Introduction
Quite a lot of new boards found their way to my lab since my last Socket 7 review and quite a few of them are using chipsets which are not from market leader Intel. It seems as if the other chipset manufacturers as well as the board designers have done their homework, so that for the first time a board with a non-Intel chipset is now the fastest Socket 7 motherboard on the market. All motherboards support 75 MHz bus speed, so that by definition all of them should be able to run the 6x86MX PR233. Only six boards give you the chance of running at 83 MHz bus speed.
Please have a look at the results first.
Abit IT5H rev. 1.5
The IT5H rev. 1.5 has already been tested before and you can find it on my motherboard page as well as on a special IT5H reviews page. I was using this board as reference so that you can compare the result to previous ones.
Unfortunately I’ve recently heard quite a bit of complaints about this board. Probably most of you know that the IT5H is not able to take advantage of ECC RAM, but this is not a really big issue, however now I was informed that the latest shipments of IT5H boards have serious overheating problems if used with K6 CPUs, which in some cases even lead to the destruction of the board’s voltage regulator. Abit is pushing the release of the revision 2.0, which will take care of the ECC RAM issue and will finally be equipped with a switching voltage regulator to solve this overheating problem.
Abit PX5 rev. 1.14
Abit has been a little bit unlucky with the first revisions of the PX5 which have been giving lots of people quite a bit of trouble. Now the new revision 1.14 is supposed to solve all those problems. Indeed the board ran quite nicely and I didn’t come across any problems, but only after I picked the right SDRAM. Don’t even think of using Corsair SDRAM with this board, it imply doesn’t like it. I was using Toshiba 10ns SDRAM successfully on it. The performance of the PX5 is still not much more than average, but it supports 75 and 83 MHz bus speed. TX Board.
The test with the 6x86MX PR233 didn’t really run as smoothly as I would expect this. The instability problems of the PX5 obviously get worse when increasing the bus speed. I managed to run the tests often enough for getting results, but under Windows 95 I occured 3 crashes.
AOpen AP5T rev. 3
The latest revsion of the AP5T is now coming with a switching voltge regulator. It also has been revised to run even more stable particularly at higher bus speeds. The perfromance of the board is also kinda avrage, but if stability is a big issue for you this board may be the way to go. This TX board runs at 75 and 83 MHz bus speed. It has the very interesting option of ‘suspend to disk’ in case you are into power management. This board can now also monitor the CPU temperature.
The most amazing feature of this board is its stability. If you are into overclocking this board should be your no. 1 choice. No other board gave me the ability to overclock the new 6x86MX I got to even 225 MHz and run all my tests. From the stability point of view this board is unfortunately a complete exception. I wish all boards would run as stable as the AP5T.
As the inductivities easily show, the new AP5T includes a switching voltage regulator.
AOpen AX5T rev. 3
Finally the AX5T goes 83 MHzas its Baby-AT form factor brother AP5T. The board is not only running very stable, it also was the best performer of the Intel chipset boards in this test. This board is a nice one, in fact my favorite TX board. It comes also with the ‘suspend to disk feature’ and you can also monitor CPU temp., board temp., fan, … in case you fancy that. It also comes with a switching voltage regulator.
This board showed also in the 6x86MX test that it’s the no.1 of the TX boards. Although without a linear burst option and only 512kB cache it got fairly close to the winning PA-2007.
FIC PA-2007
The new number one amongst the Socket 7 boards is called PA-2007 and is manufactured by FIC. This board is using the VIA Apollo VP2 chipset, also known as Eteq, PC Chips or AMD 640 chipset. It supports SDRAM and Ultra DMA just as well as its TX chipset competitors, but its big advantage is the 1 MB onboard level 2 cache, which is mainly responsible for its excellent Win95 and especially NT performance. The only thing I have to criticise is the fact that this is a Baby AT board and there isn’t any ATX version with 1 MB cache out yet. Due to its VIA Apollo VP2 chipset it also supports linear burst for Cyrix/IBM CPUs, so that this bord is a double win for every Cyrix/IBM CPU user, since it’s not only the fastest Socket 7 board anyhow, it even speeds up your 6×86 or 6x86MX CPU by another 1.5 to 2%. Using switching voltage regulator.
As if we wouldn’t have expected it, in the 6x86MX test the PA-2007 could even extend its lead over all the other socket 7 boards due to its linear burst support. The board runs stable at 75 MHz and the linear burst in combination with 1 MB level 2 cache makes it the only really cool choice for Cyrix/IBM CPU owners. What a shame that it doesn’t support 83 MHz bus speed!!!
That’s what does the job. The four level two cache modules, summing up to 1 MB and the 
BGA chip of the VIA Apollo VP2 in the background.
FIC PA-2010+
The PA-2010+ is using the VIA Apollo VPX chipset, the newer version of the Apollo VP1. This chipset wasn’t a too good performer in the past, but now it does pretty well. However it’s still not fast enough to avoid the last place in this comparison from the performance point of view. This board also supports linear burst, so 6×86 users might consider buying it.
Indeed the linear burst option of the PA-2010+ could get it closer to the rest of the field, however even in the 6x86MX test it could get rid of the last place of all the boards that actually ran the test. Nevertheless the PA-2010+ combined with a Cyrix/IBM CPU is not far at all from the competitors and it offers 83 MHz bus speed!
FIC PA-2011
The PA-2011 in the revision that I got comes only with 512 kB level 2 cache and without any COAST socket as the previous version that was shipped in AMD’s K6 hot boxes. Due to this fact it can’t reach the performance of its brother PA-2007. It also uses the VIA Apollo VP2 chipset and also supports linear burst. This board is certainly not bad at ll, but you wonder why FIC wasn’t putting 1 MB level 2 cache on this ATX board as it did on the PA-2007. Switching voltage regulator.
The linear burst support of the PA-2011 could help giving this board a better placement under Windows 95 when tested with the 6x86MX. However the 512kB L2 cache are just too little to compete against Intel’s TX chipset under Windows NT. No 83 MHz support.
Although the PA-2011 is considerably larger than the PA-2007, FIC’s engineers could only find enough space for two L2 cache chip, therefore only 512 kB L2 cache. In the background on the right you can see the switching voltage regulator.
MSI 5147
The MSI-5147 is a very interesting as well as surprising board. It is using the SiS 5571 chipset which is well known by lots of 6×86 users, since it’s used on the M-Tech R534. However the MSI 5147 certainly makes less problems than the R534 and it comes in ATX form factor. The performance of this board is better than the performance of the R534 and altogether quite surprising, especially compared to Intel’s TX chipset boards. The MSi 5147 also supports ‘linear burst’ so you might go for this one in case you are a 6×86 or 6x86MX fan. In case you should want to use SDRAM on this board take these two important advices: Get the latest BIOS first or let the store flash it on the board before you buy it. The BIOS mine came with wouldn’t boot at all with SDRAM, not even from floppy. That would leave you quite stupid in case you haven’t got any EDO or FPM RAM to at least flash the new BIOS. Secondly you’ll have to set the SDRAM timing to x-2-2-2 instead of X-1-1-1 as you would think and leave ‘Enhanced Memory Write’ disabled, because it’ll crash otherwise.
This board was again costing me several hours and a lot of nerves. THe 75 MHz bus clock, required by a 6x86MX PR233 CPU is too much of a demand for this board. Please forget about using this board in case you want higher bus speeds than 66 MHz. What a huge disappointment! No 83 MHz bus clock support, of course not!
Use these settings with SDRAM and 66 MHz bus speed:
In the front you can see the SiS5571 chipset with its typical heatsink, the upper right corner shows you the switching voltage regulator of the MSI-5147.
Shuttle HOT-603
The HOT-603 is using the AMD640 chipset (equal to the VIA Apollo VP2) and has been designed after AMD’s guidelines for a good K6 board. The bootup message of the BIOS and the AMD logo on the right where the EPA one usually resides remind you of this each time you are starting the computer. This board also comes with 1 MB of level 2 cache, which is mainly responsible for its second place under Windows NT. The revision of the test board was pre-release so it might be that this board will be tweaked up a little bit more to reach the performance of the FIC PA-2007. The board comes in Baby AT form factor and has three instead of only two DIMM slots. As all the VIA chipset using boards it has a much larger cacheable area than the TX chipset boards. Switching voltage regulator.
This board is the other reason why this update goes online so late. I tried everything but I couldn’t get 75 MHz bus speed to work at all!! Instead of this I came a cross the strangest occurance!! When I jumpered the board to 75 MHz bus speed, it depended on the type and amount of RAM I used if it would boot at 75 or only 66 MHz bus speed!! Two SDRAM DIMMs and the board would just ignore my 75 MHz setting and run with 66 MHz. If it once ran at 75 it didn’t get far though … it always crashed at OS load.
I have to let this board off though, since it’s only a pre-release version which will most certainly not be released with those strange problems. Anyway, there was no way of getting 6x86MX PR233 results with it for now.
AMD640 chipset and four L2 cache chips — the easiest way to recognise the HOT-603. In the background you can see the three DIMM slots.
Windows 95
Boards marked with * were running with linear burst option enabled.
Please note that these tests were run with a IBM DCAS 2160W instead of the Segate Cheetah ST34501W I was using in previous NT tests. Hence the results are all about 4 Winstones less than in previous tests, they can’t be compared to previous results.
Windows NT
Boards marked with * were running with linear burst option enabled.
