Gamecube had more transistors than Xbox?

[sonyps35]

http://www.pcperspective.com/images/reviews/150/transistors.jpg

According to this if you add CPU+GPU up it did.

Wow, they didn't get there moneys worth out of that, as Xbox was widely considered the most powerful console of that era, by a reasonably good margin.

Also, I did a double take that Pentium 3 (Xcpu was a P3/Celeron cross, basically p3)was only 9 million transistors, but sure enough according to wikipedia it was 9.5 million.

I believe the new Fx-62 dual core monsters from AMD are like, 230 million transistors. With 1MB cache per core. That is cell territory. How time flies..

Like previous FX processors, the FX-62 contains 2MB of L2 cache total (1MB cache per core) with 256KB total L1 cache (64K - L1 instruction + 64K - L1 data cache per core). The chip sports a 230mm2 die size and contains 227.4 million transistors. Despite the higher clock speed, max power is the same as FX-60 at 125W

http://www.firingsquad.com/hardware/amd_athlon_64_fx_62_5000/page3.asp

This could fly in the face of how efficient GC hardware was considered I suppose. It seems much of that stigma was due to microsofts losses with Xbox. Were those all down to bad royalty deals on MS part? Or are perhaps some of the actual Xbox parts larger die sizes than the transistor counts hint at?

Also the 64 MB RAM was a factor in favor of the Xbox that may have played some part in it's superior reputation, ad I wonder how many transistors GC dedicated to EDRAM? 3MB? That would be, 24 million right? That's almost half the size of the GC GPU by that measurment.

 

[Guden Oden]

According to this if you add CPU+GPU up it did.

Because of the embedded memory on the GPU. Though if you add in the extra RAM of the xbox it will regain the lead. This is an exercise in futility however, there's no actual point in comparing transistor counts between different hardware architectures.

Also, I did a double take that Pentium 3 (Xcpu was a P3/Celeron cross, basically p3)

"XCPU" is a mobile celeron, period. End of story. Nothing differs that motivates it carrying a different name than what intel's already given it.

This could fly in the face of how efficient GC hardware was considered I suppose.

What are you talking about? You're merely comparing transistor counts, which says nothing about how well any particular piece of hardware is working. Gamecube is considered an efficient platform in part because of its embedded 1T SRAM frame and texture buffers and the 1T SRAM main memory, which has about ten times faster random access time than regular (DDR) DRAM such as that used in the xbox.

Also the 64 MB RAM was a factor in favor of the Xbox that may have played some part in it's superior reputation, ad I wonder how many transistors GC dedicated to EDRAM? 3MB? That would be, 24 million right? That's almost half the size of the GC GPU by that measurment.

No, it is not, because DRAM arrays are much denser than general logic. 1T SRAM consists of multibanked DRAM plus a little logic and some SRAM buffers to hide most penalties of memory refresh and first-access delays.

I'm not really sure what genuine point you're trying to argue here, but it's not going very well... Concentrating on transistor counts ONLY generally doesn't give much useful room for discussion.

 

[sonyps35]

Though if you add in the extra RAM of the xbox it will regain the lead.

Not really, since RAM is totally different than..transistors in the CPU/GPU.

"XCPU" is a mobile celeron, period. End of story. Nothing differs that motivates it carrying a different name than what intel's already given it.

True. But in terms of desktop CPU's of the time the Celeron was crippled from the P3 in two ways, cache and front side bus. Xcpu had the higher speed FSB of the P3, but the lower cache of the Celeron. So I would say perfomancewise compared to CPU's of the day, it was basically directly in the middle, that's why I dont agree with calling it a Celeron. Even though technically, it was a mobile chip that Intel chose to arbitrarily call a Celeron.

I'm not really sure what genuine point you're trying to argue here, but it's not going very well..

And why is that? I suspect I know why you think that, and for your veiled hostility..anyway, I just happened across the chart and thought it was an interesting thing that I had not known before.

 

[Urian]

Gamecube is an equivalent to a Pentium3 1.2 Ghz with a GeForce 2 Ultra.

Xbox is an equivalent to a Pentium3 700Mhz with GeForce4 Ti4200

 

[Gubbi]

Gamecube is an equivalent to a Pentium3 1.2 Ghz

Seriously doubt that. It is neat, no disagreement there, but 1.2GHz P3 is just bollocks.

Cheers

 

[Shogmaster]

Seriously, a PPC750CXe that Gekko was based on was about par with or just slightly more powerful than equivilent speed Tualatin P3s. 500MHz G3 was about equal to a 650MHz P3 in Photoshop version of the day for instance.

I'd say GC was equivilent to a 500MHz G3 Mac with a Radeon 9000 and XBox is equivilent to a PC with Tualatin Celeron @733MHz mated to a GeForce 4 Ti4200 power wise, but of course such machines would have had tons more RAM than the respective console counterparts.

 

[sonyps35]

I thought the tulatin pentium CPU's came much later than Xbox..like they clocked up to 1.4 GHZ..not that it really matters in performance terms. I think I recall Tulatins basically being the afterthought P3, after P4 already came along.

The link appears dead..imageshacked..

http://img132.imageshack.us/img132/4963/transistors8kz.jpg

 

[London Geezer]

Sorry i must have missed the memo explaining why counting transistors is actually fun or even useful...
Manufacturers have proved time and time again that higher transistor counts don't always imply higher performance (Nvidia and ATI have shown that).

 

[sonyps35]

Sort of like Gflops, it is a broad measure, and I certainly enjoy knowing it.

Although I consider die sizes perhaps a bit more telling..

 

[Rodéric]

Comparing transistor count in consoles in as meaningful as counting grains of sand on a beach...