Blogcast Audio of Tech Engineers(?) on PS3 vs Xbox360

[notAFanB]

A:TFLOPS really comes down to I think who's willing to really push the meaning of floating point calculation and who's willing to dig in and kind of lack of a better word fake the numbers and see who can get the best number.

regardless of whatever measurement they put up surely the Hardware Documentation is gonna be an bitch to work around with this much injection of BS from both parties.

It's genuine but it doesn't make PS2 more powerful like SONY would like people to believe. OTOH Xbox IS more powerful.

see above, non-holistic figures are just getting out of control

 

[ndoogoo]

Nintendo should do some marketing with Gflops.

GFlops mean how many crap games flopped or estimated to flop and that system.

Gflops= game flops
Tflops = terrible flops

 

[PC-Engine]

Nintendo should do some marketing with Gflops.

GFlops mean how many crap games flopped or estimated to flop and that system.

Gflops= game flops
Tflops = terrible flops

 

[Shifty Geezer]

It's rather like how Streve Ballmer says cell can't be more powerfull than XB360's as cell only has one core, he's not exactly lying, its about presentation.

I disagree. This is a straight lie. MS talk of SPE's as being vector units or DSPs. They are full processing cores, able to do everything a PPC core does in terms of the types of code they can run. They aren't as effecient at branching, general purpose code. They are more effecient at floating and integer maths on streams. Specialised, yes. But full cores all the same.

AFAIK Sony have used GFLop benchmarks that have been validly derived if misrepresentative peak figures, and hyperbole that's just marketting nonsense (like StarWars in real time). I don't know that they've every lied about the hardware implementation on a rival, though I wasn't on the net when PS2 was launched so don't know what went on there. For me, I can take a bit of 'creative licensing' when it comes to specs, but actually fibbing about how your opponent's machine works really takes the biscuit.

 

[London Geezer]

Still waiting for the 256GB/s to main memory explanation...

:|

 

[gmoran]

SONY started using GFLOPS as a performance metric. MS is only doing it in return. An eye for an eye.

1TFLOPS isn't being honest, 2 TFLOPS isn't either. It's just funny that SONY is stuck in this GFLOPS/TFLOPS game that they started.

I know you are not being entirely serious here, but Sony's use of GFLOPS was as legitimate as these things can ever be termed legitimate; whereas NV's original use of NV FLOPS, and MS's and Sony's subsequent use of that metric isn't.

 

[PC-Engine]

It's rather like how Streve Ballmer says cell can't be more powerfull than XB360's as cell only has one core, he's not exactly lying, its about presentation.

I disagree. This is a straight lie. MS talk of SPE's as being vector units or DSPs. They are full processing cores, able to do everything a PPC core does in terms of the types of code they can run. They aren't as effecient at branching, general purpose code. They are more effecient at floating and integer maths on streams. Specialised, yes. But full cores all the same.

AFAIK Sony have used GFLop benchmarks that have been validly derived if misrepresentative peak figures, and hyperbole that's just marketting nonsense (like StarWars in real time). I don't know that they've every lied about the hardware implementation on a rival, though I wasn't on the net when PS2 was launched so don't know what went on there. For me, I can take a bit of 'creative licensing' when it comes to specs, but actually fibbing about how your opponent's machine works really takes the biscuit.

Can't you program a DSP into functioning like a self contained CPU?

 

[blakjedi]

It's rather like how Streve Ballmer says cell can't be more powerfull than XB360's as cell only has one core, he's not exactly lying, its about presentation.

I disagree. This is a straight lie. MS talk of SPE's as being vector units or DSPs. They are full processing cores, able to do everything a PPC core does in terms of the types of code they can run. They aren't as effecient at branching, general purpose code. They are more effecient at floating and integer maths on streams. Specialised, yes. But full cores all the same.

AFAIK Sony have used GFLop benchmarks that have been validly derived if misrepresentative peak figures, and hyperbole that's just marketting nonsense (like StarWars in real time). I don't know that they've every lied about the hardware implementation on a rival, though I wasn't on the net when PS2 was launched so don't know what went on there. For me, I can take a bit of 'creative licensing' when it comes to specs, but actually fibbing about how your opponent's machine works really takes the biscuit.

Can't you program a DSP into functioning like a self contained CPU?

Am I the only peson tired of this argument? Everyone is going to spin plain and simple.

 

[PC-Engine]

SONY started using GFLOPS as a performance metric. MS is only doing it in return. An eye for an eye.

1TFLOPS isn't being honest, 2 TFLOPS isn't either. It's just funny that SONY is stuck in this GFLOPS/TFLOPS game that they started.

I know you are not being entirely serious here, but Sony's use of GFLOPS was as legitimate as these things can ever be termed legitimate; whereas NV's original use of NV FLOPS, and MS's and Sony's subsequent use of that metric isn't.

I don't know why you're still arguing this. You cannot compare EE to a P3 because EE needs to do all the TnL that a P3 does NOT need to do. The TnL is done by the NV2A so if you want to compare GFLOPS then you need to compare the EE vs the CPU+GPU understand?

 

[gmoran]

It's rather like how Streve Ballmer says cell can't be more powerfull than XB360's as cell only has one core, he's not exactly lying, its about presentation.

I disagree. This is a straight lie. MS talk of SPE's as being vector units or DSPs. They are full processing cores, able to do everything a PPC core does in terms of the types of code they can run. They aren't as effecient at branching, general purpose code. They are more effecient at floating and integer maths on streams. Specialised, yes. But full cores all the same.

Yeah but Steve Ballmer talking to some journo isn't an official specification, and while he was definately twisting a bit, I think it was more spin than lie. ow if MS had printed it as a side by side spec analysis then yup I agree it would be a bare faced lie.

 

[gmoran]

I don't know why you're still arguing this. You cannot compare EE to a P3 because EE needs to do all the TnL that a P3 does NOT need to do. The TnL is done by the NV2A so if you want to compare GFLOPS then you need to compare the EE vs the CPU+GPU understand?

I'm not comparing EE to a P3, I'm saying Sony's original use of the term GFLOP was legitimate, and that subsequent NV FLOP specs are not.

 

[PC-Engine]

I don't know why you're still arguing this. You cannot compare EE to a P3 because EE needs to do all the TnL that a P3 does NOT need to do. The TnL is done by the NV2A so if you want to compare GFLOPS then you need to compare the EE vs the CPU+GPU understand?

I'm not comparing EE to a P3, I'm saying Sony's original use of the term GFLOP was legitimate, and that subsequent NV FLOP specs are not.

Their comparsion and intent wasn't legitimate. MS's number wasn't legit, but Xbox was more powerful than PS2 anyway so it's kinda moot.

 

[gmoran]

Their comparsion and intent wasn't legitimate. MS's number wasn't legit, but Xbox was more powerful than PS2 anyway so it's kinda moot.

I've never seen an official Sony PS2 to Xbox spec, so I can't comment. Are you sure they published one?

Xbox was more powerfull than PS2.

 

[marconelly!]

I've never seen an official Sony PS2 to Xbox spec, so I can't comment. Are you sure they published one?

Yes, there was a funny comparision chart on MS Xbox website that just completely misinterpreted many GC and PS2 specs (GC especially) and played up XBox specs.

 

[MechanizedDeath]

SONY started using GFLOPS as a performance metric. MS is only doing it in return. An eye for an eye.

1TFLOPS isn't being honest, 2 TFLOPS isn't either. It's just funny that SONY is stuck in this GFLOPS/TFLOPS game that they started.

Curious...when did Sony start this "game"? Are you talking about the PS2 press releases, of which the GFLOPS figure was merely one of a plethora of other numbers? I hope you don't mean that, b/c that press conference also took place well before the Xbox/GC were even announced. It's not like they were trying to one-up the competition, they were merely throwing out a number. You could say the same thing about the polygon count or the fillrate. And AFAIK, we knew the FLOPS rating for the DC's SH-4 before that.

If you're talking strictly about this gen, then IBM was in charge of that ISSCC presentation. And besides which, GFLOPS were being calculated by regular posters on here before that. I don't know where you're going with this. :? PEACE.

 

[Alpha_Spartan]

MS goes to greath lengths to damage control the fact that x360 overall will be less powerfull then the ps3.

And you base that on what? Sony numbers?

 

[Fafalada]

It's just funny that SONY is stuck in this GFLOPS/TFLOPS game that they started

Funny, I could have sworn Sega started it with DC(making direct comparisons to PC chip Flops of the era too), unless there's some kind of time-travel-anomalies involved.

 

[aaaaa00]

I look forward to your take on it then. Please explain to me (and I guess others here to) how the use of main RAM in PS3 is poor for general code and how XeCPU doesn't have to write data and read data from main memory.

The most time consuming parts of general code are the parts that walk in-memory structures using load and store from main memory, and make branches based on those values. Integer math ops are not the issue here.

What I think they're trying to get at is that if you want an SPU to access main memory, you have to schedule a DMA from it to local store. This is not so different from constructing an DMA packet for read and writing to a disk device, just for example.

Doing this is a lot more complicated for the developer and takes more time to execute than just executing a single "load <address in main memory>" instruction.

You can argue that with careful optimization and structuring of your code and data structures, you can make a DMA/LS system work more efficiently than a cache type system, but it takes a lot more effort and it's only practical for certain kinds of work loads.

If your code does a lot of random main memory access and structure walking/pointer chasing that won't fit in the SPU's LS, then IMHO the cache based system will tend to outperform the DMA/LS system in overall performance.

 

[Alpha_Spartan]

I think MS hit the nail on the head with this odd viral marketing campaign. They're pandering info to 1st semester CS students and internet pseduo-geeks who are going to get all hyped up about misinformation they have no educational or experiential knowledge to disassemble for validity. Thus, they go off and spout it to their friends and parents who treat them as tech experts and that sells units.

...Which makes it a perfect way to balance out the PS3 press conference which was much of the same stuff. See if these were Sony engineers, their word would be taken as gospel, but since these guys are from M$ (I'm cool cuz I used the "$" at the end denoting that all they care about is making money), they're obviously bending the truth. It's tit for tat, eye for an eye, BS for BS.

 

[Shifty Geezer]

What I think they're trying to get at is that if you want an SPU to access main memory, you have to schedule a DMA from it to local store.

Thanks for the explanation aaaaa00. That's actually a good point then. So the SPE's haven't got direct addressing per se? I thought that was implemented. Does the memory request go through PPE or the FlexIO/some other memory controller? Also what's the overhead (transistor count I guess) for implementing direct memory access that STI felt it worth going without? Considering they're suppoed to work on voluminous data streams, not having easy access to main RAM locations seems a bit daft