X-Box hacker interview
- Thread starter -tkf-
- Start date
randycat99
Veteran
aaaaa00 said:I'd call that a pretty hefty sized grudge.
I'd call it a personal assessment- an opinion. Seriously, I don't doubt he has his reasons to think that. Where is it written that only positive things can be said of M$? A real grudge would be, "M$ has wronged me in xyz way, and I vow to exploit every single hole in their OS to their detriment in the fullest extent for as long as I live." That is what M$ is smart enough to avoid by employing diplomacy in this situation.
That may be the case, but who is to say that something positive cannot come out of 2 smart minds exchanging some thoughts and ideas? It's a sad state of affairs if programmers find it beneath them to speak with anyone else who hasn't coded a game.
Randy, at the end of the day, the 2 16 meg PC800 RIMMS are blood cousins to the PC800 RIMMS that go in an i850 (or i820 for that matter). Last I read, they were the exact same spec. Now, memory controllers are wonderful things, but they dont alter the underlying memory architecture.
If you care to disagree in public, then provide some evidence supporting your claim. Alot of us PC folk digestrd buttloads of articles on Rambus over the past year or two, so if you can come up with something to completely blow over its core characteristics, by all means.
Personally, I'm all in favour of serial interfaces. Despite Rambus' proprietry-lawyer-death-murder-kill practices, I'd like to see their tech trickle down into alot more hardware.
zurich
If you care to disagree in public, then provide some evidence supporting your claim. Alot of us PC folk digestrd buttloads of articles on Rambus over the past year or two, so if you can come up with something to completely blow over its core characteristics, by all means.
Personally, I'm all in favour of serial interfaces. Despite Rambus' proprietry-lawyer-death-murder-kill practices, I'd like to see their tech trickle down into alot more hardware.
zurich
Zurich, You can't compare the RamBUS implementations used in PCs to the one in PS2. An onboard RamBUS memory controller makes all the difference.
Latency for the upcoming EV7 Alpha which has umpteen RamBUS controllers on the CPU die is said to be 75ns for an open page. The recent presentation of Opteron states around 55 ns for an open page hit on DDR333. So yes DDR has better latency, but RamBus can have *alot* more pages open at any given time, so the timings are not that different. The biggest mistake RamBUS made was that the protocol doesn't support critical word first in a burst.
So it comes down to cost. RamBUS adds cost in on-die logic, but saves cost in onboard traces.
Cheers
Gubbi
Latency for the upcoming EV7 Alpha which has umpteen RamBUS controllers on the CPU die is said to be 75ns for an open page. The recent presentation of Opteron states around 55 ns for an open page hit on DDR333. So yes DDR has better latency, but RamBus can have *alot* more pages open at any given time, so the timings are not that different. The biggest mistake RamBUS made was that the protocol doesn't support critical word first in a burst.
So it comes down to cost. RamBUS adds cost in on-die logic, but saves cost in onboard traces.
Cheers
Gubbi
aaaaa00 said:randycat99 said:
Pfft. He already has a grudge against Microsoft.
I'd call that a pretty hefty sized grudge.
i see we are again in the same schema.. he's saying negative things about microsoft.. booh he is a baad anti-MS guy..
did you at least see the documentary he was talking about ? if so, and if you don't agree with his analysis, explain me how you disagree..
it sounds like saying something negative about microsoft make you criminal of subversive thoughts, a counter-revolutionnist ?
xbox doesn't sell so well in europe => it's because european are anti-americans
a very talented hacker critizes the xbox and microsoft => it's because he have strong personnal problems with microsoft.
i'm wondering who is so biased.. it's just not "bunnie"..
you just trying to decridibilize this guy pretending he gives more priority to his personnal feelings that to truth or his knowledge.. you make him sound like a fanb*y: someone who, despite his knowledge, is eventually not very smart.
so ?
what's your point ?
Magnum PI said:aaaaa00 said:randycat99 said:
Pfft. He already has a grudge against Microsoft.
I'd call that a pretty hefty sized grudge.
i see we are again in the same schema.. he's saying negative things about microsoft.. booh he is a baad anti-MS guy..
did you at least see the documentary he was talking about ? if so, and if you don't agree with his analysis, explain me how you disagree..
it sounds like saying something negative about microsoft make you criminal of subversive thoughts, a counter-revolutionnist ?
Where did I say having a grudge against Microsoft is BAD?
I merely said I think he has a grudge against Microsoft.
Magnum PI said:
And there you go putting words in my mouth, I said no such thing.
Magnum PI said:
My point is, IMHO, Faf has a lot more credibility regarding the overall performance of PS2 than bunnie does. Likewise, I'd believe something ERP said about xbox over something that bunnie says.
The reason is that hacking the boot procedure of a console doesn't necessarily give you any insight as to how it actually performs in a game.
I have a couple places where I disagree with him:
My answer is "so what?".
The xbox is not a low power mobile device, who cares if the CPU consumes more power, as long as it's within your design limits.
The economies of scale and superior fabrication processes that Intel has can overcome the cost of the wasted silicon area. So the chips are marginally more expensive? So the chips are marginally bigger?
So what?
The benefit of not having to have developers learn a new architecture and build new tools is a pretty big bump in x86's favor.
Having a great compiler right off the bat is a big plus. What's the problem with that?
You don't choose a CPU for a game console because it's beautiful and low power and efficient and elegant.
You choose a CPU for a game console to run games.
And if the criteria for the design are such that x86 turns out to be a good choice, then so be it.
Is it? DDR-SRAM might have 10x the latency of 1T-SRAM, but it's at least comparable to the RDRAM in the PS2. Does this mean PS2 is a steaming heap of dung too?
Didn't Nintendo have IBM fab their chips?
So, who owns Nintendo's mytical fab?
Didn't Nintendo get their GPU from ATI?
So, who owns the IP on the ArtX chipset?
Didn't Nintendo get a PowerPC from IBM?
So, who owns the IP on the PowerPC?
Anyway, I'm not saying that bunnie isn't a smart guy, there's no doubt he's brilliant.
But some of the stuff he says in his tirade on icrontic just makes no sense to me.
My answer is "so what?".
The xbox is not a low power mobile device, who cares if the CPU consumes more power, as long as it's within your design limits.
The economies of scale and superior fabrication processes that Intel has can overcome the cost of the wasted silicon area. So the chips are marginally more expensive? So the chips are marginally bigger?
So what?
The benefit of not having to have developers learn a new architecture and build new tools is a pretty big bump in x86's favor.
Having a great compiler right off the bat is a big plus. What's the problem with that?
You don't choose a CPU for a game console because it's beautiful and low power and efficient and elegant.
You choose a CPU for a game console to run games.
And if the criteria for the design are such that x86 turns out to be a good choice, then so be it.
Is it? DDR-SRAM might have 10x the latency of 1T-SRAM, but it's at least comparable to the RDRAM in the PS2. Does this mean PS2 is a steaming heap of dung too?
Didn't Nintendo have IBM fab their chips?
So, who owns Nintendo's mytical fab?
Didn't Nintendo get their GPU from ATI?
So, who owns the IP on the ArtX chipset?
Didn't Nintendo get a PowerPC from IBM?
So, who owns the IP on the PowerPC?
Anyway, I'm not saying that bunnie isn't a smart guy, there's no doubt he's brilliant.
But some of the stuff he says in his tirade on icrontic just makes no sense to me.
aaaaa00 said:
you are essentialy missing his point - it's not DDR which has big latencies (compared to SDRAM or RDRAM, that is), it's the UMA which introduces a whole new level of latencies here; when a resource such as mem is generously shared you get every part in the system waiting for the one part which is bursting its data to/from mem ATM. IMHO, coming up with a low latency UMA solutution would be anything but cheap - not apt for a console.
i believe N owns the IP in this case. correct me if i'm wrong.
BenSkywalker
Regular
Randycat-
Take off the blinders for a second here. The XBox is a steaming pile of dung because it isn't the design philosophy he would have chosen?
Is the XBox the most powerful console on the market?
It's a simple question we all know the answer to, and yet in his eyes it is still a pile of crap? Do you see anyone in this thread commenting on his statements regarding the poor encryption used in the XBox? I don't. I wouldn't make a comment on it as it is very clearly obvious that he has extensive knowledge in that area while I doubt anyone here could compare.
That said, we do have a few developers in these forums and none of them agree with what he has said and in fact have pretty much stated the exact opposite. These are people that have spent hundreds or thousands of hours tweaking software to run on the XBox versus a guy who doesn't even really use it to play games, he simply worked around the security.
His talking seems like the same line of thought we hear from Mac advocates concerning hardware 'superiority'. If it is designed in a more "elegant" fashion it is superior no matter if it is more expensive and slower. He seems to suffer from the same misplaced idealism.
Yes, he has a PhD from MIT, does that mean we should all listen to every word he has to say as the gospel truth? When he can point to a console that outpowers the XBox by a clear margin he can start to talk and have some credibility on the subject. If you read an article from an automotive engineer stating that the McLarenF1 was horribly designed would you listen to them?
Take off the blinders for a second here. The XBox is a steaming pile of dung because it isn't the design philosophy he would have chosen?
Is the XBox the most powerful console on the market?
It's a simple question we all know the answer to, and yet in his eyes it is still a pile of crap? Do you see anyone in this thread commenting on his statements regarding the poor encryption used in the XBox? I don't. I wouldn't make a comment on it as it is very clearly obvious that he has extensive knowledge in that area while I doubt anyone here could compare.
That said, we do have a few developers in these forums and none of them agree with what he has said and in fact have pretty much stated the exact opposite. These are people that have spent hundreds or thousands of hours tweaking software to run on the XBox versus a guy who doesn't even really use it to play games, he simply worked around the security.
His talking seems like the same line of thought we hear from Mac advocates concerning hardware 'superiority'. If it is designed in a more "elegant" fashion it is superior no matter if it is more expensive and slower. He seems to suffer from the same misplaced idealism.
Yes, he has a PhD from MIT, does that mean we should all listen to every word he has to say as the gospel truth? When he can point to a console that outpowers the XBox by a clear margin he can start to talk and have some credibility on the subject. If you read an article from an automotive engineer stating that the McLarenF1 was horribly designed would you listen to them?
What the hell do game developers opinions have to do with memory latency and architectural efficiency? It is no more their area of expertise than the hardware encryption.
Game developer's opinion on its relative merits compared to the PS2 and GC are important, but he doesn't actually say it is worse than those from an overall performance point of view ...
If I read an article from an automotive engineer which stated that the F1 could have been build for much lower cost, and hear a hundred game developers state otherwise Ill still be inclined to put more faith in the automotive engineer.
Game developer's opinion on its relative merits compared to the PS2 and GC are important, but he doesn't actually say it is worse than those from an overall performance point of view ...
If I read an article from an automotive engineer which stated that the F1 could have been build for much lower cost, and hear a hundred game developers state otherwise Ill still be inclined to put more faith in the automotive engineer.
BenSkywalker
Regular
Unlike him, they have to deal with it for thousands of hours in real world, not philosophical terms.
How are you reading the following statement?
Considering he explicitly states it is sub par, not to mention a steaming pile of dung, I'm inclined to think he thinks there is something better. If not, then par must exist only in his imagination
There are plenty of examples of his "depth" of knowledge on the XBox in the interview-
Not only is the XBox sub par and a steaming pile of dung, but MS is also taking a $250 loss per console(do the math)
If it cost you the same as the other vehicle you were looking at, would it matter in the least to you? Philosophy versus end results. I'll take end results every single time.
marconelly!
Veteran
The cost (being end result) means little to me. I would gladly pay twice as much if the end result of an Xbox was a small, stylish package vs what they made.
A game developer's opinion on memory latency counts for a lot, because unlike an armchair architect, a game developer has actually had to make a game work on the hardware.
So while they might not know how individual architectural features contribute to the memory system performance, they do know the actual overall memory system performance.
In the case of the Xbox vs. the PS2, I think it's the case where the Intel / NVIDIA memory hierarchy has lots of caches that hide the latency. The PS2's memory hierarchy apparently does a much poorer job of hiding the latency. So while it may be true that certain parts of the PS2 hierarchy are faster than the equivalent parts of the Xbox hierarchy, when it comes to the effective latency seen by the game, the Xbox seems superior.
Exhibit 1: Cross platform games
There have been many cross platform games written to run on all three consoles. Most of these games have been optimized to run on the PS2 first, and then ported to the other platforms. It seems to me that this process would have ruthlessly exposed any glaring flaws in the Xbox architecture (compared to the PS2). But so far we haven't seen any. The PS2 ports almost always look as good on the Xbox as they did on the PS2. The bad Xbox games have typically been PC ports and/or Xbox-specific. That seems to indicate that the Xbox is at least as capable as the PS2.
Exhibit 2: GDC talks and gamedev mailing list traffic
Compare the GDC and Euro GDC talks on PC game optimization vs. GDC talks on PS2 game optimization. The PC talks are never about improving memory latency, while the PS2 talks often include discussions about how to avoid using the CPU to fetch from main memory.
Similarly, if you read the gamedev mailing list, you'll often encounter discussions on how to avoid fetching from main memory (e.g. by avoiding using virtual function pointers) in order to get decent performance on the PS2.
BenSkywalker said:
I read it in the context of him being an engineer evaluating the quality of the design.
Considering he explicitly states it is sub par, not to mention a steaming pile of dung, I'm inclined to think he thinks there is something better. If not, then par must exist only in his imagination
If my plan was to sell everyone a Cray with appropriate inputs and outputs as a console would you call that a good design?
Exactly, but with cars what m$ is doing would be called dumping and would be illegal ... m$ using their funds to write off the entire first generation of the XBox as a loss leader is bad for competition.
BenSkywalker said:
who do you think design consoles ?
software developpers ?
of course not...
using you reasonning the people who design consoles (who probably have the kind of qualifications bunnie has) are not qualified enough to discuss memory latency and architectural efficiency..
well
How are you reading the following statement?
Considering he explicitly states it is sub par, not to mention a steaming pile of dung, I'm inclined to think he thinks there is something better. If not, then par must exist only in his imagination
to paraphrase mfa:
he doesn't actually say it is worse than those from an overall performance point of view ...
this statement doesn't specifically deal with the performance aspect of the xbox but more with other ones, like its electronic design.
and when he says "sub-par", i think it isn't a comparison to other consoles but more generally to other electronic hardware. i mean there are other things than consoles that use electronics..
perhaps the problem is not what he said but what you read..
Nope, that's a common misconception. Dumping doesn't mean selling a product for less than it costs to make something. Dumping means selling a product for less in one market than you do in another market, for the purpose of driving out competitors.
What would people have said if MS had come out with the PS2, while Sony had come out with the Xbox? (Leaving box size and style out of the discussion, and purely discussing the software architecture.) I can imagine discussions like this: "Microsoft has done their typical job of slapping together a kludgy, non-standard DOS-like machine that's insanely hard to program, while Sony's provided a much more elegant, more powerful system." BenSkywalker
Regular
Philosophy versus end results.
If you included an IR graphics sub system and sold it to me for $199 I'd say a lot nicer things then 'good design' for damn sure.
In the console market it is called, well, the console market. The entire business model is based around making nothing to taking a loss on the hardware and making the money back on the software. On that front, MS has the highest tie in ratio at comparable life cycle points of any console ever. If they continue with that trend then their loss per unit would easily be offset making them come out ahead in the long run. That is ignoring all the long term benefits superior hardware can reap you in terms or market penetration and mind share.
You're comparing apples and coconuts here.
Typical modern x86 cpu has 10+ times more cache then R5900 core in EE.
CPU accessing main memory is something you always want to avoid as much as possible(pretty much regardless of the architecture), difference is that PC cpus do most of that work for you.
R5900 on the other hand is akin to running a P3 with L1 cache halved, and L2 disabled - Suddenly the problem is almost entirely thrown to the programmer to handle.
The other problem (something which less informed people usually don't realize) is that PS2 processing flow essentially is an UMA, and that only serves to accentuate the above problem.
Edit:
On that same note, calling UMA 'inferior' is essentially putting down all three consoles
, not just XBox.Johnny Awesome
Veteran
The other thing that's important to point out is that you need to look at the Xbox as an ecosystem of consumers, developers, retailers, publishers, and Microsoft. Creating the strongest ecosystem possible should be Microsoft's main focus.
Sacrificing some design elegance for ease of development was a smart choice for MS to make, because it helped developers, which breeds more software, which consumers want, which helps retailers sell product, which helps MS recoup their original investment and make a profit.
You would expect an engineer to look at only the engineering aspects of the Xbox, which are still half-decent, but Bunnie ignores the fact that by making those sacrifices MS gained a lot for the entirety of the Xbox ecosystem.
Sacrificing some design elegance for ease of development was a smart choice for MS to make, because it helped developers, which breeds more software, which consumers want, which helps retailers sell product, which helps MS recoup their original investment and make a profit.
You would expect an engineer to look at only the engineering aspects of the Xbox, which are still half-decent, but Bunnie ignores the fact that by making those sacrifices MS gained a lot for the entirety of the Xbox ecosystem.
