Review: MSI P6N-SLI Platinum
Last time we gave you an overview of the MSI P6N-SLI Platinum here. Today we will concentrate on overclocking and benchmarking it.
MSI P6N-SLI Platinum (provided by MSI)
nForce 650i SLI SPP/430 MCP
Intel Core 2 Duo E6700 (provided by Intel)
Scythe Andy Samurai Master (provided by Scythe-Europe)
Patriot 2GB PC2-6400U Kit PDC22G6400LLK (provided by Patriot)
CL5-5-5-15-CR2T at 1.80V
CL4-4-4-12-CR1T at 2.20V
AMD ATI Radeon X1950XTX (provided by AMD)
OCZ ProXStream 1000W (provided by OCZ)
Seagate Barracuda 7200.9 500GB SATA (provided by Seagate)
SilenX iXtrema Pro 14dB(A) (provided by PC-Cooling.at)
The Chipset is without a doubt a fast one and it can run 1T commmand rate on more expensive overclocking-friendly memory-modules, that should allow for some speed increase in memory hungry applications. The downside is, it gets extremly hot. Good case-cooling is required, because the cooling solution can barely cope with the heat. Increasing the northbridge voltage will crash your board without porper cooling in seconds. If you already have an SLI setup, this board may be a nice upgrade. Using only one PCIe x16 graphic-card, the second PCIe x16 slot is unusable.
MSI has choosen an AMI BIOS, which has some glitches. The CPU-fan will not slow down when you are using a standard 3-pin fan, other manufactures can do this. The overclocking options are inconsistent. You need to enter the marketing speak FSB, instead of 320 you have to enter 1280, use the delete key and enter the number, as using the +/- keys will take ages. FSB VTT Voltage can only be increased in percent-steps not in 0.05V steps. The "current" CPU frequency is always displayed wrong. The overclocking "warning" message is only annoying. When overclocking you should disable the "EIST" feature, but this setting was ignored completly. MSI has a lot of work ahead.
We received the 1.22 BIOS from MSI, which is dated March 15th, 2007 - but still not available on any MSI website. This BIOS should only have a fix for the CPU-multiplier, now lower settings are possible.
The results were disappointing. Regardsless what we tried, any attempt to set the FSB over 400MHz failed. 400MHz was the highest possible even with 1.50V on the northbridge. The Northbridge temperature quickly exceeds 60°C and the board crashes. You need to use the incredibly loud fan that MSI has supplied with the board. This is a 4cm Y.S.-Tech fan called "FD124010EB" running at an insane 7500rpm - welcome to the world of servers. 333MHz FSB was possible without a hitch, at 366MHz we needed to increase the northbridge voltage to 1.30V, 373MHz needed 1.35V for stability, booting was however possible with lower settings.
The board is limited to 1066MHz, so buying 1200MHz memory won't give you any benefit. For testing purposes we stuck to 800MHz, because we can run CR1T and higher settings will require 2T, so the benefit will be consumed by slower command rate and relaxed timings. Of course you will not always receive the frequencies you choose, because the ratios between FSB and memory are limited. Sadly you can't set the ratio, so you won't see the resulting frequency.
If you own a CPU with a high multiplier you are very lucky, because you need it to achieve high frequencies. We could boot up to 3750MHz, but it was impossible to run any tests. With 3733MHz we could run LameMT and 3DMark 06, but Gordian Knot was always crashing, so it seems our CPU doesn't like frequencies above 3.70GHz when the SSE units are used heavily, even with a dangerous Vcore of 1.65V. Setting higher voltages to the FSB and/or Southbridge didn't help either, so on all other settings they were left untouched. The Vcore drop on this board is very good resulting in 0.02V-0.03V up to 1.55V, higher Vcores could cause higher Vdrop.
We didn't bother with synthetic benchmarks, because they won't tell you anything interesting, we concentrate on real world applications, especially multimedia encoding.
Gordian Knot/XVID 1.1.2:
For our Gordian Knot testing we took a PAL epsiode of "Babylon 5" with a length of 41 minutes, 57 seconds and 8 frames.
We tried to "emulate" the most common usage of Gordian Knot:
1st: We have a perfect master, so we only de-interlace the content and resized it, without any other manipulations, we marked this as "fast".
2nd: You get bad mastering on many DVDs, especially "old" stuff or when the studios are in a hurry for the release. In this case you would want to improve the picture quality, which is done by filtering the content. You can choose from tons of filters for any purposes you can think of, we only used the most common "undot", "FluxSmooth" and "MSharpen". Of course we also de-interlaced. Filters were applied before any resizing took place (which is slower). We marked this as "slow".
If you need more info about filters, we recommend reading the doom9.org forum.
When using filters faster memory doesn't make a big difference, as the 0.5% speed increase is not worth mentioning. Without filters you'll get about 5-6% faster encoding. Of course faster memory has higher voltages and this comes with a penalty of 4-6W in power consumption per GB.
CL5 = CL5-5-5-15-CR2T, CL4 = CL4-4-4-12-CR1T
The same episode we encoded was used for our MP3-testing. We don't recommend using MP3 for encoding, because AC3 does a better job, but nearly 42 minutes gives us approximately the length of any given album.
A measurement in seconds, as many sites do, is useless, because the differences are too small. So we used the built-in play/CPU ratio, this means the CPU is encoding x-times faster then the track-length. Fast memory does not play an important role here.
For your convenience we also show you the single-threaded benchmarks, they will be produced with any other L.A.M.E. version, because only lameMT can do more than one thread and take advantage of a second core.
We used this setting: lamemt --vbr-new -q 2 -V 2 -m j --strictly-enforce-ISO --resample 48
Some of you want to know how much power is used when overclocking and/or gaming, so we did some testing. 366MHz FSB was achieved with 1.30V, 373MHz with 1.35V and 400MHz with 1.45V on the northbridge. The CPU ran at the default 1.3000V up to 3.00GHz, 3.20GHz required 1.4000V, 3.33GHZ 1.4250V, 3.66GHz 1.5000V and 3.73GHz 1.5250V Vcore. Faster memory give you a power penalty under load of 4-6W per GB and the speed increase is hardly noticeable.
Our testing should give you estimates, it depends on the power supply's efficiency and what other components you use in your system.
From a technical point of view the board is excellent. You get SLI - if you need it - a fast chipset, two IDE-ports and an excellent four phase VRM. The BIOS and the overclocking results are a mixed bag. Good overclocking results are only possible with high CPU multipliers, only the more expensive CPUs provide these.
The E4xxx series are limited in overclocking, E63x0 and E64x0 will not give you the multiplier to achieve the results all overclockers are looking for. The board has a target price of compareable Intel P965 boards, but with the limited FSB overclocking it isn't the choice for overclocking interested customers.
If you don't care about overclocking, the MSI P6N-SLI Platinum will be a stable, fast, and trouble free board for you. If you don't like the heatpipe or want to use a 3rd party cooling solution, you can buy the non-Platinum Edition, which is about €40 cheaper, but does not feature the "all-solid-capacitor" design. Instead it gives you an additional power-connector for the SLI-setup, the "Platinum" is missing this feature.
MSI should also supply quality cables and a fan that will do the job without producing so much noise. The board is good, but it can be even better, we wish FSB would overclock much better. This is our final conclusion, still 3.73GHz on air cooling is good.
The MSI P6N-SLI Platinum ships for around €135 in Europe.