Anand has the details about r520,rv530,rv515

[fallguy]

Not much more then it has for the last couple years, plus they've moved to 90nm which means they get more per wafer. Don't kid yourself, they still make a good chunk of money per card sold and yet the prices keep increasing every year. Consider the 9700 Pro (then high end) went for $400, now the high end is $600. That's a $200/33% increase in approximately three years time.
.

There are many other price increasing factors you seem to be ignoring. Such as, the 9700 Pro has 128megs, the XT has 512.

Despite that, its still the same MSRP of the 7800GTX, with twice as much ram. If you are going to complain about ATi's price, you should be really upset with NV's.

 

[Tim]

What do you mean with "rest of the pipe"?

ROPs, TMUs, Z.

 

[Rur0ni]

I can understand the whole "but its 512MB" argument, but as someone stated, nVidia will surely drop their prices a bit once ATI's cards come into play, and the better deal is apparent when the performance is identical but there is a $100 difference. Unless of course the XT and XL models perform moderately better. And all indications we've had thus far point to no? If it performs the same on defaults, but shines in AA/HDR modes then it could be worth it.

Don't like putting myself out there like that with speculation, now I feel vulnerable.

 

[Arty]

600Mhz, $599 MSRP? *sigh* XT *and* XLs have Dual-Slot designs, most likely, according to AT? Errr... Not exactly impressive. Unless there are things we are missing per-pipeline, ATI is in a great position to lose this round.

Uttar

The cards arent even out yet and we are ready to hand out the verdict already !

that would give the 7600 the about the same fillrate\pixelshading performance as the 7800gt - i cant see the only difference being memory bandwidth, the 7600 is most likey a 12 pipe card IMO.

G72M which I'm assuming is based on 7600 is a 12pp design. Take that of what you can.

 

[X-AleX]

Gosh!Ati must have been distracted with the design of Xenos.
Waiting for the R600

 

[Geo]

$599, which is what 7800GT launched at, does not seem unreasonable to me when the part has twice as much memory, and that memory is significantly faster. Availability will be more interesting, and if it is there then, like everything else, that MSRP will be undercut by street price after a little bit.

Where my cognitive dissonance comes in is only 50mhz difference on the core between XT and XL. Not believing it yet.

 

[Dave Glue]

Where my cognitive dissonance comes in is only 50mhz difference on the core between XT and XL. Not believing it yet.

It's the price for me:

X1800 XL 16 550MHz 625MHz 512MB GDDR3 $499
X1800 Pro 16 500MHz 500MHz 256MB GDDR3 $449

So the memory is clocked 125mhz lower, there's half as much, and the core is 50mhz lower.

For a grand total price discount of..$50?! Doesn't compute.

 

[digitalwanderer]

If Anand is right then the X1800 XL is the sweet spot right now.

 

[Arty]

It's the price for me:

X1800 XL 16 550MHz 625MHz 512MB GDDR3 $499
X1800 Pro 16 500MHz 500MHz 256MB GDDR3 $449

So the memory is clocked 125mhz lower, there's half as much, and the core is 50mhz lower.

For a grand total price discount of..$50?! Doesn't compute.

geo's referring to the difference between XT & XL.

 

[Tim Murray]

I wonder if ATI's margins are going to take a beating from the cost of 512MB of 700Mhz GDDR3...

 

[Hellbinder]

600mhz seems awefully low imo. Especially since *they* have been claiming for a while now that the R520 would give an increase over the GTX accross the board.

Of course that lower Fill rate may not be an issue since the R520 is TBDR.





















Just messin with ya'll

 

[digitalwanderer]

I think y'all are forgetting something; those 16 pipes aren't just regular pipes, they're extreme pipes....

 

[Rys]

Hellbinder is right, IMHO. 600MHz is a false-sense-of-NV-security clock, in this hack's opinion.

Although, given ALU configuration......

 

[scificube]

Am I lost...I thought the NV40 had higher IPC than R420 by virtue of the fact that NV40 performed so close to R420 despite R420 parts being clocked significantly higher.

Have I misunderstood things? If not, why did Anand say R420 has typically had better IPC considering R420?

This is not to say that they could not get better/higher IPC now but we have to see just how re-vamped R520 is as compared to all the other ancestors of R300 no?

...and how do you overcome a fillrate defecit with a higher mem clock...once you've hit the wall, you've hit it no? It is wrong to think you'll get closer to peak performance up until you wall but once you've slammed into it...you gotta stop.

I hope we learn more soon about how good those pipes are in R520...

 

[Geo]

. . .and he's referring to cognitive dissonance. I got it, anyway.

 

[Geo]

Although, given ALU configuration......

Wavelet!

 

[Jawed]

...and how do you overcome a fillrate defecit with a higher mem clock...once you've hit the wall, you've hit it no? It is wrong to think you'll get closer to peak performance up until you wall but once you've slammed into it...you gotta stop.

I suspect R5xx is designed to keep firmly pressed against the wall.

Also be careful to distinguish between fill-rate and shaded-fragment rate. The former is a function of the ROPs the latter a function of the fragment pipelines.

Jawed

 

[Tim]

Where my cognitive dissonance comes in is only 50mhz difference on the core between XT and XL. Not believing it yet.

Yes usually the top card is pushed the max speed where it is possible to suply a reasneble, it would be pretty strange if exactly 600Mhz was that limit - I think it safe to say that we will see a higher clocked version within a couple of months after the initial launch (assuming the 600Mhz XT is correct).

With the r420 launch Ati tried to create confusion about the final clock speed of the card, they might try doing the same this time so I don't think we can be 100% sure that the numbers from AT are final.

 

[scificube]

I suspect R5xx is designed to keep firmly pressed against the wall.

Also be careful to distinguish between fill-rate and shaded-fragment rate. The former is a function of the ROPs the latter a function of the fragment pipelines.

Jawed

Thanks.

 

[Maintank]

Well I have to say I'm impressed by the prices for the 512MB cards. I don't see how anyone can complain.

And relieved that this will finally put to rest all the 24 and 32 pipe nonsense.

So, the question remains - is 600MHz the top speed of R520 with the soft-ground problem, or with the soft-ground problem solved?

600MHz seems awfully low - although there have been hints before that 90nm was required for the process density more than speed.

Might ATI release an XTPE just in time for Christmas running at 700MHz?

Jawed

Judging by ATi the last couple of rounds. You will be lucky to see the 600Mhz variant in quantity by Christmas. 700Mhz would be a miracle.