To keep it simple gfx card perf breaks down like this in groups slowest to fastest (ie GF2MX400 is about equal in 3D speed to a GF4MX420): The GF4MX cards are 97% the same as GF2 cards, just sport better AA, image quality, dual display and TVout. The Rad cards offer better image quality, DVD playback, dual display and TVout than the GF2 and GF3 cards. The GF3 and GF4 cards have the best AA. Hope this helps to clear things up. Precise gains do depend upon your CPU etc, but here's a general idea using Athlon1ghz in 3Dmark2001 (card: total, games1-3 low, games1-3 high and game4 is DX8): The GF4MX420 should o/c very well though, apparently to GF4MX440 speeds. But then the GF2Pro should o/c to about GF2ultra speeds, so it would still be faster. With AA the GF4MX420 may be slightly faster than GF2Pro.You have to be careful saying the GF4MX420 gives similar performance as the GF2MX400 because the DDR version of the MX420 is much much faster than the SDR version which IS almost a GF2MX400.Originally posted by: AnAndAustin
GF2MX200 = GF2o/b
GF2MX400 = GF4MX420
GF2TI = GF2Pro = Rad7500 = GF4MX440 = Xabre400
GF4MX460 = Rad9000
GF3TI200 = Rad8500LE
GF3TI500 = Rad8500
GF4TI4200
3Dmark2001 (gfx card: total, games1-3 high detail FPS): nemesismk2, I don't have to be careful when I know what I'm talking about, respect. The LMA and AA does help the GF4MX cards, but the diff is minimal (5%ish), essentially the GF4MX420 is nothing more than a slightly enhanced GF2MX400, and the same for GF4MX440 vs GF2TI. I wasn't aware any GF4MX420 cards used DDR, if so I would expect it would be like the GF2MX incident where it was a matter of 64bit DDR vs 128bit SDR ... essentially no benefit at all. If you can find any info on GF4MX420 using 128bit DDR RAM it would be very informative. The nVidia standards are: Memory bandwidth is everything in GF2/GF4MX designs, even if a GF4MX420 used 128bit DDR it would still lose 20% bandwidth to the GF4MX440 which is only a tiny bit faster than the GF2Pro/TI. JFYI: Incidently the GF2Pro uses 200/400 (DDR) Memory Bandwidth 6.4gb/s and performs within about 2% of the GF2TI as the core makes little diff on GF2/GF4MX cards, again it is the Memory Bandwidth which hugely impacts perf.Originally posted by: AnAndAustin
GF4MX420: 4300, 34, 35, 40
GF2MX400: 4000, 32, 33, 36
GF4MX440: 5600, 40, 47, 47
GF2TI/Pro: 5400, 40, 42, 49
GF4MX460: 6000, 39, 51, 47
GF3TI200: 8100, 46, 66, 60
I'm going to have to agree with Nemesismk2 on this one, the GF4 MX420 is definitely superior to the GF2 MX400, and LMA is nothing to laugh at. In my experience most real world benchmarks put the GF4 MX420 performs barely reasonably similarly to the GF2 GTS-V which is essentially an underclocked GF2 GTS.
I think in this case your putting way to much faith in 3DMark as an accurate reproduction of real world gaming performance... though even still, in my own experience the GF4 MX420 usually beats the GF2 MX400 by a larger amount then that in 3DM2001 if not quite as much as the rather respectable difference in real world games.
On average, I've found the GF4 MX420 to be a good 20% faster then the GF2 MX400... which is no small performance difference.
In any case, the GF2 Pro is certainly faster then the GF4 MX420 in 99% of cases.
Gainward's GF4 MX420 uses DDR SDRAM. It's little more then marketing hype though, because as you mentioned it uses a 64bit memory bus which effectively limits it to the same theoretical bandwidth as SDR SDRAM at the same clockspeed. Plus, given that DDR SDRAM is not 100% efficient it actually tends to be a bit slower then 128bit SDR SDRAM.
After all guys I did say, "To keep it simple". I think what I stated was very accurate and concise and summed up the complex and confusing subject of where the GF4MX cards sit. I admit that the GF4MX cards are about 10% faster than the equivilent GF2 card, thanks to the enhancements like LMA-II (albeit cut-down) which should really help the GF4MX420 vs GF2MX400 due to the huge bandwidth problem associated with the SDR they use, efficiency is everything. However I believe the perf bands I gave were still pretty accurate, esp considering I was 'keeping it simple'.
Facebook
Twitter