I've got a quick question about 2 ATi Radeons, which is better an X300 128MB 375Mhz/200Mhz, or a Radeon 9550 256MB 425Mhz/200Mhz. I'm just wondering. That Radeon 9550 is amazing how much it overclocked, it went 175Mhz over and it still has more room, I'm just playing it safe :-). And another question, how much does overclocking the memory add to performance, is it better than a large clock speed? Lord_balron

I would suggest the X300. It has better support for Shaders and newer specs. I believe that the X300 can be overclocked a bit too. I would say that the X300 would be faster than an overclocked 9550 anyways. If not, then overclock the x300. :)

Thanks neonic, and does overclocking memory speed increase performance the same or better.

Lord_balron

I would think that OCing video memory wouldn't do too much, but I'm not sure. It may be that the overclocked core and stock RAM creates a bottleneck, but I'd do some more research to be sure.

While the X300 has shader support, it wouldn't be at all adequate for games that require SM3.0. I'd go for the 9550 unless you need the shader support. Even then, it would be beneficial to wait a bit and buy a better card that supports SM3.0, but it's just my opinion.

The X300 and 9550 are very similar chips. Both have 4 PS units and 2 VS units and are DX9 (SM2) chips.

The X300 only supports PCI-express and the 9550 AGP 4/8.

For which is faster check the clock speeds and bus width (higher is better). The memory capacity probably wont matter that much with these chips, 128MB will be more than enough.

As for the OCing, find out where your bottlenecks are. If the chip is bandwith limited then overclocking memory will help, if the chip is gpu limited (ALU calculations, etc) then overclocking the core will help.

They're pretty much the same card feature-wise, though there *may* be some small improvements in the silicon of the newer X300. For the most part, the Xn00 line was basically a PCIe refresh of the 9x00 line with slightly revised silicon and a temperature sensor.

The X300, however, allows vendors to produce models with only a 64bit data bus, while the 9550 is only possible with a 128bit bus. This will be the deciding factor. If the x300 has a 128bit bus to math the 9550 they will be roughly equal; if not, the 9550 will be better.

As for OCing memory, yes it does help. Graphics is a largely memory-bound operation, so increasing the transfer rate of data will provide a noticable boost. The Core/memory clock ratio is carefully tuned to provide a proper, if slightly conservative, balance.

I've found it best to, if possible, keep within a few percent of the original ratio. Given that you've overclocked the core from 250mhz to 425mhz - or 70%, you'd want to increase memory speed to 340-350Mhz to maintain the stock ratio. However, your core seems to have a lot of headroom so its likely that your memory won't be able to match this increase. The best way to OC your memory is to download the 3D mark 2003 (SM 2.0) trial, and begin overclocking the memory in 10mhz increments. After each step, run through a 3Dmark test, watching carefully for errors in the rendering. "Shimmering" is usually the first artifact you'll find, which will look like patches of flickering dark or bright pixels. Once you've found the point where shimmering begins, Back it off at least one step, or whatever you're comfortable with. Additional cooling or voltage adjustments may allow you to go further, but this leaves the realm of "simple" overclocking.

For fun, take note of your 3Dmark score before OCing, and your final results to see the benefit.

As always, OCing is never a guarantee and will void your warranty, so you do so at your own risk.