Rumor so dont post hate Im telling you up front.

[route66]

What's the difference between 2 16pipe cores or 1 32pipe core?

I call shens.

 

[Malak]

Originally posted by: HighCalibreHooch
Ati releases products? I need to sit down ....

You know, people like you were complaining less than 6 months after the previous generation of cards. SLOW DOWN. ATI sold out so many cards. People don't need to upgrade every 4 months, for pete's sake.

 

[BenSkywalker]

Buddy of mine who is usually right about 90% of the time is telling me that the R600 was designed with Multi-Core in mind like the AMD cpu's were originally designed.

GPUs are multicore and they have been for YEARS.

I wish I could get a sticky on this....

CPUs went to the multicore format as they realized that in order to continue to increase performance at close to the prior rate they were going to have to rely on parallelism, not frequency. As Moore's Law explicitly deals with trasnsistor count- not clock rates- the advances in being able to increase the amount of parallel operations in a processor significantly exceeds the ability that they have to increase the clock rate(as transistor density has enormous growth potential for todays' relatively tiny CPUs).

GPUs have been built this way since last millenium.

You know those 'pipes' people talk about? Consider them cores, or if you want to push the issue divide the number of pipes by four as all modern GPUs work on quads. 7800GTX- 6 core GPU.

If anything ATi and nVidia should be charging Intel and AMD licensing fees.

 

[otispunkmeyer]

hang on a sec im nit sure i understand

GPU's are parallel by nature.....computer graphics are infinitely parallel are they not?

so how would having say, 2 24pipe GPUS on one die, be any different than having 1 GPU with 48pipes? id bet the latter would be slightly faster also. though i can see that it maybe easier to manufacture? 2 less complicated cores on die as aposed to 1 more complicated one?

i thought GPU's have always been "multicored" since they can process more than 1 thing at a time, ie like approximating the gpu pipeline or quad to a cpu core.

 

[Steelski]

the closes to the reasoniong behind multicore would be the two simpler dies than one complex one. maybee.....:S

 

[mwmorph]

Originally posted by: otispunkmeyer
hang on a sec im nit sure i understand

GPU's are parallel by nature.....computer graphics are infinitely parallel are they not?

so how would having say, 2 24pipe GPUS on one die, be any different than having 1 GPU with 48pipes? id bet the latter would be slightly faster also. though i can see that it maybe easier to manufacture? 2 less complicated cores on die as aposed to 1 more complicated one?

i thought GPU's have always been "multicored" since they can process more than 1 thing at a time, ie like approximating the gpu pipeline or quad to a cpu core.

1. slightly faster, probably, but the cost would be massively higher and it would be massively harder to manufacture. if you have 2 24pipe cores, if 1 quad dosent work, you throw only 1/2 the chip away, in 48pipe cores,there is a much migher chance of 1 quad or more not working due to a exponentially more complex chip and the cost is much higher.

 

[Topweasel]

Originally posted by: Ackmed
ATi released the XL 2 days late. ATi released the XT 3 days early. NV relesaed the 512mb GTX a few days early too, not that they are actually available anywhere though.

If you're going to complain about one, complain about them both, and dont be one-sided.

I am pretty sure that ATI released the X1800XT a month after they anounced. Wheter it was early for their estimated production to retail date makes little difference. I have always hate specailly around update time being told that in a 2-3 Months (GF2) the new card will be available but here is proof it kicks everyones ass. Sometimes ingnoance is bliss, and why I tell everyone make your purchase the best you can afford at the time you want to upgrade.

And while I don't like the Supply issues with the GTX 512MB one bit at least for people willing to pay for it they had the cards in quantity the day it was announced, and some places selling them even earlier, prior to seeing any kind of review (which to a degree is a stupid thing to do).

 

[5150Joker]

R600 multicore: dedicated physics chip + gpu core...i wish.

 

[crazydingo]

Originally posted by: Ackmed
ATi released the XL 2 days late. ATi released the XT 3 days early. NV relesaed the 512mb GTX a few days early too, not that they are actually available anywhere though.

If you're going to complain about one, complain about them both, and dont be one-sided.

Q F T

 

[akugami]

The ATI release date jokes are really old and truthfully, nVidia is not much better. Yes, the 7x00 series was released on time and excepting the 512MB version of the 7800 it was readily available after announcement. nVidia is absolutely to be commended for this. However, I remember last year roughly the same time as now when both ATI and nVidia had problems supplying video cards.

ATI absolutely needs to get on the ball as far as product releases goes. No one questions that, however, nVidia has only released one product in the last year and change that was on time and readily available. The last time I checked, one product release on time does not make it a regular occurance.

The move to 90nm has affected more than one company and that has to be taken into account as well.

Not trying to defend ATI, but complaining about their release record is kinda unfair when their competition hasn't been much better. If ATI still screws up the release date of the R580 while nVidia is on the ball then we got a serious complaint against ATI. At that point it would be inexcusable not to get your a$$ in gear with 1+ year to work out any kinks.

nVidia might have been very smart testing out 90nm on lower end parts and working out any potential pitfalls before moving their high end line to 90nm.

 

[coldpower27]

Nvidia so far has had some good hard luanches

Geforce 7800 GTX 256
Geforce 7800 GT
Geforce 6800 GS

The 7800 GTX 512 is a mixed bag, it was a hard launch but the availability could not be sustained due to the lack 1.1ns GDDR3 modules in production.

Geforce 6600 DDR2 is also a mixed bag as only 1 manufacturer that I know of has a product based on this card and that AIB Partner is XFX. So availability wasn't that good as not many other AIB made cards for this SKU.

 

[Drayvn]

Just had an idea... Here goes.

The R600 is supposedly having a unified shader architecture right? Well what use would there be in have a a Dual Core Unified Shader GPU? Wouldnt that be pretty pointless.

If all the shader pipelines can do everything what would be the point in having another core of the same shader pipelines on top of that? The CPU only went to dual core as this can increase efficiency it thought.

While dual core CPUs can increase efficiency of programs and applications that wouldnt that be what the unified shader architecture would be effectively doing as its what probably could be called a "dual core pipeline" as it can do pixel and vertex shaders except it just cant do it at the same time.