I noted recently how the tide of products we've looked at here at Hexus had become distinctly Intel dominated with boards and processor keeping us busy. Well it seems the tide is changing just now with our recent look at the Abit NV7m motherboard, a micro-ATX solution based on the NVIDIA nForce chipset and supporting AMD Athlon and Duron processors and here today with a look at Shuttle's updated AK31 product, the revision 3.1 AK31.
Previous versions of the AK31 have featured the VIA KT266 chipset, one we looked at in the Soyo Dragon and found to be competent but slightly underwhelming. Thankfully VIA released a new version of the chipset, the KT266A, with an updated northbridge featuring a new and improved memory controller. This addressed the deficiencies in the earlier release of the chipset by deepening the memory controller to CPU buffers present in the northbridge allowing the CPU to communicate with the memory controller more efficiently, effectively increasing bandwidth.
It's the updated version of the chipset, the KT266A, that features in the revision 3.1 AK31 and that's what we are looking at today.
Without going into huge detail about the KT266A, I will offer up a quick diagram and overview of the chipset to refresh your memory.
As you can see, the chipset is comprised of the usual north and south bridges (VT8366A and VT8233) which give us out component interfaces and peripheral interconnects needed to build our motherboard.
Both chips are V-MAP designs meaning they are pin compatible with older and future V-MAP bridges. This means that Shuttle could effectively take the new KT333 northbridge, a V-MAP chip itself, and use it with the already effective AK31 PCB design and have to do very little else to make the new board work. The bridge is a pin compatible drop in design which is what defines V-MAP. This is what makes it so easy for manufacturers to bring new VIA based products to market since the PCB designs can remain identical for the most part with little tweaks here and there.
Manufacturers will use V-MAP in this way to bring KT333 based boards to market in a short time.
The southbridge provides us with our 32-bit, 33MHz PCI interface (6-slots on this board), the USB interface and the IDE interface. The northbridge as you can see is responsible for hosting the memory controller interface for the processor and the AGP interface for your graphics card. It is also in charge of the EV6 bus connect to the CPU itself. The bridges are joined via a 266MB/sec V-Link bus connect (to be upgraded to 533MB/sec in future products).
So lets take a look at the official spec of Shuttle's implementation of the chipset.
Supports AMD K7 Socket A processors
As you can see the board isn't bristling with added features but is mearly a solid implementation of the chipset. While you wont find USB2.0, RAID or a decicated audio processor, you do get a decent layout, good expansion, 4 DIMM slots and support of all of AMD's current Socket A processors.
(North bridge) VIA VT8366A
(South bridge) VIA VT8233
Supports AMD Athlon XP CPU up to 2000+Mhz
Supports AMD Duron CPU 950Mhz+
Supports AMD Morgan CPU 1.1GHz
Size: 305mm x 245mm
1 x AGP (4x AGP)
6 x PCI
1 x CNR
4 184-pin DDR DIMM Slots
On board IDE Controller
Supoort Ultra ATA 33/66/100
Transfer rate up to 100 MBytes/sec
Onboard VIA AC97 Sound Card
With nothing much to comment about feature wise, we'll jump straight to the layout and installation.
Layout and Installation
We'll start as we always do at the top left and work our way across and down.
The first thing we hit as we take our tour around the board is the ATX connector placement, vertically aligned in the top left corner. This isn't an ideal placement in many cases since it means dragging the ATX power connector across the CPU area, potentially impeding airflow. It owuld have been nice to see it on the other side of the board if possible but the PCB layout of the AK31 prevents this and at least the connector isn't in center of the board as we've seen on previous boards we shant mention!
Moving right from the ATX connector we hit the usual bank of power regulating components, mainly capacitors, that flank the CPU socket. They are far enough from the socket to no cause concern when installing large heatsinks but there are other issues that will as we'll find out soon.
The socket itself is rotated 90 degrees to the right. We've seen this before on the Soyo Dragon and it's a pet hate of mine for the socket to be rotated since it can be fiddly to install a clip based heatsink in your case when the socket is rotated like we see here. Vapochill owners will cry when they see the socket rotation since it means an extra twist on an already limited evaporator hose. From my experience with water cooling it hinders using the board if your hose alignment is calculated for the usual north-south rotation.
Finally with regards to the CPU socket, one of the pins that hold the northbridge fan to the board is too close to the socket. I was not able to fit a Swiftech MCX462 to the board because the pin fouled the heatsink base so I couldn't make good contact with the CPU (actually very little contact at all). I had to resort to a less than ideal Globalwin FOP38 heatsink which was my only backup Socket A sink at the time.
4 DDR DIMM slots flank the CPU on the right hand edge of the socket and it's a nice touch to see 4 slots on an AMD board with the norm being 3.
The two IDE connectors and the floppy connector on the board are tightly grouped along the right hand edge of the board in the lower half, below the line of the AGP slot. They could be higher up ideally but the placement isn't terrible.
As is usual on AMD boards, from the AGP slot (featuring a retention mechanism on this board) downwards, the layout is very clinical with nothing out of place. The ATX case headers are in their usual spot in the bottom right and the CD audio headers are in a fine placement on the left of the 6 PCI slots.
Physical installation was fine since the board isn't oversized. It uses the full 9-hole ATX mount but in the test chassis that wasn't an issue. It's not a small board by any means and owners of smaller chassis should check to see if it will fit, but for 99% of people, there wont be a problem. The old board was removed and the AK31 slotted right in to replace it.
Bundle and Presentation
The bundle was stead and unexciting with just the essentials for getting up and running. Items include the obligatory IDE and floppy connectors, easy to read manual, driver CD and a pair of backplane mounted USB 1.1 ports.
The box itself is an interesting silver design but nothing incredibly exciting. Overall, the presentation and bundle are nice but there's nothing to catch the eye and no extras to make you feel you are getting your moneys worth. However, at the end of the day, it's just a motherboard and how exciting can a bundle of electronics be?
Before we delve into the performance of the board, a quick rundown of the test system as always.
While I had the luxury of running an unlocked processor in the board, the Globalwin FOP38 didn't let me overclock the CPU as much as I'd like. With the use of the Swiftech, had the board permitted, or indeed a better clip style sink I would have been able to run things a bit faster. 1610MHz was the maximum stable overclock I managed to achieve from the processor using the FOP38 (I hate Delta fans!) and was the speed I was able to run all the benchmarks at without incident.
- Shuttle AK31 r3.1 Socket 462 AMD DDR Motherboard
- Unlocked AMD Athlon XP1500+ Processor (1.33GHz, 10 x 133)
- 1 x 256Mb Samsung PC2700 DDR Memory Module (CAS2)
- Gainward Ti550 GeForce3 Ti500 64Mb
- Adaptec 39160 PCI SCSI Dual Channel U160 controller
- 2 x 73Gb Seagate Cheetah U160 10,000rpm SCSI disks
- Plextor 12/10/32S SCSI CDRW
- Pioneer 6x Slot-load SCSI DVD
- Creative Labs Soundblaster Audigy Player (SB0090 version)
- Windows XP Professional Build 2600.xpclient.010817-1148
- DetonatorXP 22.40 NVIDIA drivers
- Aquamark v2.3
- Quake3 v1.30
- POVRay v3.1g.msvc.unofficial-win32 dated 28 August 2001
- 3DMark 2001 Professional
- SiSoftware Sandra v2002.2.8.64
- Serious Sam: The Second Encounter Dema
The speed was achieved by simple bus overclocking with the multiplier at the stock 10x and the front side bus increased to 161MHz. The memory clock when overclocked was also 161MHz and the memory was run at the the default 2.5V at this speed with the following timings chosen in the BIOS: CAS2, 4 Bank interleave, 2-5-2 Trp/Tras/Trcd, 4 level DRAM Queue depth and 1T Command rate.
These agressive memory timings weren't a problem for either board or memory module and the module seems to run very well with the KT266A chipset memory controller.
The BIOS allows you to push the CPU Vcore higher than most boards, all the way to +0.275V giving a maximum of 1.975V for the CPU on the AK31. Memory voltage is adjustable to 2.7V which is a little low compared to some boards but the CPU Vcore adjust makes up for it in the eyes of the overclocker. Interestingly, the test benchmark results were done at 1.75V (stock) due to the excellent processor.
You only have a choice of up to 166MHz front side bus so I was close to maxing out the front side bus on the board and I can confirm it does 166MHz FSB easily on this CPU when run with a lower multiplier.
So lets start with the Sandra memory benchmark. I'll dispense with the CPU benchmarks for this review since they don't show anything of interest other than the CPU working properly at both the benchmark speeds.
The KT266A is very efficient in term of the results you get from Sandra where it reports the chipset as generally being 95% efficient. Averaging 2.4GB/sec is the scaling you'd expect from the front side bus increase, possibly slightly low. We'll see later that the overall system increase from a combination of memory bandwidth, CPU clock speed and PCI/AGP speeds all help to increase scores across the board.
We'll use POVRay as out sole CPU benchmark and only at 1333MHz (stock) just to verify correct operation. We run the P3 binary and render the pawns.pov scene at 640x480 using our usual command line.
54 seconds is within the 1 second margin of error we expect to see from the processor at 1333MHz and for the record the KR7A-RAID managed it in 53 seconds. Effectively equal, especially since one of the POVRay runs was done in 53 seconds. Remember we take a 3 run average, discarding the highest and lowest result, just as we do in all tests.
Quake3 next and with our four.dm_66 demo (shipped with the 1.30 Point Release) it's a decent test of overall system throughput so we'll expect a good increase in scores when overclocked.
Here we see, especially at the lower resolution where overall system throughput has more of an effect, that the scores are increased handily when overclocked as we'd expect. The graphics card is untouched throughout and runs at a constant 240MHz engine clock and 500MHz DDR memory clock. The increased AGP clock speed increases the bandwidth available to the AGP port as the front side bus speed is scaled up. There is a case to me made for a fixed PCI/AGP timing when overclocking but provided your system is comfortable with the increased bus clocks, it makes for a handy increase.
The scores are actually a shade faster than the KR7A-RAID out of the box so Abit's performance flagship takes a beating in Quake 3 for the second time in recent days with the little NV7m claiming the scalp of its big brother recently for us.
Next up we have Aquamark which doesn't respond as much to system clocking like it does to an increase in graphics horsepower. Aquamark is incredibly shader heavy so quick DX8 class cards perform well here. 60fps is the target in this benchmark. Anything over that represents a comfortable all round gaming system.
We get our 8fps increase over the stock clocked results to take us over the 60 frames per second goal and again the AK31 beats the KR7A-RAID out of the box making it the fastest KT266A board we've yet seem (although the margins are very very slim). The increase in system throughput would be well paired with an overclock of the graphics card here.
With the benchmark being shader heavy, the graphics hardware plays more of a part in big scores than the rest of the system but 52 fromes per second out of the box on a 1333MHz processor is quite something. The KT266A is a good platform when paired with a DX8 class accelerator.
Lastly for our look at performance benchmarks we'll look at everyones favourite, 3DMark 2001 SE Professional. 3DMark's variety of game tests are a decent balance of all the areas of system performance. Some tests respond to CPU clocking, some to memory bandwidth and system througput and some to increases in graphics performance. More than 6000 is our target. It's a target that is somewhat easy for systems now and if the Shuttle doesn't do comfortably more we'll worry!
We sail past 6000 with ease in this test and 8000 is despatched without difficulty when overclocked indicating very strong overall system performance when the board is paired with a high powered graphics accelerator. Again, the out of the box score is better than the KR7A-RAID. The AK31 is working well with the Samsung memory and the aggressive memory timings (we used very similar timings on the KR7A-RAID too) are helping things purr along here.
It's tempting to discard 3DMark as a benchmark these days with the advent of the GeForce 4 Ti cards which annihilate the benchmark on a system like this. But for those not able to run a GF4 Ti, it's very much relevant.
Overall the performance of the AK31 is very very strong. It's the quickest KT266A board we've yet seen at Hexus, just ever so slightly beating out the KR7A-RAID which was maybe having an off day in my hands when I tested it! There is no doubting the performance of the AK31 and the KT266A chipset with the combination returning very strong results across the board, especially when paired with the GeForce 3 Ti550.
The board overclocks effortlessly and while the 166MHz front side bus limit might be restrictive for some it does get to the limit very easily. You are helped on your way with a nice range of Vcore adjustments (memory voltage adjust is slightly weak) and the memory settings are easily adjusted for maximum performance.
While the board isn't exactly setting the world alight with its features, the performance is very very good and it's a decent overclockers board to boot.
Performance wise it does well.
Overall we have to say the strengths of the board lie mainly with its performance. It's fairly featureless and the bundle and presentation are steadfast but hardly exciting. While the average user may not care for the extras, it's nice to see a bit more effort than the norm made with the presentation on boards these days.
The layout is clean barring the socket rotation which is a niggle of mine but might be fine for you. Otherwise, everything is fine. I couldn't fit the Swiftech as mentioned when I discussed the layout so watch out if you use a large heatsink like the MCX or an Alpha.
To conclude, the board has good out of the box performance, good expansion and is easily tweaked from the overclockers point of view.
Availability in the UK is thin on the ground, something that Shuttle has to work on. A quick look on pricewatch turns up nothing for the AK31 so its reach will lie with the smaller retailers so you're going to have to look hard to find it. I'd estimate price to lie just over £100 given the features.
A nice board, good performance with some niggling issues. Availibility is low.
Good CPU Vcore adjust range for the overclocker
Overall performance out of the box
Swiftech MCX478 wouldn't fit
166MHz front side bus ceiling