GeForce2 MX 200

[adntaylor]

Has anybody seen any reviews of these yet?

I want to see just how lame they _really_ are, as we'll be probably seeing them in most "high-end" retail PCs soon 😉

With half the bandwidth of an already bandwidth-limited card, will this be slower than a TNT2?

 

[vss1980]

Well, they have benchmarks showing it as slightly faster than the TNT2, but I think, when its used in games that have lots of large textures and it cant take advantage of any of the Geforce2 core's abilities it will be slower.

All in all, it is probably on a par with the TNT2, but I think a TNT2 Ultra would have the edge over it. But due to the now lacking TNT2 feature set, this will probably fill any gap the TNT2 filled.

 

[dowxp]

www.tomshardware.com

just came out with 200 and 400

 

[Rand]

I generally think of it akin to the TNT 2 M64 as they both have the same memory bandwidth, just the MX 200 has a significantly more powerful core.
I'd be guessing it'll perform somewhere in the level of the generic TNT2 is most cases... maybe somewhat higher. TNT2 Ultra at stock clock should be able to consistently outperform it though.

 

[LeoE]

"I want to see just how lame they _really_ are, as we'll be probably seeing them in most "high-end" retail PCs soon"

50% of the bandwidth of "standard" GeForce2 MX and GeForce SDR. At around 1GB/sec, it barely has enough bandwidth to display (not render) a 1600x1200x32 screen at the max 100hz. At which point, rendering performance goes towards zero, since the RAMDAC would hog the memory 😎

(assuming they use sub-par 143 or 125mhz vid memory, which they most likely will!)

 

[Rand]

Bangsailio, Tom only benched the MX 400 not the MX 200.
LeoE, the MX 200 has a 166MHz SDR 64bit memory bus which equates to 1.3GB/sec max theoretical memory bandwidth, not the "under 1GB" you mentioned.

 

[LeoE]

Rand, I've just realised the mistake & added the disclaimer.

With 166mhz memory, they'd get (500/3)E6 * (64/8) / 1024^3 = 1.24 GB/sec theoretical max;

With 143mhz, proportionally 1.07GB/sec (very likely they'll use 7ns memory)

With 125mhz (not impossible), they'd get 0.93 GB/sec 🙂

Now, multiply by (8/9) to account for average SDR CAS latency. We get:
166 mhz: 1.1GB/sec
143 mhz: .95GB/sec
125 mhz: .83GB/sec

Subtract what the RAMDAC consumes, and you have the actual leftover rendering bandwidth... woot 😉

 

[Rand]

LeoE, another slight correction, though some manufacturers may likely use slower 6.5 or 7ns rated SDRAM on MX200 cards they will almost assuredly clock it at 166MHz as per NVidia's specs.

And to correct your calculation of bandwidth...If we have a videocard utilizing SDR SDRAM, operating at 166 MHz, on a 64bit memory bus: (NVidia's specs for the MX200 which almost all manufacturers follow strictly)
64-bit (8 bytes) x 166MHz = 8 x 166 = 1328 GB/s of available theoretical bandwidth from the memory.

 

[dowxp]

thanks for catching my error rand.

 

[LeoE]

"LeoE, another slight correction, though some manufacturers may likely use slower 6.5 or 7ns rated SDRAM on MX200 cards they will almost assuredly clock it at 166MHz as per NVidia's specs."

I doubt it, because my Creative GeForce2 MX DDR has 7ns memory @143mhz, not 6ns @166mhz. The same goes for some other existing cards. And what better product to cut corners than MX200 😉

"And to correct your calculation of bandwidth...If we have a videocard utilizing SDR SDRAM, operating at 166 MHz, on a 64bit memory bus: (NVidia's specs for the MX200 which almost all manufacturers follow strictly)
64-bit (8 bytes) x 166MHz = 8 x 166 = 1328 GB/s of available theoretical bandwidth from the memory"

That's the same calculation I did, except I used greater precision on these counts:
* I represented 166mhz as a fraction (1000ms/6ns) = 166.666...
* I used the standard definition of Gigabyte as 1024*1024*1024 bytes, not 10^9 bytes which produces exaggerated GB numbers