Radeon 9600 next week?

[mczak]

According to this link, http://www.aldi-essen.de/OFFER_D/OFFER_13_PC/OFF30.HTM you can buy a PC with a "Radeon 9600TX" next wednesday.
(more information here (german link), http://www.chip.de/news/c_news_10217455.html and here (german link), http://www.heise.de/newsticker/data/kav-22.03.03-000/ though both of these just are quotes of the above article as far as the facts are concerned.)

So are we going to see some reviews real soon now?
(Or was this 9600 just a typo in the ad?)

 

[MuFu]

MuFu.

 

[Xmas]

MuFu.

 

[UberLord]

 

[mboeller]

IMHO; Yes.. we should see the R9600 on monday. Even if we don't see an official press release c't has already benchmarked the R9600 against the NV31 and NV34. So on monday it's official.

- Radeon 9600: 325/200
- Radeon 9600 Pro: 375/300
- GF-FX 5600 Ultra: 350/350
- GF-FX 5200: 250/200
- GF-FX 5200 Ultra: 325/325

@ P4-2,26 + i845 under WinXP

NVidia: 42.74
ATI: Catalyst 3.2

no info about the settings :

Codecreatures: 1024 no AA, no AF / 1024 2xAA, no AF / 1280 2xAA, no AF:

R9600:              17 / 14 / 11
R9600 Pro:          22 / 19 / 14
GF-FX 5600 Ultra:   19 / 17 / 13
GF-FX 5200 Ultra:   16 / 14 / 10
GF-FX 5200:         10 /  7 /  4


UT2003 Flyby: 1024 no AA, no AF / 1024 2xAA, 8xAF / 1280 2xAA, 8xAF

R9600:               92 / 41 / 29
R9600 Pro:          124 / 55 / 38
GF-FX 5600 Ultra:   122 / 67 / 46
GF-FX 5200 Ultra:    82 / 48 / 32
GF-FX 5200:          49 / 26 / 16


UT2003 Botmatch: 1024 no AA, no AF / 1024 2xAA, 8xAF / 1280 2xAA, 8xAF

R9600:               50 / 32 / 22
R9600 Pro:           51 / 43 / 30
GF-FX 5600 Ultra:    50 / 40 / 28
GF-FX 5200 Ultra:    54 / 31 / 21
GF-FX 5200:          36 / 17 / 11

 

[mczak]

IMHO; Yes.. we should see the R9600 on monday. Even if we don't see an official press release c't has already benchmarked the R9600 against the NV31 and NV34. So on monday it's official.

- Radeon 9600: 325/200
- Radeon 9600 Pro: 375/300
- GF-FX 5600 Ultra: 350/350
- GF-FX 5200: 250/200
- GF-FX 5200 Ultra: 325/325

Finally an answer I understand without a translator
Looks like the 9600 will compete nicely with the 5200 Ultra, and the 9600 pro with the 5600 Ultra, just as expected. I'm a bit surprised though the 5600 (non-ultra) seems to get canned, these boards were also absent from cebit AFAIK. Performance probably was much too close to the 5200 Ultra, though I'd have expected the 5200 Ultra to become the card nobody wants instead of nvidia dropping the 5600 (similar to the MX460 which costs almost as much as the Ti4200).

btw why is the 5200 non-ultra so slow? Performance should be somewhere between 60% (memory bandwidth limited) to 75% (gpu limited) compared to the ultra, but it's only 50% in a lot of cases. I don't think even dead-slow ram timings could make things that much worse. But I should probably wait for some proper reviews to draw any conclusions...

 

[Pete]

So is the 5600U going to cost nV and boardmakers a lot more to produce than the 9600P? The core has more transistors, the memory is faster, and the card has more layers. If ATi can price the 9600P nicely below the 5600U, ppl may just overlook its lagging FSAA scores.

The UT2K3 scores are disappointing. Looks like the 9600P's memory bandwidth deficit is really hurting it in AA+AF scores, which is what I'm most interested in with a midrange card: 10x7 w/AA+AF.

Interesting scores, I just wish they'd included a 9500P. If memory serves, I think it did much better than the 9600P appears to be doing with AA &/ AF.

 

[OpenGL guy]

So is the 5600U going to cost nV and boardmakers a lot more to produce than the 9600P? The core has more transistors, the memory is faster, and the card has more layers. If ATi can price the 9600P nicely below the 5600U, ppl may just overlook its lagging FSAA scores.

The UT2K3 scores are disappointing. Looks like the 9600P's memory bandwidth deficit is really hurting it in AA+AF scores, which is what I'm most interested in with a midrange card: 10x7 w/AA+AF.

What you're probably seeing is the 5600 Ultra using the "balanced" or "quality" settings which gives it a large boost... at the expense of image quality.

Also note that the AA quality on the 9600 is far better.

 

[Joe DeFuria]

- Radeon 9600: 325/200
- Radeon 9600 Pro: 375/300
- GF-FX 5600 Ultra: 350/350
- GF-FX 5200: 250/200
- GF-FX 5200 Ultra: 325/325

It's toungh to judge based on the inital results so far, particularly because they only utilize 2X FSAA, which tends to be the sweet spot for the nVidia cards (vs. 4X FSAA which tends to be the sweet spot for ATI products.)

However, in terms of performance, it looks like the 9600 Pro is going to be directly competive with the 5600 Ultra, and the 9600 Non-Pro a shade higher in performance than the 5200 Ultra.

So for OEMs, two things are likely to sway them when choosing a product:

1) cost
2) Availability of parts

I'd wager that ATI has a distinct cost advantage here, especially with the 9600 non-pro vs. the 5200 Ultra.

Not going higher than 200 Mhz DDR is going to be a big cost differentiator, given the impending push by intel to standardize on such memory with their up-coming chipsets. You can bet memory manufacturers are ramping up 200 Mhz DDR production, and 200 Mhz DDR is going to be very cheap compared to higher speed grades.

On the other hand the 0.15u 5200 ASIC is likely to be available with a larger supply than the 0.13u 9600 ASIC. It gets tricky though trying to predict the availability of the specific speed grade. It's not at all clear if the "Ultra" speed grade of the 5200 chip would be more or less plentiful than the non-ultra variant on the 9600.

 

[mr]

The german article also stated that a 5600 was not available for now. They were also not sure if the retail 9600 Pro cards will make the announced 400MHz clockspeed as their sample had 375MHz.
The article further states that the 9600 with an announced price of €190 and the 9600 Pro with €220 are both €50 cheaper than the 5600 and 5600 Ultra they are competing with.

 

[Arun]

What you're probably seeing is the 5600 Ultra using the "balanced" or "quality" settings which gives it a large boost... at the expense of image quality.

Wasn't it determined that "balanced" on the NV3x is roughly equal ( sometimes better, sometimes worse ) to "performance" on the Rx3xx? Of course, if they compared it to to "quality" on the Rx3xx, then it isn't fair.

BTW, anyone noticed you got to put three "x" to include all of the R300 family GPUs? Pff, those ATI guys!

[/quote]Also note that the AA quality on the 9600 is far better.[/quote]

Agreed. But those specific benchmarks use 2x AA - and ATI's 2x AA pattern is only *slightly* better than nVidia's 2x AA pattern. While it is good to point it out, I don't think that should influence anyone purchase decision.

Of course, 4x AA is a completely different matter, and ATI got a very signifiant lead in 4x AA IQ. This should definitively influence the buying decision of people playing at lower resolutions, since those people would benefit the most of such advantage on ATI cards ( it shouldn't influence it if you can play at 1600x1200, because the RV350 doesn't have the muscle to deliver 1600x1200 4x AA in most situations )


Uttar

 

[Dave Baumann]

The sample pattern is not all that determines AA IQ.

 

[Arun]

The sample pattern is not all that determines AA IQ.

Stupid me... Forgot about Gamma Correction. I'm not forgetting anything else now, am I?


Uttar

 

[demalion]

What you're probably seeing is the 5600 Ultra using the "balanced" or "quality" settings which gives it a large boost... at the expense of image quality.

Wasn't it determined that "balanced" on the NV3x is roughly equal ( sometimes better, sometimes worse ) to "performance" on the Rx3xx?

Well, perhaps by reviewers who actually bother to evaluate things further than the label used for settings.

If I understand what I've been reading correctly, nvidia has changed the name for "Balanced" to "Quality", and the name for "Aggressive" to "Performance". This allows a bait and switch so that when actual image quality is compared, "Application" is available to be used while the impact on performance of the other modes is perceived to be concurrently representative.

 

[OpenGL guy]

If I understand what I've been reading correctly, nvidia has changed the name for "Balanced" to "Quality", and the name for "Aggressive" to "Performance". This allows a bait and switch so that when actual image quality is compared, "Application" is available to be used while the impact on performance of the other modes is perceived to be concurrently representative.

And it just so happens that ATi's two anisotropic modes are called "quality" and "performance" with the only difference being whether trilinear filtering is enabled...

Nah, there's no conspiracy here...

 

[Pete]

I know ATi's AA is better, and I'm fairly confident their AF is better in that it's cheaper, but I haven't seen ppl test ATi's AF in flight sims, to see if the varying quality is noticable. I'm not sure why reviewers haven't thought of this, as it's been done before, IIRC (can't remember the name of the site now, tho--of course).

I'll admit nV's renaming their AF settings seems a bit sneaky. Should ATi rename their AA modes to 2x+, 4x+, and 6x+?

 

[OpenGL guy]

I'll admit nV's renaming their AF settings seems a bit sneaky. Should ATi rename their AA modes to 2x+, 4x+, and 6x+?

No, I'd go a step further: Call 2x "4x", 4x "8x" and 6x "12x". I mean, our 4x looks better than nvidia's 8x(S), afterall

P.S. Just joking BTW.
P.P.S. But not joking about the quality differences

 

[Dave Baumann]

If I understand what I've been reading correctly, nvidia has changed the name for "Balanced" to "Quality", and the name for "Aggressive" to "Performance". This allows a bait and switch so that when actual image quality is compared, "Application" is available to be used while the impact on performance of the other modes is perceived to be concurrently representative.

Where have you read this? What drivers is this? I've not seen this at all yet.

 

[Entropy]

I'll admit nV's renaming their AF settings seems a bit sneaky. Should ATi rename their AA modes to 2x+, 4x+, and 6x+?

No, I'd go a step further: Call 2x "4x", 4x "8x" and 6x "12x".

Sorry, SIS already did that. Your marketing has to do better. Particularly on creative naming of pedestrian features, your marketing guys are WAY behind them there!

Entropy

 

[demalion]

If I understand what I've been reading correctly, nvidia has changed the name for "Balanced" to "Quality", and the name for "Aggressive" to "Performance". This allows a bait and switch so that when actual image quality is compared, "Application" is available to be used while the impact on performance of the other modes is perceived to be concurrently representative.

Where have you read this? What drivers is this? I've not seen this at all yet.

I first noticed an odd reference here concerning "Aggressive" aniso, and mentioned it here (go back a page and you'll see them comparing GF "Balanced" to ATI "Quality").

Later, Xmas seemed to indicate that it was part of a new naming scheme (I'm not sure which specifice driver version(s) in particular he was mentioning), and not just a mysterious labelling choice at THG. Xmas was mentioning that "Balanced" was renamed to "Quality".

I'm still awaiting further confirmation that my understanding of these indications is correct.