have to ghave more performance than Xbox2 to be equivalent.
I don't know about some here, but I certianly won't expect a 1 Tflop processor in PS3. They aren't going to be giving everyone a super computer for $400 american.Oh, I agree, but there's to be something in between the TFLOP mark and "PS3 to be slower than PS2!!OMG!!!11!!!".
IMO the three next gen systems will sit reasonably close to each other when it comes to hardware. Of course there'll be some differences here and there, and I'd expect PS3 to have a slight advantage here based on the expected release date of the machine and that Sony will be selling it at a (huge?) loss, but I can't imagine anyone looking at Xenon's games and going "ughh, that looks like crap" (other than fanboys, I mean).
Edited by moderator
Stop trolling the forums. This is your last warning.
Expectation, PS3 cell processor: four Gigaflops (Based on slide which shows "home server" with 4 cells, "home server"= apparently the deluxe version of PS3)
Which is probably why Sony turned to Nvidia to design a powerful gpu.
Consider that slide wrong. IBM has already stated that a cell rack can reach 16 teraflops. Also get ready to eat crow on your PS3 cell estimate.[/quote]
This is first gen Cell people. Don't be shocked if it doesn't turn out to be all that.
The hype never comes true. Ever.
Peak FLOPs are garbage to judge usefulness anyway. Too generic.
Cell is an architecture from fairly low performance to high performance. A 4 GFLOPS Cell chip makes perfect sense giving the infomation publically disclosed.
Its probably the FLOPS rating for a PU without any APUs. The PowerPC PU core will have a 1 FLOP per cycle FPU at 4GHz for 4GFLOPS. It would represent the 'smallest' Cell configuration possible.
The point being with Cell that you can add extra PUs and extra APUs till you get the required FLOPS.
As PS3 Cell is going in a 3D console, it will have a considerable FLOPS rating due to a bunch of APUs. The exact FLOPS rating of the PS3 isn't currently publically disclosed, so I can't say any more.
The exact FLOPS rating of the PS3 isn't currently publically disclosed, so I can't say any more.
I definetively set my expectations on the 200+ Gigaflop/s mark with
a one PU + 8 SPUs CPU.
vBulletin® v3.8.4, Copyright ©2000-2010, Jelsoft Enterprises Ltd.