Talk:PlayStation 3/Archive 2

From Wikipedia, the free encyclopedia

You have new messages (last change).
Archive This is an archive of past discussions. Do not edit the contents of this page. If you wish to start a new discussion or revive an old one, please do so on the current talk page.

Contents

Regarding FP perfomance

It's more or less meaningless to add up the FP performance of the PPC core, the SPEs and the GPU: they do three different things. The PPC core is a completely general-purpose CPU; the SPEs are DSP-like vector accelerator units, which will be programmed in a DSP style, and probably accessed through library APIs from the main CPU (build a processing graph, make a pipeline, initialize, run, sync, stop... that sort of thing), or even be completely transparent as part of the sound / OpenGL APIs; finally the GPU is more-or-less a completely dumb I/O device that is fed command streams from the rest of the system. I think the interesting thing about the PS3 is its much greater support for DSP-like grunt-power programmability upstream from the GPU.

Whilst the PS3 has less general-purpose CPU power than the Xbox 360, it has more GPU power, with much, much, more DSP-like vector processing power behind it. This has interesting implications for physics modelling and very complex scenes, where this is exactly the kind of processing power needed: I confidently expect the PS3 to overwhelm the Xbox 360 in terms of game realism if this processing power is brought to bear in the right way. -- The Anome 10:21, May 18, 2005 (UTC)

Though I do agree with your ideas, I think we must be careful of letting our own estimations and predictions creep into the article. While I personally agree with the new wording added regarding the PlayStation 3's architecture, it seems to contain some weasel words and I think we would be best served if we can find concrete examples of speculation by "trade press reports." Again, I'm not against the information that was added (far from it), but I think it needs to be backed up by some external resources to avoid problems in the future. -- uberpenguin 10:40, 2005 May 18 (UTC)

What is SPE?

Where the specs say:

7 x SPE @3.2GHz

What does the SPE stand for? The PSP has 8 SPE@3.2GHz, how does this relate to XBox 360's triple core at 3.2GHz?

I agree, can someone define SPE and perhaps even link it to the Cell processor where this info belongs?

"Synergistic Processing Element" -- called SPU by IBM. -- The Anome 10:08, May 18, 2005 (UTC)
I've seen IBM call it by both names in various whitepapers; SPE and SPU ("Streaming Processing Unit," I think) depending on what suits them best at the time. -- uberpenguin 10:36, 2005 May 18 (UTC)
The Cell Processor with 8 working SPE's has 234M Transistors. The Cell Processor in the PS3 ships with a disabled SPE, meaning 7 working SPE's is listed as having 234M Transistors. Although this is technically correct, it doesn't actually state the number of actual working transistors and is misleading information. http://www.gamespot.com/news/2005/06/20/news_6127799.html "Kutaragi explained that ignoring one SPE as a redundancy will improve the chip's production yield and allow costs to drop dramatically. In other words, Sony can ship a Cell chip with one defective SPE (out of its eight) as a working product, since the chip only uses seven SPEs to begin with." This means there is no possibility of using the transistors on one of the SPE's. Sony is using the word "Redundancy" in terms of increasing production yield, the disabled SPE will not be doing any computations. Therefore the transistor figure should be updated to reflect the number of working transistors, which is approximately 216M.--Thax 23:25, 22 Jun 2005 (UTC)
I understand your reasoning, but it isn't the purpose of Wikipedia to include our own extrapolations. Where do you get the 216 million transistor number? This transistor count is really a trivial bit of data anyway, but I'd at least like to know if you have a source that can verify your number. -- uberpenguin 02:22, 2005 Jun 23 (UTC)
Thank you for your response. My initial calculations were based on the surface area, however I was able to get a more accurate number of 21M thus reducing the transistor count to 213M. (Although this figure doesn't include the non-functional transistors in the EIB.) "Each SPE is made of 21 million transistors: 14 million SRAM and 7 million logic." -http://arstechnica.com/articles/paedia/cpu/cell-1.ars/2
I think that it would be useful to include a reference to the working transistor at the very least. The 486sx included a math co-processor, however to state that the 486sx had a math co-processor without qualification would be misleading as it was disabled at the factory to achieve better yields. http://en.wikipedia.org/wiki/486SX --Thax 02:58, 23 Jun 2005 (UTC)
Though I'm still somewhat wary of including figures that are the result of our own calculations, I don't see any real reasons to argue its inclusion. Go ahead and add the information back in with a brief note explaining it. -- uberpenguin 14:17, 2005 Jun 23 (UTC)
Sorry I should have been more clear, the 21M figure is for the SPE is actually from the Arstechnica article. --Thax 15:47, 23 Jun 2005 (UTC)

Missing 162 GFLOPS

Expected total Cell processing power is stated to be 218 GFLOPS, and the GPU is said to have 1.8 TFLOPS. On the other hand, the article claims the overall performance is 2.18 TFLOPS. But 0.218 + 1.8 = 2.018 and certainly not 2.18. Is this simply a calculation error, or where does the number 2.18 come from?

If noone can give good arguments for why it should say 2.18 TFLOPS, I shall change the overall performance to 2 TFLOPS, which is also what Sony says it is.

Grahn 14:52, 18 May 2005 (UTC)

Sony said in the press release linked on the article that it is 2.18 TFLOPS total. Frankly I have no idea where this number comes from and it seems like a creative marketing fabrication. I tried to remove it before, but other users want it to stay for now. See the discussion about that up the talk page a bit. -- uberpenguin 15:29, 2005 May 18 (UTC)
From what I can read, it says "System Floating Point Performance 2 TFLOPS" in the linked press release. I only skimmed it very quickly, though, so it may be that is says 2.18 elsewhere in the text. However, when I searched for the string "18", only a date and the the value "218 GFLOPS" came up. Anyhow, assuming this contradictory information exists, it still seems wiser to me to go for the more logical value of 2 TFLOPS. Especially since is has only one significant digit, and doesn't rule out the possibility that it actually is 2.18. Grahn 16:00, 18 May 2005 (UTC)
Weird, I swear it did say 2.18 a day ago... Perhaps they ammended the PR? Oh well, it now accurately reflects Sony's current magic number. -- uberpenguin 18:13, 2005 May 18 (UTC)
I just can't understand why people in Wikipedia don't read talk page and don't check the sources making irresponsible changes; Watch the press conference (placed and explained for second time here) [1] and see slide show presented by Ken Kutaragi, with a direct comparisson with the Xbox 360. I will back with information and wait that the author check the sources, an official information of Sony. --Mateusc 00:04, 19 May 2005 (UTC)
Hmm... Well, by the very link you provided, [2] the number is said to be "2 teraflops of overall performance." The press release linked in the article also states " System Floating Point Performance: 2 TFLOPS." [3] We all realize these are preliminary and probably marketing-touched numbers, but before you point fingers at people for "irresponsible changes" and failing to "check the sources," perhaps I might suggest you check your own sources first. -- uberpenguin 00:10, 2005 May 19 (UTC)
Man.. you saw the slideshow? it has a decimal number. 2.18 check the source correctly please, I ask for this. --Mateusc 00:15, 19 May 2005 (UTC)
So you are saying that the slide show is magically less prone to typographical error than Sony's press release? Look, this is EXACTLY why I suggested earlier that we remove this number entirely because it is a MARKETING FABRICATION and not any kind of useful performance metric. However, you and others INSISTED on keeping it, so now you are going to have to deal with the ambiguity that this sort of thing causes. The fact is that right now we have two equally reputable sources in contention; the Sony press release and their slide show, so perhaps the phrasing should be "over 2 TFLOPS" to err on the side of generality rather than inaccuracy. Would that suit you? -- uberpenguin 00:19, 2005 May 19 (UTC)
Err.. I know about this, but this is a number divulged by Sony and we can't hide. If it will be fulfilled, and it has controversies, this is substance for another article or section.. (why you don't write one? ;)) --Mateusc 01:12, 19 May 2005 (UTC)
Because it's all speculation at this point; I will not write a section based on pure speculation. It's expected that a demo at a game conference will contain a healthy dose of marketing, and here we are. A lot of folks are just a little too interested in the internals of these game consoles for their own good, so we get situations in which a certain marketing number (*ahem*) is agonized over. Right now the article suffers from list overload, but this is tolerable until we know enough to flesh out a bit more. If you do not object, I will change the phrasing to say "Over 2 TFLOPS" since this is correct by both sources and deals with the ambiguity. -- uberpenguin 01:18, 2005 May 19 (UTC)
It's a oficial and final tech information exposed by Sony and compared in the media. Not a rumor, not a speculation, just a tech and divulged oficial info. --Mateusc 01:27, 19 May 2005 (UTC)
*sigh* Yes, but it is totally unclear how Sony arrived at that number, the number isn't really a useful metric in the first place, and even Sony itself has released two different numbers. I am changing the phrasing because it is not clear what is correct, and being a marketing number it probably matters very little. -- uberpenguin 01:55, 2005 May 19 (UTC)
It's important to Keep, because Sony it made question to show exactly numbers in your conference comparing with Xbox 360 (1.15 Tflops) You understand me? For future, I'm certain this hype number but Sony says that... --Mateusc 03:35, 19 May 2005 (UTC)
It's a marketing number that doesn't translate to real-world performance in any useful way. I'm agreeing to keep it because there is little hard factual information on the console right now and it makes several other people happy for this page to echo Sony press releases. -- uberpenguin 03:43, 2005 May 19 (UTC)
If it's a market number or not, Sony divulged! We can't hide a number that divulged, Jesus, You can't understand this? I'm not question what this number is, if is real and practical or isn't, that number was divulged Here in Wikipedia we not can't hide this. Wrong or not, this a information divulged by Sony. --Mateusc 17:19, 19 May 2005 (UTC)
Sony ALSO provided the number 2 TFLOPS in their press release, meaning that even they haven't got their numbers straight yet. Have you been reading anything I've been saying? Would you like me to request mediation on this since you either seem to misunderstand or ignore everything I've been typing? -- uberpenguin 18:58, 2005 May 19 (UTC)
Look, to show the loopyness of your reasoning, shall I also place another number in that section that reads 2 TFLOPS? Sony divulged the number 2 TFLOPS in their press release, I suppose that means the information should be included on Wikipedia. Hell, why not just go ahead and copy their entire body of press releases and marketing slides here? -- uberpenguin 19:01, 2005 May 19 (UTC)
The number it's on Sony E3 Press Conference slideshow, look in this talk page, I'm not post for third time. I will back with 2.18 number - because it's presented by Sony on E3 Press Conference. --Mateusc 20:42, 19 May 2005 (UTC)
And the number 2 is in their press release, as I have said at least three times now. Which is more reputable? Do you suggest one to be less prone to error than the other? Please share with me your reasoning here. So far all you have said is "one source from Sony says 2.18" but have completely ignored the fact that another source from Sony gives a different figure. I feel as if perhaps something is being lost in translation here... -- uberpenguin 02:18, 2005 May 20 (UTC)
You're can't understand-me. I'm not disagree with you, I found only expose an crucial information that Sony shown in your press conference. I know (We Know) Sony hypes and strategies. We need information for people what Sony says, to promote your product. I ask for you expand the section, writing a thing like this: "Sony in your E3-Press Conference shown that PS3 have 2.18 Teraflops of overal system performance, disclosing unreal number not-practical performance". You understand-me now? People that access Wikipedia need to know this, because same thing has made with Emotion Engine in 2000. -- --Mateusc 02:46, 20 May 2005 (UTC)
Fine, will do... Hopefully changing the phrasing won't cause a thousand anonymous users to jump all over it... -- uberpenguin 02:51, 2005 May 20 (UTC)
Wonderfull, it's perfect now. --Mateusc 20:21, 20 May 2005 (UTC)
There seems to be little I need to add to this discussion, but I would like to point out that, before my "irresponsible" edit, the article said: "According to a press release by Sony at the May 16, 2005 E3 Conference, the final specifications of the PlayStation 3 are as follows." followed by "Estimated floating-point capability 2.18 TFLOPS". Since this was not what the press release said, the article was incorrect, regardless of what some slide show said. Mateusc, perhaps you shouldn't be so eager to throw accusations around, and check the sources yourself before you post. (Not that I can get why this has to be such a big deal anyway.) -- Grahn 08:11, 27 May 2005 (UTC)
Maybe the press release was just rounding it off to 2? --Wulf 01:17, Jun 24, 2005 (UTC)

GPU

300M transistors in RSX, source: Sony E3 2005 Press Conference

Actually they only confirmed it would be at LEAST 300M. The exact count is currently unknown.

Price confirmed?

Amazon is offering preorders of the ps3 for 300 US dollars: [4]

Umm, no. I wish and hope it will be $300 USD, but the truth is that no one really knows (except maybe the head guys at Sony). Amazon.com now says "This item is currently not available. Customers also shopped for these similar items." Anyway, I would never take pre-orders prices by dealers as an actual price. See IGN's PS3 FAQ [5]-
How much will the PlayStation 3 cost?
Unknown at this point, though Sony guru Ken Kutaragi has stated that "It will be expensive" -- so it will almost certainly be more than the sweet-spot launch price of $299. In a recent interview with Japanese economic website Tokyo Keizai, Kutaragi said the company wants "for consumers to think to themselves 'I will work more hours to buy one'. We want people to feel that they want it, irrespective of anything else." -IGN
Except for the Ken Kutaragi quote, they are just speculating, so don't think that it must be more then $299 either. However, we do know the price of the Xbox 360 - see IGN's interview with Peter Moore on the official pricing of the Xbox 360. [6] There will be US$299 "core" system and a US$399 "premium" system. More at Xbox 360#Prices. -Hyad 06:35, August 19, 2005 (UTC)
In the wikepedia article for PlayStation 3 there has been an edit by 150.216.124.44 from 40.000 Yen to 50.000 Yen, without providing a source, this is not how it should be done. I won't change it back without better information but someone who can, please maintain the "Cost and release date" part.ShotokanTuning 10:21, 17 October 2005 (UTC)

PC performs at .08 TFLOPS?

Who added this information and what justification do you have for it? Considering the SIMD units in big desktop microprocessors, this seems like a very pessimistic estimate. Why are we even trying to compare the vector capabilities of a game console to those of a PC? Most PC users have very little need of high vector performance, and thus a SIMD unit or two is sufficient. If you wanted to make a FAIR comparison, compare the PlayStation 3's esimated FLOPS to that of a vector architecture, like the NEC SX-6. I'll probably remove this phrasing unless someone can think of a really good reason to leave it. As a point of comparison it is questionable and definitely misleading. -- uberpenguin 06:11, 2005 May 22 (UTC)

Presumably it came from Sony's press conference, where they showed off that not-very-convincing statistical analysis. Should probably be noted in the article that it's Sony's estimate and may be unrealistic. Sockatume 14:18, 22 May 2005 (UTC)
We don't know the specs and the Sony criteria shown on the slideshow, but it was presented of this form on Press Conference (16 may). --Mateusc 17:45, 22 May 2005 (UTC)
Maybe so, but it isn't the job of Wikipedia to simply copy and paste one company's marketing information. The number for the PC is highly questionable, there is no indication of what kind of PC they used for the number, what sort of code was running on it, whether or not it used SIMD optimized instructions, etc. I think we should remove the PC comparison for clarity's sake, because it only adds to confusion and suggests that floating point performance is the end-all measure by which we compare computers. -- uberpenguin 12:20, 2005 May 23 (UTC)
An 'average' PC (eg P4@3GHz) performs at no more than 10GFLOPS according to Linpack and other benchmarks.
Do I even need to point out the amount of subjectivity in that kind of benchmark? Even within the LINPACK suite there is a lot of variation... What if one performed the benchmarks without the Atlas library? Could that be considered 'fair?' What is a 'fair benchmark?' I certainly don't know anymore... The point is that FLOPS are at best a limited indicator of computer performance, and in an article like this, that doesn't go into detail regarding the ins and outs of computer function and capability, it's misleading to slap a "comparison to PC" number on the page as if it is some magical universal ruler by which computers can be gauged. -- uberpenguin 04:35, 2005 Jun 6 (UTC)
At the moment, it's marked out pretty well as a dubious marketing figure. Perhaps the best option would be to simply ditch the comparison and simply include the claimed PS3 figure; we're going to have to collapse the specifications down to something on a par with the PlayStation and other articles' spec lists at some point anyway. Sockatume 05:47, 6 Jun 2005 (UTC)
Yes, of course, but right now there is so little information available and so little to write about the PlayStation 3, that this will have to suffice. In all honesty I think a lot of this should be cut out since the article is still dangerously close to a list, but this format seems to keep most people reasonably happy, so we'll stick with it until better text can be written as new information becomes available. Until then we're doomed to revert all those anon edits that would have this article become nothing but a collection of all the internet rumors and endless speculation on the subject :) -- uberpenguin 14:43, 2005 Jun 6 (UTC)
The point is that an 'average' PC does not perform at more than 10GFLOPS, while Sony CLAIMS its PS3 performing at about 100 times this number. We do not expect this to be true (in any sense) but it tells something by itself. Maybe it should be noted how highly subjective this kind of comparing is but such a claim on performance 'kick' should not be passed by. 193.92.150.203 03:31, 7 Jun 2005 (UTC)
On other somewhat dubious information, does anyone think the listing for vertices performance hiding something? Aren't vertices the positions that make up a polygon triangle face and it takes three vertices to make a polygon? Extrapolated, this means that the polygon fill rate is something like 370 million, more than 100 million less than the XBox 360? Granted, this doesn't mean that the vertices can be linked together and produce more than that, but it does seem a bit fishy. I would like to see more information on how this benchmark was conducted on both systems. --YoungFreud 02:59, 21 August 2005 (UTC)

The USB (2.0) ports

Will I be able to connect my USB mass storage device and save my Game Saves to it?

Do you think we are authorities on a prototype console that was just announced and will go through another year+ of revision before it is released? -- uberpenguin 13:31, 2005 Jun 3 (UTC)

lol yes uber!!! /sarcasm. anyways, sony has said they will feature an attachable harddrive very similar in fashion to the current one, in that respect you would have to have the recognized sony drive to save onto unless you modified the system which is very possible as we saw on the xbox mod chips (you could replace the hard drives with a larger one, for not so legal reasons sometimes) but i wouldnt count on having 3rd party hard drive support out of the box. However since sony hasn't commented on this i cant tell you for fact.

IF they do they had better accept the mac os format discs, windows dosent and it pisses me off.

1. It's much more likely that Sony uses a proprietary format; if they can get away with it then they will be poised to make more profit. 2. Who cares? Unfounded speculation is useless. 3. There are several NT IFS drivers for HFS and HFS+ volumes, so I don't know what you're complaining about anyway... Microsoft chooses not to support a relatively uncommon filesystem (compared to NTFS, VFAT, EXT2 derivatives, etc) from a third-party vendor that would likely charge licensing fees in their OS, but they certainly provide the capabilities for HFS support via IFS. How can you hold that against them? -- uberpenguin 22:56, 13 September 2005 (UTC)

Memory prefixes

I just reverted the memory specification prefixes to the metric measurements (KB, MB). I believe that the average reader will not have knowledge of IEC binary prefixes and that it will simply cause confusion. This isn't a scientific article, so I don't think it's necessary to be quite so pedantic. Taking a look at articles such as PlayStation, Xbox and Pentium, I see that metric prefixes are used, so I think we should stick with the more common notation until IEC binary prefixes are more widely known. Carbonite | Talk 13:01, 16 Jun 2005 (UTC)

The ambiguity of the metric measurements is the reason that the IEC prefixes exist. 99% of readers will see MiB and just read MB. I have never seen a problem with readers becoming confused by the binary prefixes. If it suits you, I'll make the first instance of the measurements wikilinks to the pages that explain their purpose. I see no reason in this circumstance to avoid using the correct measure just because the extra letter in the abbreviation might look strange to some readers. You might notice that more and more Wikipedia articles are moving towards using IEC binary prefixes, so citing a few articles that still use the ambiguous prefixes shouldn't be justification for changing the text of this one. If a reader is terribly detratced by the use of the binary prefixes then they can click do what every Wikipedia reader should do and click the relevant links to gain understanding of the terminology used.
I'm pretty adamant on this point; it just makes good sense to use the correct measures. If you honestly see justification for this, why not go and change the hard drive and other computer-related articles to be "less confusing?" The IEC binary prefixes are designed specifically to avoid confusion. -- uberpenguin 13:26, 2005 Jun 16 (UTC)
Also, I believe this is particularly relevant here because later in the article a TRUE metric measurement is used; GB/s, which literally means 10^9 binary octets per second, not 2^30 binary octets per second. Maybe if the article required only one interpretation of the metric prefixes it wouldn't be such a big deal, but using uniform metric prefixes in a situation that requires interpreting them both with their SI and their (incorrect) colloquial computer usage is confusing. -- uberpenguin 13:34, 2005 Jun 16 (UTC)
I understand your intentions and agree with many of your points. However, I don't believe that this is an article that should place technical accuracy above comprehension. I do think that many people will be puzzled by seeing MiBs and KiBs, even with the Wikilinks. The average reader of this article is a gamer, not a computer scientist. This article is different from the hard drive article, because that articles discusses the topic of marketing capacity vs. true capacity. I wouldn't be opposed to including both measurements (like the hard drive article), but only including IEC prefixes doesn't seem to be the best course of action. Carbonite | Talk 15:29, 16 Jun 2005 (UTC)
I too see your point, and it is perfectly valid. However we seem to have an irreconcilable difference of opinion here, so perhaps we should sit and wait a bit for some other people to comment, or ask some others to provide their POV via Wikipedia:Third opinion. That sound okay to you? -- uberpenguin 17:30, 2005 Jun 16 (UTC)
I don't think that's necessary. I'm not going to force the issue, nor will I revert to metric. Perhaps it may be worth persuing the issue of binary prefixes in the Manual of Style or some other broader forum. Carbonite | Talk 18:56, 16 Jun 2005 (UTC)
Perhaps, I'll have to take a look at that sometime. Thanks for the understanding! -- uberpenguin 19:10, 2005 Jun 16 (UTC)

The new units are not standard SI units, until they are, we should stick with the standard KB, MB, etc. (See http://physics.nist.gov/cuu/Units/binary.html) It will confuse people. And personally, I wouldn't want to use units I can't pronounce and that would make me sound like an idiot when speaking to others.

"Quite often is the short answer. But what are kibibytes, and indeed mebibytes, gibibytes, tebibytes, pebibytes and exbibytes? The answers are all in IEC 60027-2, developed by TC 25 (Quantities and units, and their letter symbols), published in November 2000 and now gradually being adopted in the IT world. Essential details of the new units, their derivations, symbols and approximate relation to commonly, if sometimes incorrectly, used metric equivalents in the Système international d'unités (SI), are shown in the accompanying table. How do these new standardized units differ from those that have become so familiar during the last two or three decades’ explosion in personal computing?" (See http://www.iec.ch/zone/si/si_bytes.htm)
The binary units are standard units. They will never be metric SI standard units because they are not metric (Base 10), they are base 2. "Mehbeebyte" If you can't pronouce some words used in the encylopedia it does not mean that those words should be removed.--Thax 4 July 2005 05:32 (UTC)
Okay 24.36.1.94 was cleaver and went to the Xbox 360 page and edited the units to KiB and MiB and the change was subsiquently reversed. Certainly this brings forward a greater need of consistancy throughout the encyclopedia, it isn't feasible to have different units of measure on related pages. --Thax 5 July 2005 18:26 (UTC)
I brought this forward at http://en.wikipedia.org/wiki/Wikipedia_talk:Technical_terms_and_definitions#Unit_Disagreement.2C_MiB_vs_MB --Thax 5 July 2005 18:25 (UTC)
Okay, I moved this to the village pump as I was not getting any activity in it's current location. http://en.wikipedia.org/wiki/Wikipedia:Village_pump_%28policy%29#Unit_Disagreement.2C_MiB_vs._MB --Thax 8 July 2005 02:54 (UTC)
I wasn't being clever, I edited it for consistency after the PS3 article was reverted to the new units.

A vote has been started on whether Wikipedia should use these prefixes all the time, only in highly technical contexts, or never. - Omegatron 14:48, July 12, 2005 (UTC)

I cannot seem to find a vote, or anything at the village pump...
The vote ended like a month ago and has been archived here: http://en.wikipedia.org/wiki/Wikipedia_talk:Manual_of_Style_%28dates_and_numbers%29/archive22
The policy was added to the following section: http://en.wikipedia.org/wiki/Wikipedia:Manual_of_Style_%28dates_and_numbers%29#Non-base_ten_numbers
--Thax 02:51, 16 August 2005 (UTC)

Gallery

Would it be any easier to put each picture on the page instead of having a gallery? Let me know what you think. --Numberonedad 18:02, 25 May 2005 (UTC)

Ps3 devkit

Can someone post this picture somewhere? http://ps3.ign.com/articles/628/628048p1.html

Sure, why don't I just take the latest Interweb rumour picture and stick it right in this wikipedia article now? -- uberpenguin June 28, 2005 13:04 (UTC)

The teraflops have to be wrong...

By no means is this a biased post. However, I cannot come to believe Sony that PS3 performs at 2.18 teraflops, when the Xbox 360 has 3 processors, each with dual core/SMT/hyperthreading technology, which means that the processors can virtually duplicate themselves, enabling them to perform 2 separate tasks at once (essentially, 6 individual processors each running at 3.2GHz). That just doesn't make sense. There is no way that a SINGLE processor that runs at around 3.0GHz - 4.6GHz can out-do 6 processors at 3.2 each. Now we can look at the Graphics Processing Unit. The graphic codes are streamlined through one single path in the Xbox 360, enabling the card to process the graphics quicker and more efficiently than standard cards. Also, the GPU has a 10MB memory cache built in, so the card doesn't have to fully rely on the system's main RAM. Sure, the PlayStation 3's GPU is 50MHz faster, but the processes go through different paths, making it less efficient. Not to mention that PS3's GPU has to fully rely on the system's RAM. The RAM in Xbox 360 is the same MB count as in PS3 (512). However, Microsoft gave a 1MB L2 Cache for the processors to use, and the GPU has it's own 10MB cache. The PS3's Cell has to rely mostly on the system RAM, and the GPU has to heavily rely on the system RAM also. Another thing that boggles me is that the Xbox 360 has a built in 20GB hard drive, while the PS3 has none (even though one might come in the future). I don't know, it just doesn't seem logical, with the 360's distinct technical advantage, that the PS3 would have teraflop performance of almost DOUBLE that of 360.

First you need to understand the difference between cores, processors and multiple threading. The x360 has 1 CPU, or processor which has 3 cores and each core can execute 2 hardware threads. This does not mean that the core is 100% efficient at processing 2 threads, each thread shares the core in an efficient manner. This means when one thread is waiting around the other thread can execute instructions on the core. So if the x360 CPU were executing a stream of floating point operations, each core would be 100% utilitized and the dual threads would not add a great deal of extra performance. Where the x360 is able to gain some ground are the VMX 128 vector units which are able to perform fast floating point operations.
Secondly you omit the fact that the PS3 Cell has 1 Core, but 7 SPE's which are doing a majority of the floating point calculations. Each SPE is especially designed to execute floating point calculations quickly and can easily outperform the any component on the x360 in terms of floating point calculations.
"The graphic codes are streamlined through one single path in the Xbox 360, enabling the card to process the graphics quicker and more efficiently than standard cards" - Even though I am not sure what you are trying to say here, I can tell you it is wrong. If you look a the bandwidth diagram I created for the x360 on the Xbox page you will see that the GPU can directly access memory, as well as the CPU and is also the connection to the southbridge. In addition to this the eDRAM daughter die is able to execute instructions using it's internal logic. There is not one single path.
"Also, the GPU has a 10MB memory cache built in, so the card doesn't have to fully rely on the system's main RAM." - The 10MB of eDRAM is on a daughter die, so it isn't really built in yet.
In addition I don't see how the having a hard drive helps the x360 have better floating point operations. Your post is inaccurate and misleading --Thax 4 July 2005 16:45 (UTC)
This is a frivolous discussion about computers who have yet to be benchmarked by third parties. All the numbers we get from Microsoft and Sony are theoretical peaks, which rarely accurately reflect the real-world performance. Furthermore, you're getting far too caught up on double precision floating point capability. Believe it or not, computers (game consoles included) do have to perform more tasks than just crunching IEEE floating point numbers. This discussion doesn't belong here; it doesn't involve the article or its contents since the article accurately reflects what is currently known about the console. -- uberpenguin July 4, 2005 18:09 (UTC)
Retrieved from "http://en.wikipedia.org../../../p/l/a/Talk%7EPlayStation_3_Archive_2_83e7.html"