ASUS A7V vs FIC AZ11 – Duron Overclocking

SUMMARY: Two very different Socket A motherboards – a good example of “you get what you pay for”. The ASUS A7V outshines the FIC AZ11 in features and performance (see page 2 for ASUS undocumented settings and cooling mods).

We outlined earlier how folks are modifying the FIC AZ11 (one courtesy of PC Nut, one we bought) so you can overclock AMD’s latest CPUs (HERE). This mod is not rocket science, but it is exacting work; further, one slip and a $125 motherboard is toast. For most people, this is not the way to go.

PC Nut got a shipment of ASUS A7Vs with the required dipswitch on it already, so we bought one to try out with our Duron 600s. Slick – the way it should be.

Flip some switches and watch the Duron fly! These switches are not documented in the manual (BTW: The manuals are pretty good – lots of detail and very clear). However the settings were silk-screened on the board; I would expect at some point the manual will be updated.

I have used both boards for about one day, so what follows is a first look between the two boards – more testing will follow.

Starting to get the picture? My initial impression is that the FIC is a board rushed to market, whereas the ASUS in comparison is much more polished and user friendly. Whether or not you feel the price difference is justified is an individual decision, but I’m hard pressed to see the price/performance advantage for the FIC vs ASUS.

OK – what about performance? We have two Duron 600s and found that one was 50 MHz better than the other, so we used the higher performer for these tests. Each is unlocked, but we modified the voltage setting so they run at 1.85 volts. In hindsight with the ASUS – not necessary.

Both boards ran Si Soft Sandra benchmarks at virtually the same level, so the table below for the ASUS is representative of what the FIC will do:

What a second – did I make a mistake on memory scores? No, they don’t scale the way we have seen with Intel CPUs. Why? Simple – we are varying the CPU’s multiplier, NOT FSBs, so we don’t see the increases in memory performance as we have seen with other Intel-based comparisons.

The last entry (904 MHz) clearly shows the impact of higher FSBs – an 11% increase in memory performance at 113 FSB compared to almost the same CPU speed at 100 MHz.

As a further example of the impact of FSBs on memory performance, I benchmarked memory at the following settings:

More like what we have seen with FSB overclocking – increasing FSB by 10% yields increases in memory performance of about 9% overall. So rather than use 8 x 100 for 800 MHz, why not 6 x 133? Correct in theory but wrong in practice. I could not get the ASUS to work at any FSB over 113 MHz. In the picture of the board’s jumpers, the four on the left jumper set FSBs of 100, 103, 105 and 110. In BIOS, you have a whole bunch of settings ranging from 90 to 145 MHz. USELESS!

But understandable – with the VIA KT133 chipset running at 200 MHz, FSB increases run up very quickly. We’ll be looking into this in more detail.

The “hidden” issue here is app performance: I don’t think, for memory-intensive apps, a 50% increase in CPU speed will translate into a 50% increase in application performance. My simple mind says “More CPU speed but at a decreasing memory bandwidth to CPU ratio”. We’ll be testing this area further – I don’t have an answer on this yet.

However, the high 200 MHz FSBs result in noticeably more heat from two components – The KT133 chipset and the PLL controller.

The puny, non-greased “greenie” we all know and love was warm to the touch – not hot, but warm enough that you can’t keep your finger on it. The PLL chip, however, was HOT! You will burn you finger. I threw my Radio Shack I/O on it and it went off scale. Both chips need better cooling, so I am using a TennMax Lasagna on the KT133 and some small IC heatsinks on the PLL chip.

Pushing the board will require some serious cooling for stability.

I did find a significant performance difference between the ASUS and FIC: The ASUS looks to be stable at 900 MHz, while the FIC was stable at 850 MHz. I did get the FIC up to 900 MHz but not totally stable with a water-cooled peltier. In both tests, I used water cooling.

AMD’s chips run hot. That’s the long and short of it. I seriously question how air-cooling is going to allow these chips to hit one GHz+ speeds. I was pleasantly surprised to see the following:

Now, only someone who has epoxied bolts onto motherboards would view these holes with pleasure – it means that you can bolt on all manner of cooling goodies EASILY. Not also that the socket has three mounting lugs on each side – much better than the anemic two on socket 370s (the off-center lugs are virtually useless anyway).

So when I first saw the FIC, I said “BRAVO!” The when I saw the ASUS, I said “Wait a minute – this looks like some kind of cooling spec.” This is a hunch, but with both boards having the SAME HOLE PATTERN AND SPACING, I think we are going to see some aggressive cooling solutions using these holes (or could be that ASUS and FIC source the board from the same company that uses the holes for something else).

Update: “Office Boy” sent me an email and it turns out that this is part of AMD’s socket 7 spec (HERE), apparently at the request of KryoTech; so all socket 7 boards will have these holes. More later.

I placed the ASUS temp probe next to the CPU (I much prefer internal diode) to monitor CPU temps. I trimmed the leading edge so that the thermistor was closer to the CPU. With water cooling and running at 900 MHz, the highest temp I saw was 33 C (ambient 23 C).

I also noticed that the power transistors were warm and some directed cooling in that area can’t hurt.

Duron overclocking is a different animal – not bad, just different. We will be doing more benchmarking and living with these boards to see their ins and outs. Of the two, the ASUS A7V is a clear winner.

Many thanks to PC Nut for the FIC and overnighting the ASUS to us.

Email Joe

What follows are a few things about the ASUS A7V you might find interesting:

Turns out you can set FSBs from 90 to 113 MHz using the on-board dipswitch – the Manual only shows four. I tried all the combinations (I hope) with the following results:

This chip runs warm, and as a matter of routine I replace the inadequate “greenies” with a TennMax Lasagna. I found that the hole pattern used on the ASUS is just small enough that you have to adapt it slightly to fit. I drilled out the tab, making it just a bit wider, so that it the holes line up and it easily clears the capacitor.

Similar to the ASUS P3V4X, this chip runs hot. With the A7V, this is finger-burning hot – my Radio Shack I/O was off-scale, so it is over 65 C at least. Users found that cooling the PLL chip on the P3V4X lead to better stability, so I would expect the same benefits here.

Pictured above is “stop-gap” cooling – I am using two small IC heatsinks held in place just by thermal grease. Because of the tight spacing, directly mounting a TennMax Lasagna is not possible. I think I can do it by gluing a small aluminum or copper spacer to the chip with thermal epoxy and then gluing active cooling onto it.

Any other ideas welcome – drop me a line.

Email Joe

UPDATE 8/18/00:

I am humbled by James Barsing’s expertise:

“Hi Joe – just read your update on the Asus A7V jumper free situation and thought I’d let you know that there are no tricks required to get my A7V to operate as you describe. Basically this is how I have it setup:

• Voltage jumpers to 1.85 setting,
• FSB dipswitches all OFF,
• Multiplier dipswitches set according to multiplier required,
• Jumper free jumped to “jumper free” position.

With that setup, I have full control in BIOS over FSB, Mem, Core voltage etc, and can change my multipliers at will – no tricks required. The board is a rev 1.01, no audio running BIOS 1003.

I don’t know whether it’s the combination of what I’m doing with the jumpers/dispwitches, but it’s worked like this for me since day one. The only way I lose the a BIOS settings and create a POST error is if I don’t have all the FSB dispwitches OFF.

Obviously there are more slightly different versions of this board out there than Asus would have us believe..

Regards, James”

I tried it and it works flawlessly. The ASUS Manual states that voltage jumpers for jumper free must be set to DEFAULT, so this is NOT what ASUS recommends. Thank you Jim!

UPDATE 8/17/00:

Lots of email on the “Jumper Free” issue; seems like some have the same problem, some don’t. I played around some more and found out what the issue is, and it’s not an obvious one.

Looks like when you change from using jumpers to “Jumper Free” mode, there is a little dance you must do or you’ll never get back to “Jumper Free” with the multiplier dipswitch activated. Here’s what I had to do to:

• Set all the switches for “Jumper Free” mode. Set switches 5 and 6 on the multiplier dipswitch to OFF.
• Boot up. Board runs fine, BIOS uses DEFAULT settings. Make any changes to FSB and VCore and save, reboot.
• Board fires up at new settings but at CPU default multiplier.
• Power down, switch 5 and 6 to ON and reboot.
• The board will not boot. At this point, I was doing a power-off reset and getting the same result. HOWEVER, if you hit RESET, the board boots up, you get an error message and the BIOS resets FSB, RAM and VCore settings to DEFAULT.
• Change whatever you want in BIOS and reboot.
• I then get the same error message, so I go into BIOS and find the settings I made unchanged. Save and reboot.
• Now it works fine with the dipswitch multiplier active.

HOWEVER, if I change the multiplier, I go through the same routine again! I think this is a BIOS issue that I hope will be fixed in a new release. Thanks to all who sent emails on this – it spurred me on to a workable solution. I can’t tell you every A7V must go through this routine, but this is what works for me. The board is rev 1.01 – maybe the rev 1.02 boards don’t have the same issue.

BTW: The latest ASUS A7V Manual states that Jumper Free must be disabled for the multiplier switches to work – go figure!

Dave Jeffers
of Outside Loop Computers sent this email which seems to confirm the A7V’s quirky behavior:

“After spending a decent amount of time and effort in overclocking the Durons and Thunderbirds with the A7V (version 1.01, w/o audio), I found it to be a very sensitive and challenging proposition.

Specifically, gaining a post after a lockup, or windows error, was very difficult – even after clearing the CMOS, reseating EVERYTHING on the board, and jumping back and forth between jumper and jumper-free mode. I was, however, successful in overclocking in jumper-free mode, which is the way all my overclocked A7V
combo’s are configured.

Yes, you must bump the core voltage up in the bios before shutting down the system and setting the dipswitches to overclock, but I had no problem with a subsequent post after raising the core voltage
(Using the PAL6035), assuming reasonable overclocks (800 at 950 or a
Gig). In jumper free mode, I was getting funky overclocked numbers, 50mhz above the listed switch setting! Go figure. Any one else see this behavior?

Interestingly, I found a real difference in stability (typically 50 mhz) between using the orb as opposed to the alpha. The orb just can’t dissipate the heat well enough, when generated at the full 1.85V Vcore.

Seeing as this lackluster performance of the ORB really pissed me off, I had my machinist bore out a Socket A Golden Orb and press fit a 1.25″ diameter x 0.65 deep solid copper puck into the orb base. I’ve been testing the thing the past two days.

Results? On an 800 T-bird at 1.85V, the thing loads Windows and ran 48 hours on Prime95 Torture Test without a hitch. Using the PAL6035, the T-bird posts but WOULD NOT make Windows after repeated attempts! No kidding.

Next, I lowered the T-bird to 850, 1.85V and repeated with the Copper-ORB and PAL6035. Steady state temps?

Copper-ORB: 57C
PAL6035: 54C…Hmmmmmmm?

My Theory? The large copper plug in the Copper-ORB has a larger heat
capacity so will allow an initial windows load when the PAL would
not. But in the long-run, the temperature comes up a little higher than the PAL since it doesn’t have as large a fan or as many cooling
towers? Bottom line is it allowed a higher post and subsequent successful torture test.

Anyone else’s thoughts? I’m thinking about having my guy
make a few more of these Copper-ORBs if anyone is interested.”

SUMMARY 8/15/00: One more spot to cool and two things to look out for – one which might affect your purchase decision.

I was puzzled by something and I thought the BIOS might be “off” – in “Jumper Free” mode, I was not able to boot up the board. Someplace I was doing something wrong and I could not figure out what. Humphrey of PC NUT called to chat and we found we both had the same problem.

I looked into this further and discovered it’s not a problem – the board is doing what it’s supposed to. In “Jumper Free” mode, ALL the dipswitches must be OFF; this means that the dipswitch that sets the multiplier must be OFF also. Ergo, you can not change the CPU’s multiplier in “Jumper Free” mode, ONLY the FSBs.

In my book, this renders “Jumper Free” meaningless for us; so if you are thinking of the ASUS A7V, unless there is a BIOS fix that can take care of this, motherboard adjustments will be by dipswitch, NOT by BIOS. If I’m wrong on this, please email me.

Using the “finger test”, I found more hot-spots on the A7V. The voltage chips get pretty hot, especially the ones closest to the CPU. I measured these under stress with the following results:

I have ordered some thermal epoxy and will mount some heatsinks on the chips – meanwhile, make sure that the air circulation in this area is tops.

And while we are talking about jumpers, talk about crowding! ASUS managed to cram three different functions into as small a space as they could: Fan headers, CPU voltage and thermistor header. Changing these does require some patience and I usually have to move the fan connector to get enough space. Not a great design.

If any users have additional comments, send them in.

Email Joe