predict how actual Xbox Next will differ from leaked specs

I never said that.

I've stated its a clean sheet design but haven't said how more cores or what ISA it uses. Its can be clean sheet and have the rumoured configuration.

After all Cell is a single core PowerPC (as confirmed by the ISSCC)... but it still very much a clean sheet design.

OT but slight funny, I forget wether it was ISSCC or ISCCC. For some reason I can't imagine Cell being disclosed at the Irish Setter Club of Central Connecticut

 

How much of the Xbox's user base is predicated on having the most powerful hardware?

MS runs a risk by being first one out of not having the most powerful hardware. Not just with the silicon but in the areas of optical disc, local storage.

Is the strategy to carry the curren user base forward or try to work with a clean slate?

Besides hardware spec., backwards compatibility and software (will the library be mostly the hardcore titles or a lot of sequels?) will provide a clue about their strategy.

 

Sorry to put in your words something that you did not said.

I was made this:
clean sheet design + its not remotely related to a G5 + [only] shares a similar ISA = no PowerPC

here http://www.beyond3d.com/forum/viewtopic.php?p=430998&highlight=#430998

The tricore coments is only made because the "leak" without relation to you.

I should pay more attention to this http://www.beyond3d.com/forum/viewtopic.php?p=436009&highlight=#436009

Meybe I should rethink my thoughts about XeCPU, now I hope even more that their GPU is really more capable of general porpose ops...or have hope that they can do a CPU has good has Cell

Thanks the correction

 

going by the diagram i say the r500 has roughly(assuming same clockspeed) twice the shader power than a x850XTPE(540MHZ) including vs/ps and 50% the pixel fillrate

 

According to the recent patent Xenon is a kind of multicored GameCube (XCube? ).

 

I dont know - it doesn't look that way to me.

X800/X850 has 16 pixel pipelines, each with 2 ALU's ( split to give vec+scalar ) plus a texture fetch unit.

In the diagram there's a comment about 16 bilin tex. fetches / cycle -> implying 16 pipelines again ( Could be read another way, but doesn't make too much difference ) and 48 ALU ops /cycle. This would give 3 ALU's per pipeline ( a 50% increase )

 

I thought the more interesting part was that it happens to reaffirm much of what has been said in the leaks. Including some of the GC similarities

Different kind of chips. Considering that Cell samples at ISSCC(Deano you better not be misleading me with spelling here ) were apparently running at 4.6Ghz, high mhz are apparently not out of IBMs reach.

 

In the xenon hardware overview it's 48ALU's each can execute a scalar and vector op per clock similar to what x850 does. That's where the 48billion shader ops come from in the first place(96shader ops * 500MHZ = 48000)

 

Well, that would be funny though .

 

The NV2A had 8 pipes and a 233mhz clock frequency, which gives 932 megaPIXELS/s and 1864 megaTEXELS/s. People constantly confuse this two terms.

BTW, geforce3 has 8 pipes and a 200 mhz clock frequency, which gives 800 mpixels/s and 1600 mtexels/s. And NV2A had 2 vertex shaders, so it was a underclocked gf4 ti.

 

NV2A has 4 pipes, not 8. 233x4 = 932 MegaPixel/s.
When it's not doing color writes it can output more than one zixel per pipe per clock. It seems the maximun sustainable fillrate in this case is 1600 megapixel/s AFAIK

ciao,
Marco

 

Cpu dual tri core cpus at 3-3.5ghz with 1meg cache per core .

Gpu a 16- 24 pipe line r500 with new fsaa modes that use less bandwith . I'm expecting 3x the power of the x800xt .

Ram 256 megs of gddr 550mhz ram with a 512 bit bus


120 gig sata 10rpm drive with 8 meg cache .


Thats all my thoughts for now , back to bed.

 

I think some of you guys are insane.

Look for a $200-$300 price point at a mid Q3 to mid Q4 launch and then re-evaluate your ideas on what will be more plausible in that time frame. If you don't then I highly suspect a lot of you will surely be disappointed.

Tommy McClain

 

...cause I'm goin insane... Insane in the membrane... Insanse in the brain!

heh that always pops into my head when I see the word insane.

 

I totally agree.

And I think some people expect the PS3 to have 4 gigs of RAM and 32 CPU's.

People need to come down to earth. It's a console - it will cost between $300 and $400 bucks.

If you want top of the line HW - then spend $3K on a PC and constantly upgrade it.

 

who is crazy ?

Ms can launch a 500$ system at launch and have it quickly scale down in price.


Even if u get 30$ for the gpu at 90nm at 80nm it should drop a bit , 65nm it should drop more , 45nm will drop even more . That should all happen in the first 3 years of the console life

The 512 bit buss wont go down with price , but its a way to keep extremely low quanity high speed ram out of the system. Instead of 1000mhz ram they can get away with 500mhz ram which is already being used in the mid end cards from ati and nvidia. That 500mhz gddr ram would be cheap adn on its way to being dirt cheap by 2005.

120 gig sta 10krpm 8 meg cache drive is going for about 120 now , I'm sure if ms bought them in bulk in the millions they'd get them for under 100$ most likely around 50ish with the price constantly droping .


So yea ms wants to save some cash , but they want to save it in the long run not up front. The problem with the xbox is that they were buying the nv2a at a fixed cost through out its life

 

The manufacturing process dropping twice in the first three years? I don't think so. It will liekly drop once over the course of the consoles life. I think you underestimate what it takes to reduce cost. you need to sell enough units to warrent putting money back into a system shrinkage.

Not a chance of that happening IMO. 120 dollars price for one component is way too much for a machine that is supposed to drop in price over 5 years. you'll find that harddrive prices drop at a rate that is much slower then other parts used in consoles.

 

PS2 was produced at 180nm, no? Wasn´t it later reduced to 130nm and then to 90nm??

I agree, IMO it´s crazy to believe that a 120 USD component can be fitted inside the machine, especially if it´s a HDD.

 

Why not ? did not the nv2a start at 150nm process ? Looks to me that ati will have a 90nm part in early 2005 while the xbox is still being made. The main problem was nvidia would get the chips fabed then sell them to microsoft so the costs would not go down quick enough.

No ms can be as agressive as they want .

If a new process comes along and I have no doubt that 65nm wil lbe avabile in 2006 if not sooner than that. Ms can move production from 90nm to 65nm if they feel they will get the desired yields and cost benfits ms can and will switch to the newer process.

Some people here feel in 2006 sony will launch the cell chip in the ps3 on 45nm , why can't ms go from 90nm to 65nm in the same time if not smaller ?


I will agree that the hardrive may be a hard sell. But it may still end up getting in. Perhaps an 80 gig sata drive. I was able to get one for 60$ . It had 8meg cache. I'm sure buying in bulk would get that down to the 30s .


Remember in a 4-5 year life span the first year isn't a big deal. If they take a loss in the first year but are able to scale the price down quick enough to price the unit agressively against the ps3 and ns6 while recouping costs or breaking even in its 2nd year going on they wil lbe in a fine position.

 

I believe you just answered your own question.