Support NeoGAF

[Panajev2001a]

Sony is going to have harder time reaching that bar if MS decides to go with 512mb...if at all.

Right, because GDDR-3 at 25 GB/s is so MUCH cheaper than XDR DRAM... :rolleyes.

Do you have a detailed cost analysis that supports your statement ?

 

[Panajev2001a]

Is the data transfer rate faster on 12x speed DVD drive than the standard HD-DVD or BR drive?

12x DVD = 16.62 MB/sec = 132 Mbps.

 

[Fafalada]

Do you have a detailed cost analysis that supports your statement ?

Problem is, NOONE debating on these forums does (or if anyone does, they refuse to provide any evidence to the matter). But it's been dubbed "common knowledge", so everyone just runs with it as they please.

Not unlike how everyone was slamming any machine using RDRam for "latency problems"(which was 'common knowledge' too) regardless of the fact that PS2 memory latencies were on par, if not faster then DDR/SDR solutions during the same time frame.

Not that I'm trying to argue XDR isn't more expensive, but I would like to know a little real detail about price difference before arguing what can be afforded or not.

 

[Pug]

I think a bigger cost than the memory on the PS3 would be a 4x BR. I just can't see how 1x drive is going to be of use, 2x would be okay but I think thats still slower than a 12x DVD. As for memory, lets be honest theres not going to be much of a difference between the machines.

 

[gofreak]

My post didn't mention anything about absolute maximum either.. just like KickyFast post, you will need to retool the design to add more ram to Cell, its not as flexible as what MS is using.. this would cause Sony to re-fabricate the cell design going into PS3 and I'm sure Sony is pass that stage.

a) We don't know yet if the limitation even exists..

b) "retool" is possibly an exaggeration of what could be required if such a limitation exists. Realworld tech speculates that only a minor change would be necessary.

But they'd have to go with either a different version of Cell than we've seen

It wouldn't have to be very different, though, at least according to Realworldtech. Consider, also, that the chip shown at the ISSCC will be a good year old (or more, even) by the time PS3 starts manufacturing. If Sony had the foresight to be dealing with this, that's probably more than enough time to effect necessary changes - though I'm no chip engineer, perhaps someone like Pana or Faf would care to comment.

Furthermore, there are other possible solutions - like giving the GPU its own memory, or others have speculated putting more memory out over the FlexIO inteface.

Anyway, this is still assuming such a limitation exists, which may not be the case..

Also, re. memory pricing, I would agree that it is silly to speculate at this stage how memory costs will compare across both systems, but from what we know, or what's been leaked of Xbox2's memory, it's of similar bandwidth to XDR...if XDR is expensive, Xenon's memory won't be cheap either

 

[Timedog]

If the patent shows that 256mb is the limit, then for the time being, with the limited information we have, how is that an invalid assumption?

 

[Kobold]

kobold is it you that used to post over at Bizarre creations? If so Heard anything on the grapevine regarding PGR3?

Ya, it's me! Heh, I feel so famous Was happening?

I'm not a moderator anymore there either, too busy with my work to keep playing PGR all the time and checking the forums which are quite busy. I hardly had an overview. I basically played so much preview builds of PGR2 that when I finally got the retail the passion died down a little.

No info for ya on PGR3. That it's coming seems a given though with all the Job postings, it ought to be sweet!

 

[mrklaw]

*if* 256MB is a current limitation for PS3, assumptions are being made that it is more complicated to add more memory due to retooling etc.

But why is there the assumption that Xbox2 can simply slap another DIMM in there? Although there is no information that it is limited in the same way as PS3, there is also no information that I've seen that it is easy

 

[Izzy]

The obvious advantage of the XDR memory system is the bandwidth that it provides to the CELL processor. However, in the configuration illustrated in figure 9, the maximum of 4 DRAM devices means that the CELL processor is limited to 256 MB of memory, given that the highest capacity XDR DRAM device is currently 512 Mbits. Fortunately, XDR DRAM devices could in theory be reconfigured in such a way so that upwards of 36 XDR devices can be connected to the same 36 bit wide channel and provide 1 bit wide data bus each to the 36 bit wide point-to-point interconnect. In such a configuration, a two channel XDR memory can support upwards of 16 GB of ECC protected memory with 256 Mbit DRAM devices or 32 GB of ECC protected memory with 512 Mbit DRAM devices.

Unfortunately, rather than providing insights into the issue of the CELL processor's memory capacity, the statements above may have in fact contributed to the impression that the XDR memory system is currently constrained in some way. The truth of the matter is that the XDR DRAM devices themselves are capable of supporting the 72 device configuration that would allow each CELL processor to directly address 4 GB of ECC protected memory (given 512 Mbit XDR devices). However, in order to support the XDR DRAM device in such a configuration, specific support must be built into the XDR DRAM controller interface. To date, IBM has not released details on the memory controller interface indicating whether the current incarnation of the CELL processor can support a 72 DRAM device configuration in the XDR memory system, or a less amount, i.e. 36 DRAM devices. Fortunately, regardless of whether the current CELL processor can explicitly support the 72 DRAM device configuration in the XDR memory system, the ability to address 72 XDR DRAM devices would require at most a relatively minor design change for the CELL processor. As a result, even if a DRAM device-count limitation exists for the current incarnation of the CELL processor, future CELL processors can be easily designed to rectify that limitation. "

 

[gofreak]

If the patent shows that 256mb is the limit, then for the time being

Does it? The "limitation" talk came out of the ISSCC, not the patents IIRC. And that's since been clarified by Realworldtech aka we don't actually know yet.

And anyway, the patents aren't gospel, obviously. They were written years ago, obviously things can be different as they developed the chip, and we've already seen differences in the chip shown at the ISSCC.

 

[Izzy]

If the patent shows that 256mb is the limit, then for the time being, with the limited information we have, how is that an invalid assumption?

256 Megabits is the limit? Oh notz!

 

[Izzy]

no problem, it happens to me all the time.

on topic: MS could make Sony sweat by upping to 512 mb RAM. even if PS3 edges xbox2 performance wise you may still see better looking games among other things on xbox2 by adding 512mb of ram.

I'm sure PS3 will have more than 512 Megabits of RAM.

 

[Shompola]

Ok I have talked to an MS rep and this is what he said to me.

three powerpc cores in one single chip operating at 3ghz(IBM has had some serious problems with heat and such so they have lowered the frequency from the initial 3.5Ghz and it will be on 90nm die size and not 65nm).

The cpus will have a shared cache memory of 2 megabytes with some ultra fast connection between the cache pool and the cpus resulting in no penalty whatsoever when they simultainously try to access the cache memory. The cache memory will be configured with a method called release consistency giving the programmer a lot of freedom when accessing the cache memory and main RAM.

THE GDDR-3 or something like that will be 512 megabytes, operation at 1066MHz externally. The memory controller will be on the three core cpu chipset for fast access.

An additional 10 megabytes of edram will be implemented for fast and rapid z-buffer updates and alike.

The end result is a console exceeding 1 teraflopes per second. Of course these are peak numbers and far from average numbers.

 

[Izzy]

So the cache has increased? The original figure was 1 MB. Most impressive.

 

[gofreak]

Ok I have talked to an MS rep and this is what he said to me.

three powerpc cores in one single chip operating at 3ghz(IBM has had some serious problems with heat and such so they have lowered the frequency from the initial 3.5Ghz and it will be on 90nm die size and not 65nm).

The cpus will have a shared cache memory of 2 megabytes with some ultra fast connection between the cache pool and the cpus resulting in no penalty whatsoever when they simultainously try to access the cache memory. The cache memory will be configured with a method called release consistency giving the programmer a lot of freedom when accessing the cache memory and main RAM.

THE GDDR-3 or something like that will be 512 megabytes, operation at 1066MHz externally. The memory controller will be on the three core cpu chipset for fast access.

An additional 10 megabytes of edram will be implemented for fast and rapid z-buffer updates and alike.

Thanks, I hope you're right

Also, this was a MS rep, or someone inside MS? Surely a PR person wouldn't be divulging this..

 

[Shompola]

It was my imagination :lol

 

[Ghost of Bill Gates]

*if* 256MB is a current limitation for PS3, assumptions are being made that it is more complicated to add more memory due to retooling etc.

But why is there the assumption that Xbox2 can simply slap another DIMM in there? Although there is no information that it is limited in the same way as PS3, there is also no information that I've seen that it is easy

The bits per chip is fixed, AFAIK. Sony has already stated that they are going with 512 Megabit sticks. At the ISSCC, the Cell shown had 4 slots, so that makes 2048 Megabits of memory (256 Megabytes).

In order to access more RAM, which is entirely possible, you have to access it through another bus. The bus will definitely be used for CPU <-> GPU communication at least.

It certainly makes more sense to have more slots for RAM (IMO) than to go with this alternate route. The alternate route makes some group of data in memory the black sheep of the program which would be... unfortunate to have to handle in software.

 

[gofreak]

It was my imagination :lol

:lol

I guess that's one way to comment on the seeming ubiquity of people with "sources" on here. I often wonder why people are so unquestioning of what certain people who claim to have contacts say - perhaps I missed when they proved their credibility, I probably did. I'm just too shy to ask outright in case of offending someone

In order to access more RAM, which is entirely possible, you have to access it through another bus.

...or, you add more banks. If there's no such 4-bank limitation. As there may well not be.

Or put memory off the GPU. That wouldn't be any different from the current PC paradigm in terms of coding, just with presumably faster buses. But, yeah, it'd be nice to have everything in one place for common access (logically, if not physically) by the CPU/GPU.

 

[xexex]

man, next-gen consoles pale in comparison to GSCube as far as RAM goes--that thing had 2560 MB RAM! :lol

 

[Ghost of Bill Gates]

It was my imagination :lol

:lol , I wish it were true.

I know developers always want more but you can see what the lack of ram does in games.

Halo 2 with LOD problems, Doom 3 with shorter levels and more load zones, 30FPS in racing games, Fable with less features and loading.Other than that, framerate problems, awful textures appeared in games.

512mb should be the starting point for next gen consoles.. It's the HD era!

 

[Redbeard]

:lol , I wish it were true.

I know developers always want more but you can see what the lack of ram does in games.

Halo 2 with LOD problems, Doom 3 with shorter levels and more load zones, 30FPS in racing games, Fable with less features and loading.Other than that, framerate problems, awful textures appeared in games.

But then you have games like Stranger's Wrath with huge, detailed levels and zero loading...

I guess the idea is to make it easier for the sloppy devs, but I think their games would be crippled no matter how much ram they've got to work with.

 

[Ghost of Bill Gates]

I should have a confirmation before the end of the week.

Any word yet Blimblim?

 

[Blimblim]

Any word yet Blimblim?

Nope, I keep getting this information from everywhere, but I still did not get this confirmation from a really trusty source. Most annoying.

 

[Ghost of Bill Gates]

understood.

 

[Duckhuntdog]

It's still 3 cores, but the clock has been dropped to 3 ghz.

 

[xexex]

It's still 3 cores, but the clock has been dropped to 3 ghz.

that's understandable, that the clock has been dropped from 3.5+ GHz to 3 GHz.

now hopefully the GPU gets increased from 500+ MHz to 700+ MHz ^__^

 

[Pimpwerx]

It's still 3 cores, but the clock has been dropped to 3 ghz.

Source? Seriously, let's cut down on the scuttlebutt. PEACE.

 

[Kleegamefan]

Ok I have talked to an MS rep and this is what he said to me.

three powerpc cores in one single chip operating at 3ghz(IBM has had some serious problems with heat and such so they have lowered the frequency from the initial 3.5Ghz and it will be on 90nm die size and not 65nm).

The cpus will have a shared cache memory of 2 megabytes with some ultra fast connection between the cache pool and the cpus resulting in no penalty whatsoever when they simultainously try to access the cache memory. The cache memory will be configured with a method called release consistency giving the programmer a lot of freedom when accessing the cache memory and main RAM.

THE GDDR-3 or something like that will be 512 megabytes, operation at 1066MHz externally. The memory controller will be on the three core cpu chipset for fast access.

An additional 10 megabytes of edram will be implemented for fast and rapid z-buffer updates and alike.

Supposedly this is bulshyte but it sounds pretty convincing and is probably very close to the truth anyway...I wouldn't be surprised if we see a triple core, dual-threaded PPC running at 3Ghz at least....

 

[Blimblim]

Source? Seriously, let's cut down on the scuttlebutt. PEACE.

That's pretty much confirmed by anybody on the dev side. The only spec that's still undecided seems to be the RAM.

 

[SentientStone]

I was talking to some guy who is a programmer at EA on Saturday and he said that the RAM just got upgraded to 512 just a while ago. They also have yet to get any PS3 development kits apparently.

 

[Rhindle]

I was talking to some guy who is a programmer at EA on Saturday and he said that the RAM just got upgraded to 512 just a while ago.

I really, really hope that's true.

They also have yet to get any PS3 development kits apparently.

Paying attention, GI.BIZ fans?

 

[Duckhuntdog]

Source? Seriously, let's cut down on the scuttlebutt. PEACE.

Sure, let me get you the email address of my friend at MS, I am sure he would love to have you emailing him throughout the day.

 

[gofreak]

Paying attention, GI.BIZ fans?

Hehe. Just because his friend hasn't seen them or gotten them doesn't mean somebody in EA doesn't have it..

They should probably expect them before E3 if they haven't got them already. The timetable was for Tool v1.0 to be out at some point between now and E3.

Also, not sure how it relates to the GI.biz article..key middleware houses always are among the first to get dev kits first, even before the likes of EA. Edge magazine also went with this story, and I'd trust them.

 

[inpHilltr8r]

I know developers always want more but you can see what the lack of ram does in games.

Halo 2 with LOD problems...

Yeah, that was a RAM problem

 

[pcostabel]

I really, really hope that's true.

Paying attention, GI.BIZ fans?

Didn't GI.BIZ said that only TOP developers have kits?

 

[Ghost of Bill Gates]

Originally Posted by Leshita:
I was talking to some guy who is a programmer at EA on Saturday and he said that the RAM just got upgraded to 512 just a while ago.

I really, really hope that's true.

I really, really really hope that's true.