The EVGA nForce 650i Ultra motherboard gave me no problems throughout the review. All tests ran through without so much as a lockup although Super PI's 32M test refused to run without a Vcore increase to 1.3V. I also bumped the memory voltage up to 2.0V from 1.65V and the NB voltage to 1.3. The stock Intel cooler performed very well other than the noise the fan puts out as it amazed me in maintaining CPU idle temps at as low as 35C. During testing, especially with Super PI the CPU core temps were kept in the mid 50s with ambient room temps around 26C with humidity running less than 60%. That is much better than I expected from the stock cooler.
While on the subject of heat issues, I regularly took temperature readings of the NB heatsink base and the SB bare core. The NB passive heatsink would get rather warm getting into the lower 50C range range while the SB NF-430 would hit 57C running on 1.2V running 3DMark2005. I later solved these issues and was able to up the overclock substantially, although not a part of this review due the fact I do not have the time to verify the stability of the results, again refer you to the addendum for the low cost solutions at the end of the review.
The system booted into Windows with no problems with 1.3Vcore as confirmed by CPUID's
nifty little utility, CPU-Z v1.39.
Note: Remember, overclocking is not as simple as it sometime sounds and all the
results so far have been very good. My first foray with the nForce 650i chipset was a more long
drawn out experience and took several attempts and many hours to get this far. This
is my third session with this chipset and like motherboards please keep that in
mind as your overclock attempts may be different due to components and various other
I have included the processor arithmetic and multi-media tests along with the memory
bandwidth test. CPU and memory were run at the default settings.
Sisoft Sandra Performance
The E6600 processor with the EVGA nForce 650i Ultra actually edged out both my other
nForce 650i motherboards in the processor benchmarks at default clocks while the memory bandwidth
staying the same (although I was using different memory). In the overclocked setting, an increase in bandwidth of better than 10K or a 16% calculated increase in bandwidth resulted over the default clock settings.
In Super PI's 1M test the E6600 ran the test in 21.141 seconds at the default 2.4GHz clock setting, which splits the difference between the times I have gotten on the Asus P5N SLI and MSI P6N SLI Platinum motherboards. At the overclocked clock setting of 3.0GHz, the time dropped to 17.078 seconds. The 32M is a good stability test for overclocked settings. The
EVGA nForce 650i Ultra ran the 32M test easily at 3.0GHz in 17 minutes, 58:20.3
seconds or in about 18 minutes with no errors.
HDR and 3.0 Shaders are some of the additions to this benchmark suite. The default test resolution of 1280x1024 seemed to be just a little outside the
comfort zone for the GeForce 7600 GS so I purposely ran the resolution at 1024x768.
overclock gives the score a little bump, but not much.
The criteria for this review was to use the GeForce 7600 GS with the 650i Ultra as a potential
gaming platform. I chose the game F.E.A.R., along with a new increasingly popular title,
S.T.A.L.K.E.R. Both games will severely stress the GeForce 7600 GS to maintain sufficient
frame rates at 12x10 resolutions with moderate settings.
As a preliminary, the best performance
I could get in Aquamark 3 at 1024x768 with no AA/AF, overclocked or not, was 74
fps. Therefore, I will be looking for a minimum of 40 fps and preferably 60 fps in F.E.A.R. as
well as in the second game of the set, S.T.A.L.K.E.R. Most everything will be set
to maximum at resolutions of 1024x768 and 1280x960 (or 1280x1024).
I have not been
able to use the EVGA nForce 650i Ultra in any extended gaming sessions although
the motherboard itself seems to perform very similar to the nForce 650i boards that I currently
have running. I have briefly commented on both of those motherboards and believe
the potential for this motherboard is no different with consideration of the various
topics that have been covered in this review.
Three runs of the in-game benchmark were made and an average recorded at a resolution of 1024x768 with no AA. Any resolution beyond 1024x768 will require reduced in-game graphic settings.
I did not get the opportunity
to test it out in any gaming sessions during the course of this review but frame
rates need to average 60 fps to keep the game smooth throughout.
Three runs of the in-game benchmark were made and an average recorded at a resolution of 800x600 and 1024x768 with no AA and 4x AF. S.T.A.L.K.E.R. was playable at a resolution of 800x600 with in-game settings as indicated
below and this is about the limit for the GeForce 7600 GS unless the in-game settings are
The dips in frame rates seem to be infrequent and inherent to the game and did not cause any noticeable stuttering or pauses that I could detect as game play was smooth throughout the first map. I have recorded
similar minimum frame rates with my GeFOrce 7950 GX2 but also have not noticed problems
related to the sudden drop in frame rate.