Rumor: PS3 hardware "slow and broken"?
by Christopher Grant 
{ Jun 5th 2006 at 11:50AM }
Another week, another PS3 rumor fire quickly decimating rational thought on the internet. We should start billing Sony for putting these things out. This time, British computing tabloid, The Inquirer, is running a piece titled, "PS3 hardware slow and broken." It's full of the usual Inquirer stuff: bad spelling, horrible grammar, and (most importantly) specious arguments. So, before we even get to what they're claiming, understand that their story could be considered entertainment, much the way the Enquirer is. Here goes:An Inquirer writer had the good fortune to sit next to Anonymous Guy on a recent flight to Japan who dished about the PS3's new Cell and RSX chips. "Oh my God, did you hear? Cell is like, totally broken and junk. It's MBs are all, like, messed up where the GBs should be ... and stuff." Alright, that's not a direct quote, but a decent approximation. It isn't worth going through and breaking down the arguments since a little common sense would go much further.
Alright, internet, I know we're all still upset about "The Price." I know sometimes Sony says funny stuff that we all laugh at. (Too cheap? You kill me Kutaragi!) There are some kinds of rumors we're willing to believe. Yields are low? Alright, that happens. One of the largest and most well respected chip developers in the world, whose new, custom-built, ground breaking chip architecture has a huge design flaw that we extrapolated from a slide and a single-serving airplane buddy? See, it just doesn't seem as plausible.
In regards to a system as unique -- and uniquely complex -- as the PS3, this level of "journalism" seems woefully inadequate to tackle what may or may not be actual design problems with the processor. Places to look, if you're so inclined: Sony's Cell page. Too biased? How about one of Ars Technica's famously exhaustive looks at the thing ... in two parts. The Ars guys are part processor, it appears, and they didn't manage to identify some singular shortcoming that will doom the PlayStation 3 as an also-ran.
So go ahead. Read it, enjoy it, but please don't take it as Truth. The console wars are already ugly enough without the rumors.
See also:
Clearing up some PS3 confusion
Rumor: no pre-owned games for PS3?
[Thanks, I guess, to everyone that sent this in]
(Update: The misunderstanding evident in the linked story relates to the distinction between local video memory and local system memory. The slow read speed under discussion is indicative of the feature's lack of utility. This is even reflected on the slide's statement: "No, this isn't a typo ..." A contact at Sony confirmed this telling me, "Again I cannot imagine a situation where you have any SPU reading from the RSX local memory." Nothing to see here, folks.)
Reader Comments (Page 1 of 2)
JesusXP @ Jun 5th 2006 11:57AM
Remember, the cell wasn't designed for PS3 Specifically, or for Games specifically...
Gavin @ Jun 5th 2006 11:57AM
Read the comments on Slashdot concerning the very same article. It is debunked. The supposed "slow and broken" aspects of the Cell and RSX chips are established ways of handling 3D that are pretty regular on PCs already. This isn't anything even remotely wrong in any way, shape, or form.
False alarm.
Benny @ Jun 5th 2006 11:58AM
Don't take any notice, the Inquirer is widely regarded for making up lots of garbage stories. Not to mention the fact they have trouble with basic spelling...
nootau @ Jun 5th 2006 12:07PM
Sigh. I think the games at E3 show that the PS3 is doing just fine. I know that it wasnt final silicon, but come on! Even though i do read the inquirer, the game developers have a lot riding on the cell, so i doubt the rumor has any validity. Sony would rather implode than allow this to happen.
But thanks for sharing.
boots @ Jun 5th 2006 12:09PM
Sigh, I knew the 360 was more powerful and easier to develop for. Screw the shitty ps3.
zero2dash @ Jun 5th 2006 12:10PM
Many columns before have practically said the same thing; Cell is a not a performance-driven processor, it is a numbers-driven processor. Slow is not far-fetched at all.
360 and Wii are geared towards performance, not raw number-crunching ability, therefore (in most cases) you'll see better performance (on the cpu side) from those two consoles compared to the sluggish Ps3.
Anandtech covered all three systems and their cpus a long time ago.
Cell having 1 PPE (primary processing engine) and 7 SPE's (synergistic processing engines) versus 360 having 3 dual core PPEs versus Wii having either 1 single core or dual core PPE (don't recall which it is at this time). Eliminating the technogeek babble - Ps3 can handle more with its 7 SPEs but the 360 having 3 dual core PPEs does a better job at managing all that data, therefore it performs faster (read: better).
The above article might throw in some fanboy euphemisms but the fact remains - the Ps3 is the slowest and least performance friendly processor in the next gen consoles.
trogdor @ Jun 5th 2006 12:12PM
Sigh, I knew that boots knew that the 360 was more powerful and easier to develop for.
Meatstick @ Jun 5th 2006 12:13PM
Well, somethingawful have a topic on this, and they pretty much cover all aspects of this: http://forums.somethingawful.com/showthread.php?s=53d04eebd612d47fb9f30f2ef35f16b5&threadid=1892991&perpage=40&pagenumber=1
Also, The Inquirer is reliable, I don't know where you got that from, did you pull it out of your ass? The Inquirer is not The National Enquirer, and is actually reliable. Its a UK tech news publication, not the tabloid Inquirer, you got mixed up. So I don't know how this is debunked.
zero2dash @ Jun 5th 2006 12:13PM
Correction: it wasn't Anandtech, but ArsTechnica. (didn't see that mention in the original post...sorry)
Joystiq crew - future suggestion (if possible): the ability to edit posts =)
puh_fifer @ Jun 5th 2006 12:17PM
I caught the hidden Fight Club reference, do I win a shirt?
"How's that working out for you?"
"What?"
"Being Clever"
Sense @ Jun 5th 2006 12:22PM
All this techno babble really doesn't matter. If the PS3 has a user base, developers will support it. From a technical standpoint, the PS2 was the hardest to develop for, but it still dominated.
The same thing could easily happen for the PS3, although I personally believe the field will be slightly more evenly split this time around.
ozymandias @ Jun 5th 2006 12:23PM
christopher, you just opened a can of worms. my guess is 108 angry fanboys.
pedantic @ Jun 5th 2006 12:25PM
The values are correct in the slides, the problem is what is meant by "Local Memory". "Local Memory" refers to the RSX's memory. The Cell can write to it fairly quickly but won't read from it quickly. I see no problems with that, PCs can't exactly read from the memory on your graphics card very quickly either.
zsavior @ Jun 5th 2006 12:25PM
As much as I agree with the article, by Christopher Grant and his accurate assumption of the Inquirer being crap there is a problem I do find in it. The problem is this, though the Inquirer is prove crap on a stick, most video game journalism isn't much better. Truth is we can't depend on EGM, Game PRO, or any other gaming maganize to give us the real low down on Sony, or MS, or even Nintendo in their hey day.
It is cool to attack the Inquirer crap like that needs to be gone, but seriously on a whole gaming journalist need to step up and actually be considered an asset to the gaming industry rather than a mouth piece for the popular consoles of the moment. Nothing against joystiq, but a site that post blogs, and is free shouldn't be giving better, and more reliably neutral news than that of the current game magazines and actual new papers. Not that I agree with every comment on the site but the fact that, there is not one happy group on this site means something.
The Worst example of video game journalism in past was Dan Hsu's taking apart Peter Moore like a Hungry dog on a chicken bone. It wasn't bad cause Hsu's approach and thoughts were invalid, hell no they were right on the money. It was sad cause the industry stood at a stand still shocked and amamzed that a gaming jounralist actually asked hard hitting questions. I mean really what does that say for the rest of the industry and the expectations the readers have from the print, and televised media they see from the journalist.
Tom B. @ Jun 5th 2006 12:30PM
16MB/s vs. 4GB/s.... Isn't that more like 2 orders of magnitude?
Typical sensationalist exaggeration.
32_Footsteps @ Jun 5th 2006 12:34PM
"Sigh. I think the games at E3 show that the PS3 is doing just fine."
Well, if by "doing fine" you strictly mean "not a danger to destroy property or cause fires," then I'm willing to agree with you.
While I personally doubt the veracity of this story (as Joystiq obviously does too), keep in mind that the E3 demos shown were at best aroun 30% complete (judging by the little "complete" placards Sony puts in front of games at the show). So it may be premature to say that a full game, requiring much more processing power than a game only a quarter finished, won't cause interesting pyrotechnics with the architecture as it currently stands.
We may yet see interesting issues with the system, but I don't think this story is anything more than a blatant grab for readership by spreading gossip.
cap-n-crunch @ Jun 5th 2006 12:35PM
So how did Joystiq debunk the rumor? By saying the Inquirer is uncredible, and linked to another site?
That was easy.
mmagic @ Jun 5th 2006 12:41PM
Meatstick if you beleive the inquirer is legit info then all power to you the rest of us know better
LunarDuality @ Jun 5th 2006 12:49PM
@ ozymandias
Why stop with 108 angry Sony fanboys? Why not 1080 "pee"d off Sony fanboys instead?
C. Grant @ Jun 5th 2006 12:51PM
Problem being? The entire internet is latching onto this (we got it tipped about a dozen times in a couple hours), we're encouraging people to not take it at face value. If you understand the actual technical underpinnings of the console (I sure as shit don't) then look it up. All I know is that the Inquirer's story reeked of BS and generalizations.
zsavior: Thanks for the compliment. That's a key part of the reason I think blogs like ours have become so popular.
And Meatstick: I know the Inquirer and Enquirer are two different things, but if you don't think it's a tabloid and is, instead, a "reliable tech news publication," I got a bridge you might be interested in buying.
LaughingTarget @ Jun 5th 2006 12:59PM
Here is something I'd like some clarification on from some more savvy people.
From what I've researched, local memory is memory that is used exclusively by a single processor, so in the case of the PS3, only by the SPE and the main processor. The specs show that half of the 512 megs is being used as local memory.
So, I am under the impression that the local memory is completely useless for anything else other than what that specific SPE is doing, so the SPE cannot write to the RAM and the RSX graphics chip cannot read it while it is in there. That tells me that the SPE will have to read it back then put it into system memory before it could be of any use. So, I am under the impression that this will cause significant problems because the SPE has to pull the information at 16 megs/second.
I'm guessing this won't cause issues in early PS3 development, but will bottleneck the Cell's full potential since the SPE's are eating up half the system RAM and cannot read at the same speed as the rest of the system.
Mr. Samuel @ Jun 5th 2006 1:02PM
Props for the Fight Club reference. That is all.
Jdoki @ Jun 5th 2006 1:02PM
The INQ is about as credible as 99% of other sites when it comes to speculation, heresay and so on.
As a long term sufferer of the IT industry I find it an enjoyable read, if taken with a big pinch of salt at times.
Bit of an unfair bashing by Christopher Grant in my opinion. I wonder how many people who have posted the 'INQ sucks' equivalents have actually visited the site?
Kirk @ Jun 5th 2006 1:05PM
I still think the price is broken.
cringer8 @ Jun 5th 2006 1:07PM
I like the part about "triangles." It makes me think of confetti which makes me think of parties. Parties are fun. That's about all I get out of this article.
We'll see in November :)
SuicideNinja @ Jun 5th 2006 1:12PM
"Many columns before have practically said the same thing; Cell is a not a performance-driven processor, it is a numbers-driven processor. Slow is not far-fetched at all."
The Cell processor seems to me that it would be better for servers than for a multifunctional media device. But isn't that its intention? For a multi-processor server, the cell would seem to perform at it's best.
"From a technical standpoint, the PS2 was the hardest to develop for, but it still dominated."
Will devs want to put up with Sony's overly-proprietary and overly-difficult hardware forever? I wouldn't think so personally. Programming is time-consuming enough without adding in a processor that works against you. If the provided tools are sub par, then that amplifies the headaches.
elmimmo @ Jun 5th 2006 1:16PM
The thing is not wether The Inquirer is crap or not. Did it state crap or not? Joystiq's did not answer that, and hence it statement was rather lame.
True Smoka @ Jun 5th 2006 1:16PM
damn the ps3 really does suck...............bummer
Andir @ Jun 5th 2006 1:19PM
@LaughingTarget:
Your confusing the SPE memory with video memory as well. This presentation deal squarely with the RSX processor and the "Local" memory is akin to the video memory in your PC. The Cell doesn't need to access this memory. In the slide Local Memory is the video memory and Main is the system memory.
sully @ Jun 5th 2006 1:20PM
Bravo for the little aside here about The Inquirer and their incredibly horrific article writing, both from a moral and technical perspective.
I'm an Xbox 360 fan all the way, but 'printing' this even as a rumour is unworthy of even an amateur blogging site.
LuckyT @ Jun 5th 2006 1:22PM
There seems to be a lot of confusion over this, surprise surprise.
It's really very simple. Look at the chart. "Local memory" is GDDR3, because the slide is coming from the point of view of RSX. The interconnect between Cell and RSX has been optimised totally for RSX access - and rightly so, because it needs a lot of bandwidth. This means it has full, clear access to XDR, but Cell has very limited access to GDDR3. This is a bit a lot of people seem to be missing - Cell still has full, clear access to its own XDR memory, it simply has very limited access to GDDR3.
The upshot of this is actually good news, for people like me who were sceptical about the bandwidth going into and out of RSX. The memory system design is clearly very GPU-centric, and this is how it should be! Cell is fine with just its access to XDR, but RSX really needs more, and thankfully it has got it. The only implication of this is that if you want Cell to access data from GDDR3, then really you should be getting RSX to write it out to Cell or XDR, rather than Cell reading it from GDDR3 itself, since RSX has much higher bandwidth going in that direction.
Basically it boils down to which chip was prioritised in the FlexIO (interconnect) design, and they prioritised RSX as they should have!
C. Grant @ Jun 5th 2006 1:22PM
Jdoki: I read it at times, I enjoy it at times, but I don't ever mistake it for a "reliable tech news publication." They pay the bills with tabloid-style coverage of the tech industry. Many of their claims are specious; they *almost* sound reasonable, and are often attractive to consider from a tech standpoint. Just calling a spade a spade. Sorry you fund it an unfair categorization.
Cabbage @ Jun 5th 2006 1:31PM
The issue being brought up in the article is that the CPU does not read from this 256MB of "local memory" very quickly. At all. What does this mean? Every time you need to crunch some data, that data is in the memory. This means that it doesn't matter how fast the CPU can crunch data because it takes forever just to read that data in. That's like doing your math homework when you have to go next door each time you want to see the next problem.
KEEP READING: The problem with this article is that the local memory being referenced is the half that the GPU is using to store textures and all its other data. The GPU can access its 256MB quickly, and the CPU can access its 256MB quickly, but the CPU can't read the GPU's memory very well. WHO CARES! It can send information to the GPU just fine.
The Inquirer is just trying to get attention, and guess what... it worked! They are doing what tabloids frequently do and taking an image, twisting what it really means, and shouting that the world will end!
LuckyT @ Jun 5th 2006 1:34PM
31 - Moreover, the GPU has lots of nice bandwidth to the CPU's own memory. Very GPU-centric - me likey.
Obviously there's an added latency for the GPU accessing the CPU's memory, but if you're not bound by texturing, texturing from XDR should be no different than texturing from GDDR3. And that's a big win.
zsavior @ Jun 5th 2006 1:48PM
JDoki to be fair at Slashdot, I actually checked out what one of the writers from the Philly Inquirer had to say, and as an editorial it was actually very good. The colum is here
http://www.philly.com/mld/philly/news/columnists/14733980.htm
The writer Johnathan Last wrote very valid points, so I will conceed to some of your argument. The problem I have is, does this sort of intergrity go to every article at the Inquirer. I think if we were to examine various stories that answer would be seen quite clearly.
oBLIQUE @ Jun 5th 2006 1:58PM
ah the 360 doesnt have 3 dual core processors it has 3 cores.
learn something
SHagi @ Jun 5th 2006 2:06PM
Your bashing of the Inquirer is totally off base. If I want to find out the absolute latest news on new hardware I don't go to joystiq. The INquirer usually has major hardware announcements and other hardware related articles of interest up DAYS before joystiq ever hears about it.
I always thought you guys crusied INQ yourself to find stories to write about (BEacuse of how slow joystiq usually is).
As far as calling them unreliable I've never actually read anything that wasn't specified as a rumor or speculation that was unreliable.
Maybe someday I'll think of Joystiq as a news site like INQ but for now I come here to kill time and watch the trolls.
ethan @ Jun 5th 2006 2:10PM
man you people are retarded, I don't understand why joystiq posts something false and then says it false throughout the whole article. If it has no relevance to the truth, don't F*****' post it. You just confuse all the retards that look at the article because they take it seriously.
27. damn the ps3 really does suck...............bummer
point right there proven.
blahman @ Jun 5th 2006 2:10PM
Yeah, I remember when I saw the gameplay footage of Heavenly Sword, the one thing running through my mind was, "this is hideous, WHERE ARE ALL THE TRIANGLES AT?"
martin @ Jun 5th 2006 2:19PM
I read in the Inquirer that the Loch Ness monster is already in line for a PS3 in Maine, and Bigfoot knows a person who knows a person who knows his brother who hooked him up with a Wii dev kit and is already playing LOZ:TP!
Andir @ Jun 5th 2006 2:19PM
@ethan: One thing I've noticed lately about Joystiq is that only one thing matters. Page hits. It doesn't matter if it's false, makes no sense, or is newsworthy. They post these to get "fanboi" responses, page hits and possible advertising income.
C. Grant @ Jun 5th 2006 2:21PM
We post it ethan because it's already spread around the entire internet as fact. We got it tipped a dozen times in a couple hours. That means it's big. Think of the post as a public service announcement.
Thanks for the thought SHagi. Don't think we're really trying to match them. Sort of a different focus, don'cha think?
zsavior: The Philadelphia Inquirer and TheInquirer.net are two very different animals. The Philadelphia Inquirer is our paper of record in this here city. See here for more on TheInquirer.net
http://en.wikipedia.org/wiki/The_Inquirer
C. Grant @ Jun 5th 2006 2:38PM
Andir: nice theory. So you've been noticing this lately? Some examples please. Surely you have more than your gut feeling.
If all we wanted was page hits, we'd be running a porno site. We write about video games and we write about the cultural zeitgeist of gaming. If a dubious story hits digg and slashdot, currently two of the largest influencers of online tech hubbub, you don't think it's worhty of discussion (or dismissal) at the largest video game blog? Do we do it for hits? Sure. That's why we're here. We write about video games for people to come read us. If we do a good job, more people read us. If you're going to play the "corporate" bloggers card, go ahead and play it, but don't hurl accusations you aren't willing to back up.
Epic @ Jun 5th 2006 2:44PM
1. The "local memory" read bandwidth for the Cell is low because the Cell has little business reading from the GPU. The Cell tells the GPU what to render, that's it. The GPU isn't doing any processing that the Cell needs to be aware of, so there's no point to having a high local read. This is how PCs have always handled graphics.
2. The "triangle" comparison is referring to vertices. While it's true that the X360 can render millions of vertices more than PS3, this only holds if the X360 is not rendering pixels or textures. In other words, it's a meaningless comparison, since nobody makes a game with just triangles. The reason for all this is because X360 has 48 unified pipes that can handle either pixels, textures or vertices, while the PS3 has dedicated pipes to each of those tasks. So, yes, I guess the X360 could render more triangles if they didn't save any of the pipes for pixels or textures.
Basically, the author of the article either is completely clueless about the hardware, or he is intentionally distorting it. I should mention that a lot of the technical analysis is not my own, hit up Hypn0sis on the GameFAQs PS3 board -- he's the local expert.
Foxx @ Jun 5th 2006 2:44PM
Good to see some huge amounts of misinformation flying around, and PS3-bashers sucking it all in and howling with delight.
Speaking as a developer (not PS3 though) I can say that there is nothing wrong with this. It's similar to the situation with AGP on PCs.
For the CPU, it's fast to write to VRAM, slow to read from it.
Jeph @ Jun 5th 2006 2:51PM
GameDev.Net has commented on this.
http://www.gamedev.net/community/forums/topic.asp?topic_id=396644
I don't think it's a rumor at all. The fact that the enquirer reported it doesn't mean it's false, it doesn't matter who reported it, it's where that picture of the slide came from.
if that's real, there's nothing do it besides the Cell Processor sucks.
as GameDev.net says:
"There was another console which was released with buggy CPUs in the past. Its name? The Atari Jaguar"
bladestar @ Jun 5th 2006 2:54PM
joystiq obviously doesn't like the Inquirer. Which is why this post tries to invalidate them by writing, "It's full of the usual Inquirer stuff: bad spelling, horrible grammar, and (most importantly) specious arguments. So, before we even get to what they're claiming, understand that their story could be considered entertainment, much the way the Enquirer is.”; and changing their quotes, "Oh my God, did you hear? Cell is like, totally broken and junk. It's MBs are all, like, messed up where the GBs should be ... and stuff." and "So go ahead. Read it, enjoy it, but please don't take it as Truth. " damn, I don't want to read it any more. I feel guilty if I do after all the negative things you wrote about it. I mean you said, "The Inquirer sucks; don't waste your time reading their articles. Why do we post this information? uhh I don't know.". Alright, that's not a direct quote, but a decent approximation. It isn't worth going through and breaking down the arguments since a little common sense would go much further (mmm. sounds familiar).
cringer8 @ Jun 5th 2006 2:55PM
Don't bash Joystiq for posting this. If they just sat on their asses waiting for company press releases, what the hell WOULD they post?
This is a BLOG site. How many times do you people need to be reminded of that? Blogging is about conversation, not news. We want to talk about this article and what it actually means. And, btw...SUCCESS! We're all talking. And some of us now have a better idea of what these numbers imply.
Gnat Gnawer @ Jun 5th 2006 2:57PM
Why are people saying that The Inquirer is unreilable?
I visit it frequently and I don't remember it posting terrible biased or unreliable things in the past, they wouldn't of posted this without proof, and it seems so, although its not much of a problem in my eyes.
Jason @ Jun 5th 2006 2:59PM
The PS3 will run just fine... the question shouldn't be with the read right speeds to and from the different types of memory...
The question is... Is the CELL a good processor for gaming? Well like all other things Sony... NO it is not the best processor they could have used... but then again Blu-Ray isn't the best drive they could have used either...
The PS3 is hear for Sony to launch Blu-Ray... the sooner GAMERS understand that the better... There WILL be some GREAT games for it I am sure, but bickering about the flow of memory from GPU to CPU is stupid... Sony said themselves it is NOT just a game machine... so you should expect there to be some things that don't make perfect sense for GAMING.
Developers don't care about that kind of stuff... they care about how many games they can sell... If there are lots of people WHO BUY GAMES for the PS3 then the developers will support it... no matter how much harder it is to program for... we will see though how many GAMERS buy the PS3... when it costs more than a 360 and Wii combined you may see a lot of gamers go that route... they will only be missing out on a few first party exclusives... there will be LOTS of people buying PS3s that may NEVER buy a game for it.