Review: nVidia still can't do good chipsets.
Nvidia is famous for it's renaming-game on their graphics-chips and they have picked the same strategy for the new 750i. The new chipset is basically an old Nforce 650i is paired with a NF200 PCIe 2.0 x16 chip and branded as 750i. eVGA is now trying to get the best out of it. It is little known that Evga 750i motherboard was designed by the ex-Epox motherboard design team and if you remember they used to make some world class motherboards.
You know our own English tends to turn out bad (really? sub.ed.), but we still can't imagine what "Engineered For The Win" should mean.
Analog Devices ADP3189 VRM driver
ti TSB43AB22A 1394a Firewire controller
Realtek 8211B PCIe network PHY
Winbond W83627DHG Super I/O controller
passive cooling of chipsets/VRM
8Mb BIOS, version 1.3
2x PCIe 2.0 x16
1x PCIe x2
2x Dual-Channel DDR2-slots for PC2-6400U memory up to 8GB
4-Port SATA II featuring RAID 0, 1
1x PS/2 keyboard
1x PS/2 mouse
1x Gb LAN
6x USB 2.0
7.1 analog output
1x SP/DIF digital out
1x eVGA Fan
1x 4-port USB 2.0 bracket
4x SATA cable with clips
2x HDD-power to 2x SATA power cable
1x Floppy cable
1x IDE cable
Bus Speeds: Auto, 400MHz to 2500MHz in 1MHz increments
Memory Ratios: Auto, linked, unlinked, fixed ratios: 1:1, 3:2, 5:4, sync
FSB Strap: none
DRAM command rate: Auto, 1T, 2T
DRAM timing control: Auto, manual
PCIe Frequency: Auto, 100 to 200 MHz in 1MHz increments
CPU Clock Multiplier: Auto, manual
Core 2 Duo: 6x-11x in 0.5x increments - Core-2 Duo, downwards unlocked
Core 2 Extreme: 6x-16X in 0.5x increments, downwards unlocked
CPU Voltage: auto, 0.5000V to 1.8000V in 0.0125V increments
DRAM Voltage: auto, 1.80V to 2.50V in 0.025V increments
FSB Termination: auto, 1.10V to 1.55 in 0.05V increments
Northbridge Voltage: auto, 1.30V to 1.55V in 0.05V increments
NF200: 1.20V to 1.55V in 0.05V increments
eVGA 750i SLI FTW (provided by eVGA)
Intel Core 2 Duo E8400
Scythe Andy Samurai Master (provided by Scythe-Europe)
Kingston 2GB Kit PC2-9600U KHX9600D2K2/2G (provided by Kingston)
CL5-5-5-20 CR2T @1.85V
Jetway Radeon HD3870 (provided by mec-electronics)
Seasonic S12-II 500W
Western Digital WD4000KD (provided by Ditech)
SilenX iXtrema Pro 14dB(A) (provided by PC-Cooling.at)
Cooler Master Stacker 831 Lite (provided by Cooler Master)
Let's talk about the good stuff first, the layout is the best we have seen in quite a while. Of course, it's not perfect either, because the Port 80 diagnostic LED is placed too much toward the middle; the same goes for the USB header, but otherwise you get plenty of space.
The VRM is an analog dual 3-phase design driven by an ADP3189. Sadly, eVGA did not opt for a digital VRM solution. While testing we ran into some troubles, when overclocking the VRM did sometimes exceed Intel's specifications and dropped over 0.05V. Most of the time it barely stayed below the magic 0.05V mark.
The memory slots are far away from the primary PCIe x16 slot and changing memory modules in an installed environment is not a problem at all. Below you find the IDE port and the 24pin power connector.
The slots are configured for a dual SLI configuration. When using two dual-slot graphics cards you will have at least one PCI slot free. The downside is it will block the Port 80 diagnostic display. On the left edge you also find LEDs for power and HDD which doubles also for power and reset switches. Near the EEPROM chip a small red button resets the CMOS.
Network connectivity is provided by the chipset, the Realtek Gb 8211B chip provides the connection to the Gb LAN port. A Texas Instruments TSB43AB22A chip provides two Firewire-ports, one on the back panel, one header onboard. As usual, a Realtek chip covers onboard sound.
The SATA connectors are, as usual, on the bottom edge of the board. The funny thing is, they are split in two ports each with the floppy connector in-between. All connectors are angled so you can connect nicely even with full-size graphics cards installed.
The printing of the package states "10 USB 2.0" ports, and our test sample came with only one header. We hope that eVGA ships two brackets with the retail board.
If you like to break the 500MHz FSB mark, you need something to cool the Northbridge cooler. When you look the cooling-solution you know in an instant that it just can't work, because only one heatpipe going into the MOSFETs cooler of the VRM, can't get rid off the heat produced by the Northbridge. So eVGA was wise and included a fan. Unfortunately, the fan is roaring at 5400 RPM, which is incredibly noisy. Nvidia did this to its 680i board as well, and unfortunately eVGA didn't change the reference fan.
If you listen to the fan unattached, it's not that bad, but when attached the fins create a lot of turbulence, increasing noise exponentially. Next time eVGA should not opt for the nVidia cooling solution, but rather develop something different themselves. Even eVGA should have noticed the trend is going to "silent" solutions, even for gamers, because they want to keep their love-life intact.
eVGA uses a Phoenix-Award BIOS, with lots of options. While we were not happy with the intervals of voltage-settings, they get the job done. If you have finer steppings you can better try to lower power consumption because you don't need that much over-voltaging. As always, there are some silly settings, such as COM1 enabled - the package doesn't contain a COM header and some other.
The range of the options are ok. When overclocking and leaving settings @ auto, the BIOS does over-voltage on its own. That is a very bad behavior. "Auto" settings should do only the voltages required for normal operation. We strongly suggest you set everything yourself.
eVGA intended this board as a cheaper replacement for a 790i board, so they invested some engineering for the FSB. We got it up to 525MHz stable; this is the highest setting with an nVidia chipset we ever did.
Our CPU can do 4400MHz, and the board can boot up with such settings, but it was impossible to run it stable. Due the high voltage-drop the CPU need more voltage which increases temperature.
On the package the board says it will do 1066MHz. We tested this with 1066MHz exactly, and even with high FSB-overclocking, it works.
FSB BIOS clocking:
We have to state that we always recalculate the bench results to nominal frequencies. Most vendors slightly overclock their products, maybe to get more bench points, but we nullify such attempts. eVGA keeps the clock close to the nominal values:
Please note that different bios revisions may give different results. All benches are done with AUTO settings without altering any BIOS option besides CPU VCore, NB VCore and FSB Termination Voltage.
x264 is a h.264/AVC codec which supports four threads, and it's available for free. We took a PAL episode of "Babylon 5" with a length of 41 minutes, 57 seconds and 8 frames. We tried to "emulate" the most common usage when you encode your movies:
1st: We have a perfect master, so we only de-interlace the content and resize 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 want to improve the picture quality, which is done by filtering the content. You can choose from lots of filters for any purposes you can think of, but we only used the most common "undot," "FluxSmooth" and "MSharpen." Of course, we also de-interlaced, filters were done before any resizing took place (which is slower). We marked this as "slow."
*Note: The DFI P35 mainboard doesn't support half multipliers, so it was using 417x8.
The same episode we encoded, we used for our MP3-testing. We don't recommend using MP3 for encoding, because AC3 can do the job better, 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 the single-threaded benchmarks figures, they can be re-produced with any version of L.A.M.E. Only LameMT can do multi-thread and take advantage of multi-core processors.
We used this setting: lamemt --vbr-new -q 2 -V 2 -m j --strictly-enforce-ISO --resample 48
*Note: The DFI P35 mainboard doesn't support half multipliers, so it was using 417x8.
eVGA did quite a good job. While nVidia chipsets are known to be slower compared to Intel, this time the gap is quite small, a nice achievement.
As expected, the results are bad. While the MSI X48 and the ASUS Extreme are using an IDT PCIe switch chip, which does eat about 10W, the nVidia chipset can top this easily. Due to the quite high voltage drop with high voltages, the gap widens the higher the overclock is.
Please note that other configurations may yield other results.
The price for the board now is about €170,- in Europe, so we get ripped off, as usual, compared to the $199,- in US. For this money you get a nice layout, but you have to pay extra bucks for the high power consumption. The betting champ is its SLI support and if you don't need it, Intel-based boards will do you good.
nVidia failed to improve its chipsets, they just recycled an old chipset, paired it with a new companion chip and gave it a new name. There isn't much innovation. It would have been much wiser to cash 10 bucks from Intel per board and give them the SLI license and make the dual card thing more popular.
All that is not eVGAs fault. They really tried hard, and ex-Epox team has designed and really improved the reference design of the Nforce 750i motherboard.
The first testing tells us that eVGA 750i is actually faster than some tier one Nforce 750i, but we are still testing and we are not ready to post the final scores. We feel that the VRM needs more engineering to give stable operation while overclocking with high voltages but again we scored more than FSB500.
Besides that, they did a nice job. The main problem is the price, as P35 will sell for about €100 in a good motherboard and if you don't need SLI that might be a wiser choice; and the Gigabyte EP35C-DS3R that we reviewed two months ago is a good example. It can overclock higher, uses less power and features twice the amount of SATA-connectors.
On the other hand, eVGA beats many Nforce 780i boards around and since we know that almost no one uses Tri-SLI, eVGA at significantly lower price is probably a much better choice.