Not to mention the fact that the employees still can't play Half-Life 2 on it in SXGA resolution decently enough to waste their working hours.
Intel does have a line of good 945G boards, with DDR2-800 support, Firewire and other goodies, but, as usual, flexible BIOS-level overclocking is not there - unless you count the meagre 4% 'burn-in' mode available. So, we must look to the island of Taiwan for some good 3rd-party 945G-based boards with overclockability.
Abit LG-80 and 81 are the orange-coloured, 945G-based MicroATX-sized offerings (LG-81, which we tested, lack Firewire), and they do differ quite a bit from the reference Intel 945GTP board I have.
Firstly, only two DIMM slots in there - not exactly a minus, as a low-end user with such board might never go beyond 1 GB, plus it gets somewhat simpler tracing for memory overclocking. Otherwise, the features are the basic MicroATX 945G setup - one PCI-E 16X graphics slot, one PCI slot and two PCE-E 1x I/O slots, with 4 SATA2 ports, On board 7.1 channel Intel Audio, Gigabit Ethernet PCI-E controller plus of course the Abit SoftMenu.
BIOS setup-wise, LG-81 has decent overclocking flexibility, both on the FSB and voltage side. Unfortunately, changing the multiplier on the 3.73 GHz Intel Pentium 4 EE didn't really work, so it was assuming all Intel CPUs inserted had a fixed multiplier. We tried it with 2.8 GHz dual-core Pentium D as well at 800 MHz FSB. Unlike the Intel 945G boards, the Abit entry did not support DDR2-800 memory in the BIOS. Using the the old Corsair HydroCool 200, we pushed the Pentium D to 3.04 GHz and FSB866 to run - reliably over quite a few nights, rendering 3D.
I tested the LG-81 using both Intel generic heat sink / fan combo, as well as Corsair HydroCool 2 water cooling system. Both benchmarks - 3DMark05 v120 and Sandra 2005 SP1 - were run on both the single-core and dual-core setups, with both Corsair and Crucial DDR2-667 memories. In the case of P4-3733, we tried both HT and non-HT CPU settings (HT could have an impact on the thread that does the CPU-bound vertex shader for integrated graphics), as well as the ATI X600 as well as integrated GMA950 graphics to see the effects on the benchmarks - here are the results.
|CPU INT MIPS||9983||9925||11013||16533|
|CPU FP Mflops||4618||4612||7755||7544|
|MM INT it/s||21087||21087||26836||34401|
|MM FP it/s||24883||24888||35609||40653|
|MEM INT MB/s||6223||6002||6024||4956|
|MEM FP MB/s||6225||6004||6016||4968|
Well, quite a few surprises! I thought that having the second core which could dedicate itself to the vertex shader job would substantially improve the graphics scores despite 20% lower CPU clock & memory bandwidth. Well, there was some benefit to dual-core operation, but not enough to offset the clock difference. By the way, I tried to run PCMark05 as well, but it failed to complete on both CPUs.
Overall, a nice board with just about right functionality for midrange PC setup - basic overclocking is there, a good combination of slots, all that in MicroATX size. Yes, the graphics still can't catch up to even the cheapest add-on card, but it is OK for office and limited 3-D (i.e. simple 3 year old games) use. When you outgrow it, well there is always a PCI-E 16X slot there, and by that time the RV515, RV530 or their Nvidia equivalents may just turn even cheaper, just in time for a good upgrade without wiping the wallet clean. µ
Sign up for INQbot – a weekly roundup of the best from the INQ