Crysis built-in 32bit benchmark. All details set to HIGH except for anti-aliasing quality which was set to 4x. First run is ignored (warmup), the remaining 3 are averaged.

• 1600x1200 - min 10.94, avg 30.98, max 46.93

Comparison benchmarks:
Turion 64 at 2.4Ghz, GTX 275, 1x registered DDR 533 stick, CR1 ( Re: Battle of the platforms: socket 754! ), no gain

FEAR Perseus Mandate (2007), version 1.0. It falls into Windows Vista era (2007-2009).

Maximum details, 1600x1200, FSAA 4x, Anisotropic 16x, SS off. Enabled HW audio mixing and EAX 2.0 since we have Audigy 2 ZS.

Comparison benchmarks:
Turion 64 at 2.4Ghz, GTX 275, 1x registered DDR 533 stick, CR1 ( Re: Battle of the platforms: socket 754! ), average improved by 4 fps, minimum by 3 fps

Company of Heroes Complete Edition (2009) benchmark with Turion 64 at 2400Mhz and GTX 275:

4x AA was used, the rest is maximum quality. Some options allow Ultra quality that cannot be selected (Shader Quality, Terrain Detail). It can probably be enabled only with dx10. Environmental reverb cannot be enabled.

Minimum Frame Rate is probably a benchmark glitch. The value is not reliable.

Comparison benchmarks:
Turion 64 at 2.4Ghz, GTX 275, 1x registered DDR 533 stick, CR1 ( Re: Battle of the platforms: socket 754! ), we gain 1 fps, within margin of error

While I have results also for Far Cry, Prey, Serious Sam 2 and Splinter Cell Chaos Theory I'm not going to post them as they are the same as with GTX 275, within margin of error. Those games run very well on GTX 275 and there is no need to upgrade.

Conclusion for GTX 285 is that it provides benefits mostly in synthetic benchmarks, helps with maximums in some games and thus pulls averages up a little bit, there is little practical gain as we are CPU bottlenecked. This GPU is better suited for s939 X2 CPUs. Get it only if you find it very cheap, would like to max out the hardware, but do not expect real benefits.

Regarding converting registered ECC sticks to work in non-ECC boards, it seems this is possible only with SDRAM, it requires pulling down pin 147 (REGE) to turn the register into buffer if the pin is not connected on the board Converted registered SDRAM . There is PDF from Texas Instruments https://www.ti.com/lit/pdf/smoa001 (Logic Solutions for PC100 SDRAM Registered DIMMs) that explains how this REGE pin is used. Registers used in SDRAM had LE (latch enable) pin that is connected to 147 (REGE) and weakly pulled up by 10kOhm resistor so that it can be pulled down by pin 147. shevalier did such modification on one of his boards.

Unfortunately, in DDR1 the REGE pin was phased out

I did a quick benchmark of GTX 480 and the results are similar to GTX 285, except with some small regressions - in cinebench 2003 (OpenGL HW score) and lower minimums in Crysis (about 6-7fps instead of 10-11). The card excels in 3d mark 2003 with 28900 points in 1600x1200 and max details. I used the latest driver 368.81. It doesn't make any sense to use it in s754. I think it is a good fit for low clocked X2 CPUs in AM2.

I also put in a few GTX 260s Core 216 that I have and those usually come factory OCed. EVGA (GTX 260 Core 216) is clocked high and basically equivalent to stock GTX 275. Gainward Core 216 GS is somewhere between these two and stock, but has excellent heatsink with 2 fans so should have no problems running at EVGA clocks.

EVGA GTX 260 Core 216 SSC has clocks GPU 675 MHz, Shader 1453 MHz, Memory 1151 MHz - demonstrates how high you can clock these cards
MSI GTX 260 Core 216 has clocks GPU 655 MHz, Shader 1408 MHz, Memory 1050 MHz
Gainward GTX 260 Core 216 GS has clocks GPU 625 MHz, Shader 1348 MHz, Memory 1100 MHz

These can be considered basically equivalent to GTX 275 at EVGA clocks and can be used instead. MSI and Gainward should have no trouble reaching EVGA clocks as they both have 2 fan design and excellent heatsink. Do not buy GTX 260 that is not Core 216 (Shading units count) and not GT200B!

You can save money by buying GTX 260 Core 216 as GTX 275 tend to be more expensive.

GTX 285 is unnecessary in s754 and belongs to s939.

Use them with NVidia driver 197.77

https://www.3dfxzone.it/dir/tools/benchmark/video/ contains game benchmark tools, including games not tested so far:
- Enemy Territory Quake Wars (2007)
- Call of Duty 2 (2005) - it unfortunately crashes when it tries to compile results. Perhaps it works on older game version. Single core is not sufficient for the demo, dual core is badly needed. There are lot of Americans attacking German positions on mountain wall.
- Half Life 2 Episode One (2006) & Two (2007)

I added results of Enemy Territory Quake Wars benchmark to Re: Battle of the platforms: socket 754! (2.4ghz, gtx 275, DDR 533).

I can also report that Half-Life 2 Episode Two (2007) runs fine on 2.4ghz, gtx 275. With vsync on, only at the beginning, when looking around the forrest you get 50-60 fps, but once you get into smaller spaces it's 85 fps (vsync limit). This game has upgraded graphics compared to Episode One and original Half-Life 2.

Enemy Territory Quake Wars (2007) runs noticeably worse, the game seems to have 30 fps limit (was primarily an online game). After unlocking fps limit it seems to run at about 40-50 fps.