This GPU is presented by PowerColor RADEON 9800 Platinum Edition, 128MB, AGP card.

Features:

• ATI RADEON 9800 GPU, 380MHz clock rate, 128-but GPU-memory bus (!);
• 128 MB in Samsung 4 BGA chips. 300 (600) MHz max. clock rate, 3.3 ns fetch. 256-bit bus;
• Peak fill rate in the multitexturing mode is 3000 megapixels/sec and 3000 megatexels/sec.

Theoretical and analytical videocard reviews, containing functional analysis of ATI graphics cards

By the end of the month the product costs $165-180 on average, supports DirectX 9.0.

AGP x8/x4/x2, 128MB DDR SDRAM in 4 Samsung chips. 3.3ns indicates 300 (600) MHz clock. 128-bit bus(!).

It has been rather long since ATI tried to make a good GPU out of this material. The matter is the company is still trying to make something else of the same thing and sell it as a new product.

RADEON 9800 Platinum Edition is a rather negative solution. Not because of bugs or something. But because they are trying to sell us a dubious hybrid under a nice name.

So, TUL (releases the well-known PowerColor) decided to make something original. That time ATI had lots of cheap R360 chips with a half-defective memory controller. So why not purchase a consignment of these alongside RADEON 9700 PCB. Besides, all R3xx are pin-compatible, so why not make something like Radeon 9500 Pro and serve it under a new name?

Such a cocktail. Though we have already seen similar solution on the example of RADEON 9800SE. But it had good controller (in spite of half-defective chip) 256-bit bus. And here we have everything vice versa: all 8 pipelines, 380 MHz clock rate, but only 128-bit bus.

Let's open the official specs of RADEON 9800: 256-bit DDR memory interface. Therefore, such bus is a compulsory condition for any RADEON 9800 product with any suffix.

They could name it RADEON 9770 for example, and it would have been OK more or less. But how could they release a RADEON 97xx late in 2004? Not good. And even the vendor would be happy to avoid R9800 name, but the Canadian "mother" ordered to name all R350-360 chips RADEON 9800 (plus suffix). Even if there's only 1 good pipeline, 1 texture module and 150 MHz clock rate - such a Voodoo3 or TNT2 with DX9 support. And still it would be RADEON 9800.

And TUL used such a nice suffix: Platinum Edition (ATI names its top platinum as well, e.g. X800 XT PE). But! This implies improving basic product design. So how do you think, if you take RADEON 9800 (even 9800 PRO) and halve its bus, is it an improvement? Or not?

Formally, TUL is acting correct - it's RADEON 9800 PE, not 9800 PRO PE. I.e. 325 MHz of 9800 was increased to 380 MHz. And as memory operates at just 300 (600) MHz instead of 340 (680) MHz required by Radeon 9800 Pro, it's somehow incorrect to add the PRO suffix. Besides, the experience of Sapphire is still fresh. As well as its sad result.

But what should we do with a pared-down bus? According to ATI specs above, RADEON 9800 requires 256-bit bus. More to this, when RADEON 9800SE was released, ATI informed that only 256-bit cards are the real ones. Though even then some ATI partners were releasing 128-bit 9800SE to get rid of old PCBs.

The only result we get is that RADEON 9800 Platinum Edition is an obvious trick and a lie! I believe such cards have a place to live - in the clothes world it's called Second-hand market.

Such cards might even be successful, but anyway they must be named according to their essense. If TUL considers halved bus a trifle comparing to increased clock rate, then it's deeply mistaken. You can see how our test results prove it. Only in a couple of benchmarks bandwidth is not important and 380 MHz core clock provides more than 325 MHz of 256-bit 9800. But we all remember that the last price for RADEON 9800 was about $160. I strongly doubt that TUL's new solution will cost $160 or even $150.

And what's the most important, such names can confuse unsuspecting users!

On November 10, 2004 the latest drivers from ATI were 6.490 (CATALYST 4.11 WHQL, Release), 6.497 (CATALYST beta) for Windows XP.