PureBasic Forums - English

hmmm..... Honestly i'm not all that surprised...


I do have to ask this question though:

Why are so many of you convinced that PPC processors are faster, even though in every 3rd party* benchmark i've seen x86 PC's beat the crap out of Apple's G5's?

*that is to say, Apple's benchmarks show them on top... like always, but other reviewers (magazines, webpages, etc) are consistant about apple losing.


I mean.. there's "Theoretical speed" then "practical speed" PPC has the former, X86 has the latter.

The powerpc processor (on average) at 1 ghz is comparable to an average 2 ghz x86 (I say average because there are so many combinations of cache and fsb).

If you get x86 processors with 2MB cache and 800mb fsb and 3.6 ghz internal clock speed & also add the almost second core of HT then you will floor a PowerPC, but your processor will cost a small fortune.

I have never been a fan (b4) of the PowerPC (I really like MIPS), but the future for this chip does look slighly more rosy than the x86. The PowerPC had gained multithreading and multicore for the new line of chips for the XBOX 360, Revolution and the PS3. They also all run at least 3Ghz, some even have direct access to 3.2Ghz fsb RAM!!!!

That is not even taking into account the video processing abilities of the CELL vector processors added in Sony's version. To get an idea of the power of this chip regarding video apps, Toshiba have demos it decoding/displaying over 20 dvds all in software using a single cell processor all at the same time without breaking a sweat! Don't forget that hundreds millions of these processors are now about to be made, far more that intel or amd processors combined. The cost of these chips is going to be so small (they will be in almost all hdtvs made in japan) that based upon price/performance the x86 cannot compete. Also remember that Cell processors can be tagged together to produce even greater speeds if needed. So for the same cost as a fast x86 you will be able to get a superfast cell/powerpc system (soon).

The main reason I think that Apples move is a crazy one is that the future right now is a very uncertain one. I think its just because they are annoyed that IBM now has more interest in microsoft/nintendo/sony/toshiba than them. Steve Jobs is crazy after all.

hmm... i think your estimates about the speed are a bit off.

http://www.pcworld.com/news/article/0,a ... g,8,00.asp

Basically this is a roundup of PC's and 2 of the g5 processors... sorry it's slightly dated.

Basically, a 2.2ghz Athlon64 beat apple's 2.0ghz G5 in most tests.... and the 2Ghz A64 also beat the G5's in a number of tests (rated at the same MHz)

Anyway, i don't really buy into the mhz/ghz. it's been proven time and again that it is not an accurate measurement of processor speed. Basically, if brand X has a 1ghz chip that's as fast as the competitors chip at 2ghz... it doesn't mean anything if the competitor has a 4ghz model currently selling, for cheaper i might add. (at least the system price)


now... All that aside though... One has to question why it is that these new chips that IBM has been making have not actually ended up in apple computers. There must be other technical barriers that apple ran into, after all, apple was already having issues clocking the G5's up to 3ghz and that's just a single core. as it is they're top end is sitting at 2.7ghz, after the help of highly specialized cooling system (h2o)

I based my timings on average cycle time per operation to do a particular job. In this respect the PowerPC has about a 2x advantage against the x86. Much the same as MIPS.

I think that the speed problem must be in the support chips that are provided and/or the memory interface and/or the actual os involved.

The is no other possible explaination because the machines (average powerpc) at instruction level (and number of instructions to do an operation) are much faster (than the average x86).

It would be ironic if Microsoft bring out an own brand PC, based upon the PowerPC and abandoning the x86 (they must have considered the x86 for xbox2 and found it too expensive/under powered). And Apple bring out a x86 Mac and abandon the PowerPC. Now that would be funny

aszid wrote:I do have to ask this question though:

Why are so many of you convinced that PPC processors are faster, even though in every 3rd party* benchmark i've seen x86 PC's beat the crap out of Apple's G5's?

To be honest, I don't give a damn about which architecture is fastest. A computer is only a tool, it's what you do with it that counts.
For me a 1.3Ghz G4 is more than enough grunt for what I need it to do, just the same as my 040/40 Amiga is, and my K6-450 PC.

The thing I don't understand about this announcement, is that they said they're switching to Intel cos they're faster than IBMs chips. If that's the case, why are they putting them into the low-mid range Macs first and not the high end ones that people buy for serious video/music/gfx work?

I waited to see how true this would turn out, thinking it was another BS rumor. Surprised to find out how true it is. All I can say is, watch Apple sales drop quick. It will be interesting to see if Apple will even be able to make this transition without loosing alot of current and would be users. There will be no point on buying a Mac until 2007 or later as I'm sure there will come issues.

Check OFF TOPIC thread, this is all true!

viewtopic.php?p=92000

Num3 wrote:Check OFF TOPIC thread, this is all true!

Yes, re-read this thread, it's already been said

traumatic wrote:

Num3 wrote:Check OFF TOPIC thread, this is all true!

Yes, re-read this thread, it's already been said

If you guys are commenting on my post, you might want to read more carefully. As I stated that I know it's true now.

dell_jockey wrote:Fred,

although the OS-count remains three, there's one processor architecture less to target, so this should result in some workload reduction.
Perhaps a good time to think about targetting PocketPC's ?

Oh! Yes... I agree...

Fred, I know it is painful to see work gone. Looks like I am the very few who is happy about the change as this means more market and profit for Intel or higher stock price... Then I can sell my options...

J. Baker wrote:If you guys are commenting on my post, you might want to read more carefully. As I stated that I know it's true now.

I said that, didn't I?

I deffenitly think fred has not wasted one minute of work because when the cell is PPC he will need PB to work with it wouldn´t it be cool to see PB working with a ps 3 or a computer or cell phone using the CELL
bye Nik

Guys, the PPC CPU used in the MAC isn't the same as the Cell CPU that will be in the PS3 - I've seen nothing to indicate that there is any relationship with PPC / Power5 or Cell CPUs other than that IBM did the design work - so although of course you're right in saying that there are huge economies to be made by putting Cell CPUs into everything from a HDTV to a games console, that won't affect the Apple because they use a different CPU - the PPC, not the Cell.

BTW - there's a scary estimated stat at the end of this post

Cycle for Cycle, the PPC and Power5 CPUs used in IBM Unix Servers and Apple MACs outperform Intel, but thats not the whole story. By building closed architecture systems, Apple don't weigh the box down with bloat as MS do. Apple also do clever things like using OpenGL to render the GUI - its not an efficient use of the GPU, but its better than burdening the CPU.

Similarly, the current XBox with its little 866MHz Pentium CPU flies because its a closed architecture box (even though most of the components are standard PC stuff). all the other compatability and non-essential stuff is removed

The XBox 360 will however be using 3 x 3GHz Watercooled PPCs. I think the PS3 will use 8 Cells (but I'm not sure - pls correct me)

The Cell based PS3 is expected to be <> 10x the performance of a current 3GHz P4 PC for Multimedia.
The XBox360 is expected to be about <> 5x performance of current 3GHz PC for Multimedia.

BUT

The xBox360 has 5 times the (280GB/s) memory bandwidth and a more powerful GPU than the PS3 (48GB/s). For general workloads (inc games) the xBox360 is expected to be <> 3 times more powerful than the PS3.

But even these expectations are very broad. And are based on specific Multimedia workloads. Of course, most workloads come into a more general category (even games) - and here, the PPC will outperform the Cell CPU (cycle for cycle).

The strength of the Cell is the linear scalability of the CPUs when clustered tightly together and its Multimedia performance.
The strength of the PPC is the scalability of its CPUs for general workloads,

If you want to view some *serious* benchmarks, take a look at:

http://www.tpc.org/tpcc/results/tpcc_perf_results.asp

- for comprison:
a 32W 1.9GHz IBM Power5 595 AIX5 scores 1,601,785
a 32W 2.0GHz Pentium Xeon UNISYS ES7K W23K scores 234,325
- thats one hell of a difference whichever way you look at it...

IBM with the PPC (actually Power5) take the top spots for performance. The good news for us is that like in F1 racing, developments at the top-end trickle down to consumer devices (like PS3's and xBox360's) and the economies of volume manufacturing make the top end systems economically attractive to the Corporate Buyers as well which results in more development.

IBM has had the top end all sewn up for a while (since HP threw everything away on Itanium). The Xbox360 / PS3 could do it for the bottom end.
Meanwhile Intel and AMD will slug it out in the middle ground for desktops and low end Servers. I agree, if Steve Jobs wants to get tangled up in that arena, he must be nuts...

- Blimey that was a long post (apols) - well done anyone who stayed awake to read to the end

The scary estimated stat is this:

If the IBM 32W 1.9GHz p5 Model 595 is over 7 times more powerful than the 32W 2GHz UNISYS ES7000. Then the way I work it out is that the XBox360 with 3x3GHz CPUs could be 21 times more powerful than the top of the
range 3GHz Pentium Desktop PC you're using now!!!!

DoubleDutch wrote:The powerpc processor (on average) at 1 ghz is comparable to an average 2 ghz x86 (I say average because there are so many combinations of cache and fsb).

DD, just where are you getting this stuff about the Cell? I think you're reading far too much fluff about the cpu than real nuts and bolts documentation. The reason Apple didn't go with Cell is because it really is a primo vector processor. Its obscenely fast for some things, and rather average for the majority of other things general purpose computers are used for. As a games machine it will be aweseome due to its 8x vector units. However, its been noted time and again that even todays general purpose computers spend a good amount of their time just waiting for something to do. Unless you are doing high-end computational calculations, the benefit and main reason for having a cell processor architecture are lost.


And, one could go on for a quite a bit on how the cell is not quite as easy to program as PPC, however I'd just be regurgitating what has come out since Apple's announcement.


Anything that is useful on the Cell can and will be pulled into x86-64 if Intel or AMD want it there. From what I've read, running a general purpose OS on Cell would be like commuting to work in a dump truck. An almost infinate amount of torque, but functionally no better than your average car in daily use.

And, with Cell pretty much ready to go in the near future 6 months or so, Job would have waited if Apple thought it would be a better chip. Even with Job's at the helm, theres a reason why Apple didn't go with Cell. If moving OS X to Cell would have been much easier and continued to allow Apple to remain relatively proprietary and sheltered, you gotta think there's a reason they still decided to jump the PPC ship.

And, how would doing an ARM port help Fred with income? I still think he should dump Amiga, because OS 4 and MOS are dead ducks, especailly now if you read the Amiga boards.

The main advantage of the cell processor compared to the normal ppc is the aspect of cost. This things are going to be cheap and (with os support) will be capable of sharing tasks. Most people use hi-end pc's for games and entertainment, not word. The 8 "units" in the cell are not just vector "add-on" chips to the ppc (like the vector usits in the ps2 mips), they are actual processors in their own right. They can be configured to work in several different ways. As a collective, independant or cooperative, among others. They can also be used to speed up things that most computers have to use the cpu or have in their sound of gfx card. Toshiba has demonstrated this with the cell setup for decoding mpeg streams. One advantage over other processors is the memory interface to main ram, its over 3ghz! Apple could have used this cheap multi-cpu ppc and the added advantage of the cheaper cell technology to also lower costs.