ASUS Crosshair IV Formula AM3 AMD 890FX Motherboard Review

Motherboards/AMD AM3 by leeghoofd @ 2010-06-04

It is time to present the first 890FX motherboard that entered the Shrimps lab. And boy does it look stunning, ladies and gentlemen let me introduce you: The Asus Crosshair IV Formula. The previous Crosshair III was a master in efficiency, can the new 890FX chipset top the older generation? How does it pair with the new 6-Core Thuban CPU ? Let us go and explore this new addition to the Republic Of Gamer product line-up.

  • prev
  • next

A month in the life of an overclocking endeavor

So what's this…

Once we got the Thuban sample ( 1090T ) in the lab, the quest for high ram clocks began. Previous CPU's and motherboards were hitting a wall at around +/-1960Mhz ram speed. The 890FX chipset, bundled with the new hexacore CPU would completely throw that limitation over board. Yes lads we have seen +2100Mhz ram speeds with pretty tight timings all over the net.

After staring hours at Memtest86+ I really got a hard time to stabilise 2000Mhz ram speed at Cas 7-7-7-21 1T. It was always test 5 that capped out on me. It was hit and miss, for 1 run it was fully stable, the next run my whole screen just turned red.

Tweaking voltages, sub timings and co got me Memtest stable at 2000mhz Unganged setting. The PC booted nicely into Windows 7, a very good sign. Running Superpi1mb was no biggy at all. 32mb however crapped out on loop1. Back to the drawing board. Moving the dimms to the black slots stabilised the rams a bit more... yet pi32mb was still no go...

Tip 1: Use the Black slots.

Moving onto another bios, 0601,0701 and suddenly there was version 0707. This one being optimalised for memory compatibility. An Asus employee was running an 8GB setup at 2000mhz C7-8-7-21 after some tweaking.

How frustrating is that...I couldn't even stabilise 4Gb. My whole face turned blue due to the countless staring at the Memtest screen. Crosschecking with Massman ( who had another CH IV board ) he experienced the same issues as me. Lending out rams, settings... got us both a bit further but 2000Mhz Unganged 32Mb stable was never going to happen at the Shrimps lab. If we slackened to Ganged mode there was no problem to run 2000 even 2100mhz C7 flawless. Yet efficiency was far from ideal. After some debating at local forums the culprit for our underperformance seemed to be the ACBBE stepping of our samples. Retail samples are CBBE and seem to run 2000mhz Unganged mode without exuberant tweaking.

At the time of writing 0801 bios was amongst us. Most user being pretty happy with it, though one of my first LN2 sessions went up into smoke. At a certain moment I couldn't even get into windows at 4.5GHz CPU speed at -180°C. Adding more Vcore didn't help. Neither did lowering the total CPU speed. What was going on ? I spotted a far too low CPU NB voltage. While 1.38 was set, it only read out 1.151 in the bios. My Digital Multimeter confirmed the bios reading.

Madshrimps (c)


Why did the board lower the CPU NB volts so much ? It couldn't have been safe mode as I didn't experience an Overclocking Failed post. Also when resaving the bios settings, the rig posted nicely without any warning. But Windows failed to load again. Back into the bios, still 1.151 being really fed into the CPU's NB. Upping the NB voltage resulted in a higher value, but still far below optimal voltage levels. At start I was one of the few with this erratic behaviour, yet there are numerous users having remarked this undervoltage. Asus Bios engineers are working hard on the matter and a new beta bios, it should pop up soon.

Tip 2: Leeghoofd’s Workaround

Until the new bios appears this is my solution to reset the correct NB voltage:



We start off with a too low CPU NB voltage. Trying to boot numerous times didn't set it back to the correct value. What does: Switch off the PSU to trigger the overclock failed state. Then go into the bios by pressing F1, save the settings and after the next boot the NB voltage value should be corrected.

Most users will never encounter this syndrome as they might never push the system as far as some do. Yet for benchers it can be very annoying as every extra Mhz counts... It also seems to occur far more often if you set the voltages manually instead of using the Offset value mode.

A New “Alpha” bios has been distributed amongst testers and it cures the NB voltage issue, just stay away from 0801 bios if anyone advises you, await the fixed final version. EDIT : Beta bios 0903 seems to have been leaked. Final version will have a fixed NB issue.

Also when benching with Gamer for the HardwareInfo compo, we noticed big swings in the total CPU speed and related HT clocks. We are not talking 10Mhz here, but over 80Mhz difference. This makes it very hard to push a CPU to the edge when benching at the limit. I got this tool to check the real clockspeeds : THG Clock tool.

Clocks were a bit more stable with this tool. But setting e.g. 208HT clock in the bios lead to 210 being really applied (CPU-Z, TurboV and Clock tool confirmed). Apparently resetting the bios made it less erratic.

Tip 3: When swapping a CPU, clear the CMOS first!

My first LN2 session with this board ended abruptly with a system halted error. This was really scary as I though I had bricked the board.

Madshrimps (c)


After a good dry up, the board and CPU posted, as nothing ever happened. The next night I got the same error message, but this time already after 15 minutes of benching. Something didn’t like the cold at all. Asus asked me of a picture of how I insulated the board and advised me to cover the area, where the TurboV and ITE chip were located, with kneaded gum. I have never ever insulated an area so far from the socket, but hey it was worth a try. And it worked !

Tip 4: LN2 Benchers Advice!

For LN2 benching , cover the below area with Kneaded gum :

Madshrimps (c)



Tip 5: Think twice before BIOS flash

Before flashing a new bios, really consider it if it's worthwhile. As a flash to an older version is a no go due to the new AGESA code.

I am not going to bore you with the technical mumbo jumbo ( as I don't understand it all either) but a flash to an old version could corrupt your board. So think twice before updating !

Tip 5: Check Northbridge and Southbridge Heatsinks

In case you should read out pretty high NB and or SB temps, please check correct contact of the heatsink with the concerned chips. I didn't have spot any high temps on my 2 sample boards (around 40°C). In case of out of the line temps: Carefully remove the heatsink.

Madshrimps (c)


This user had zero contact on the NB, the culprit is found below. An offset that was not properly screwed on.

Madshrimps (c)


Check the screw fittings are correctly inserted into the heatsink design. Remove the yellow pads then and replace the ASUS TIM with e.g. Arctic ceramique.

But don't let this scare you. Most boards have to undergo a few bios updates to correct juvenile flaws. As long as the manufacturer accepts input from the End user, things can get sorted out faster...

  • prev
  • next
Comment from leeghoofd @ 2010/06/04
New Beta bios is up for grabs here : CrosshairIV-Formula-ASUS-0903.zip



Gary Key confirmed me the final version will have some small updates. Not tested this bios myself.

image courtesy of ChriSch
Comment from Teemto @ 2010/07/05
Hi guys,

Just got myself a new CH IV board as well. Flashed it to Bios 0905 as the inital version it got shipped with displayed the Spinpoint F3 issue.

Spinpoint F3 can't be detected by the Bios. In the latest bios this is fixed as you can select the controller to work in SATA-300 mode.

What kind of voltages were you using to achieve the OC's?
I'm going for 20x200 @1,40-1,45V but I'm not sure about the NB/SB voltage.
Any tips would be welcome.
What kinds of temperatures should the NB and SB be able to handle (as measured in ASUS PC probe)?

Grtz
Teemto

PS. The 555 from Gamer is coming my way that'll be for a second bench setup
Comment from leeghoofd @ 2010/07/05
NB voltages were mostly not that high around 1.35ish. CPU NB is dependant on the CPU type itself. For Thuban I never needed above 1.35 for the CPU NB, 965 BE needs 1.4 at least to get anywhere close to 3000mhz NB speed. it all depends on the cooling you use. It's still better to go 1800ish speeds with tight timings then to show off at 2000mhz with C7 or worse...

Ha you bought the 555. You can grab some points with it at Hwbot

if any issues you can always pm me for MSN contact...
Comment from Teemto @ 2010/07/19
Been testing a bit. So far the Crosshair IV allows higher clocks than the Crosshair III. Have my 1090T (with VNB=1.3V) now tested at:
4GHz 20x200 1.45V
4.1GHz 20.5x200 1.50V
4.2GHz 21x200 1.50V
4.3GHz 21.5x200 1.50V
Stable in SuperPi1M, Wprime32M, 3DMark Vantage, 3DMark06, Cinebench 11.5, Orthos (1h run w/o problems)
4.4GHz 22x200 1.55V (1.50V crashed during WPrime1024M)
Stable in SuperPi1M, Wprime32M. Further tests need to be done.
In the meantime upgraded my 2xHD5850's with Gamer's 5870 Lightning (second 5870 planned for crossfire)

My question is related to the RAM: I currently have my Patriot's running at 1600MHz 7-7-7-20-36-1T 1.9V for all CPU voltages. CPU/NB is set to Auto and HT Link is set to Auto as well (should probably set the HT Link to 2000 or slightly less depending on RAM speed). Would 1800 RAM with 888 timings be better? The Patriot's won't do more than 1667 at 888.
Comment from leeghoofd @ 2010/07/19
nah mate, either keep ya current kit or upgrade to Hypers or PCS rams... 1800Mhz C6 is where the love is... for benchers at least...

I'll be testing the Corsair GTX2, Gskill Flare and some cheapo kits soon... This to see if it really all pays off or you can keep ya money safely in ya wallet...
Comment from Teemto @ 2010/07/20
Looking forward to your review then
Comment from Teemto @ 2010/08/16
Quote:
Originally Posted by Teemto View Post
Looking forward to your review then
Where can I buy them? Didn't find any store in Belgium that had em.

 

reply