update: Anandtech now has a review of this and other KT333 boards, which can be found here
well, my Corsair mem is finally working at spec (in DIMM slot 1), but with other faults. 4:5 cpu/mem divider, Turbo "System Performance" option. (this does give noticeable boost over Optimal even with manual mem timings not by SPD) cas 2.0/3/3/5 1T auto (thanks mechBgon!) Seems like this MB has a lot of annoying quirks, the main one being the memory overvoltage/undocumented jumpers. add in the fact that they are very difficult to reach without removing cards from your system (or some serious tweezers) and it is worse
Anyhow, when I keep the jumpers on to get my mem to run at spec, my USB hub doesn't work! Which is unfortunate, considering I need it to use two gamepads to play games on a TV, and use it frequently with digital cameras.
BIOS Flashing
I've seen a lot of feedback about NOT flashing from windows if you are at a BIOS of 1004 or earlier. There's a serious bug of sorts if you don't have 1005. Of course windows flashing in general is risky, but with 1005 I flashed to 1007 just peachy. If unsure, it's worth it to buy a BIOS savior IMHO...especially with MyLogo.
The BIOS Savior RD1-2M is NOT compatible with this board. (but the RD1-8X is)
Nowhere on Asus site/manuals have I seen it documented, nor the RDI site (only says A7V) ...it uses a smaller, different shaped BIOS chip and I believe you need the (newer) RD1-8X 32-pin 4Mbit. (mwave has both, I got the wrong one
) If flashing bothers you, get an RD1-8X 
Boot tab
At least with the 1005 BIOS, I experienced some problems dependent on boot order. booting from CD with my HDD freshly formatted (#1 in order, CD #2) hangs after my NIC boot-on-lan ... possible issues with SCSI/RAID if orders are wrong? solution was to move the CDROM to #1 in order (+/- on numpad) until OS was installed. board seems a bit fussy. Might also want to make sure bootup virus detection is OFF since this is usually at an OS install.
feel free to add any comments and I'll edit them to show up top, I'll be adding a lot once I fiddle w/ these timings a bit more though.
this thread will be filled with lots of general help information and links as the day goes by and I (hopefully) manage to get this memory to run. This board has a lot of strange undocumented quirks but when it works, it is SOLID. (read: if you make it into windows, sandra & Q3 run ok... you're good) I felt the need to start it because I've read/replied to info in well over 10 threads in the past 2 days.
Full system specs:
Asus A7V333-raid (flashed to 1007 BIOS)
Athlon XP 1900+ w/ retail HSF & AS3 running cool @ 49C max load (or so the sensor says, 55 tops I'd say)
512MB Corsair XMS PC2700 @ cas 2.0/3/3/5 w/ 1T 4:5 turbo
MSI Geforce4 Ti4400 @ stock speeds w/ MSI drivers (ref 29.42's now)
Antec True430W PSU
3com nic, SB live value, USR 56k ext. (woot)
link to compare with my Sandra scores (which compare decent to this review) when you consider the timings. and yes, I don't like to read into these marks too much but they are the best measure for this kind of rapid-reboot testing. Basically this is for reference of what to try and what not to try. I found this thread which makes me believe it's probably the memory, as opposed to this board (which is quite nice!)
update: I setup 98SE and it was just as fussy as XP, sandra rebooted when I tried w/ fastest timings...gotta be the mem
good news for A7V333 owners though. update2: It sounds like this board's strange voltage settings have something to do with the mem not going it's fastest...when I leave the jumpers ON it works, but other stuff stops. (ugh)
for all settings, 4:5 is CPU/mem divider, Optimal/Turbo is for SPD settings (even when I'm not using it, it seems to do other changes as well) ..auto means dimm bank settings (4-bank,2-bank, auto), 1T/2T is precharge time. A indicates memory now in bank 1, not 3.
Sisoft Sandra Memory Benchmarks (sort of in order of speed/tested)
config div- SPD CAS PRE BANK INT FLOAT STABLE
num ider ------------------------------------------------ (y/n/?)
1 ----- 4:5 opti - 2/3/3/6 2T AUTO 1952 1848 y
2 ----- 4:5 opti 2.5/3/3/6 1T AUTO 1976 1881 y
3 ----- 4:5 opti - 2/3/3/5 1T AUTO 1992 1902 n not fully
4 ----- 4:5 turb - 2/3/3/6 1T AUTO 2032 1945 n bluescreen on 2nd test (IRQL NOT LESS OR EQUAL)
5 ----- 4:5 opti - 2/3/3/5 1T AUTO 1995 1900 n not fully, oddly it works better than the above but Q3 fails after 2 benchmarks. (very high ones!)
6A --- 4:5 opti - 2/3/3/5 2T AUTO 1950 1875 y fully
7A --- 4:5 turb - 2/3/3/5 1T AUTO 1975 1883 n no 3dmark
8A --- 4:5 turb - 2/3/3/5 1T AUTO 2016 1935 n no 3dmark
9A --- 4:5 turb - 2/3/3/5 2T AUTO 2002 1906 y fully (so far)
currently stuck @ 4:5 turbo 2/3/3/5 2T auto (manual not by SPD) ^^
10A-- 4:5 turb - 2/3/3/5 1T AUTO 2026 1945 y fully *
3dmark2001SE (1024x768 defaults) scores for setups:
#6A: 9616
#9A: 9852
that's about 250 3dmarks going from optimal to turbo (harder to get stable, but it is
) ... worth noting I guess ^^
#10A in 3dmark: up to 10064 (with 29.42's)
1007 BIOS and OPTIMAL/TURBO "System Performance" Setting
This BIOS flash adds this option, and turbo has a positive influence on performance and with the new BIOS I believe it was a lot easier to get stable at lower timings
(that and DIMM bank 1)
setting lists that DO NOT WORK (erased, it's my mem) update: If anything, I am guessing this is a problem with my memory. Especially because neither Corsair's site, nor Mwave could tell me the exact specs of this mem. I'm going to email Corsair today. It isn't XP, 98SE is the same way.
Q3 numbers (v 1.31, 28.32 dets) more soon):
most surprising and identifiable hint of hidden power:
4:5 optimal 2/3/3/5 1T auto
1024x768 max settings
193.1 fps, 175.5 fps, crash
vs
4:5 optimal by SPD (2.5/3/3/6) 1T auto
167.5 fps, 168.2fps, 172.5fps (169.4 avg)
more Q3
Q3 (cont) same drivers/version/etc still, took 3 timedemos in a row and avg'd together
comparison of turbo to optimal...
both are 4:5 2/3/3/5 2T auto (not by SPD), in 1024 max settings
optimal || turbo
172.2 || 179.0
I'm not sure what these settings change besides SPD, perhaps AGP speed? /boggle ...finally some solid higher performance (it's just 10fps...but once I turn on 8-tap anisotropic filtering and/or quincunx FSAA that gives me more headroom)
update: SoFII has no benchmark yet, demos are broken. JKII all settings (even 8-tap aniso + quincunx @ 1280x1024 down to 640x480 w/o) get 70fps...very odd. vsync is definitely off, both in and out of game.
conclusions so far:
1) that 1T precharge helps a lot, as does (not surprisingly) the initial cas 2.0, and also the fourth number (SDRAM Active to Precharge Delay) when it is at 5T. 6T seems less stable than 5T. The middle two 3's (RAS to CAS, and RAS precharge) seem to be the least important. the final 1T is proving to be the most stubborn, and that I *KNOW* Corsair says it can do. update: works with Asus' DIMM overvoltage, but that makes my USB hub conk out
*searches for it's external power adapter*
2) WinXP Pro and 98SE both act funky with low timings...and in the aftermath, XP handles memory issues a lot better
mem-related, not board related, IMO. (!voltage probs are board related though!)
3)I did some serious DV camera w/ firewire editing on this box today (w/ the stable settings) and divx encoding speeds were unbelievable... near real-time... I can only imagine if this memory would work at spec and I setup RAID.
4) MSI vs 28.32's don't make a huge difference until the memory timings are really up there. I haven't posted 2 pages of Q3 benchmarks yet because most of them are near identical...except when you pimp out the 28.32's with my as of yet unstable config. I can't wait to try the new 29.42 WHQL's once my rig arrives here in Washington
those 29.42's add some serious performance.
5) Asus overvoltage to DIMMs is not a good thing. With it, USB acts strangely. Without it, mem won't work at full timings. If you don't have good mem, supposedly it could fry your DIMMs at default configuration. I have no other explanation, as my PSU should be MORE than sufficient, yet not on this board.
I've tried my DIMM in the third bank, and now I am keeping it first bank with slightly better results. I am starting to like this board a lot!
HTH
well, my Corsair mem is finally working at spec (in DIMM slot 1), but with other faults. 4:5 cpu/mem divider, Turbo "System Performance" option. (this does give noticeable boost over Optimal even with manual mem timings not by SPD) cas 2.0/3/3/5 1T auto (thanks mechBgon!) Seems like this MB has a lot of annoying quirks, the main one being the memory overvoltage/undocumented jumpers. add in the fact that they are very difficult to reach without removing cards from your system (or some serious tweezers) and it is worse
BIOS Flashing
I've seen a lot of feedback about NOT flashing from windows if you are at a BIOS of 1004 or earlier. There's a serious bug of sorts if you don't have 1005. Of course windows flashing in general is risky, but with 1005 I flashed to 1007 just peachy. If unsure, it's worth it to buy a BIOS savior IMHO...especially with MyLogo.
The BIOS Savior RD1-2M is NOT compatible with this board. (but the RD1-8X is)
Nowhere on Asus site/manuals have I seen it documented, nor the RDI site (only says A7V) ...it uses a smaller, different shaped BIOS chip and I believe you need the (newer) RD1-8X 32-pin 4Mbit. (mwave has both, I got the wrong one
Boot tab
At least with the 1005 BIOS, I experienced some problems dependent on boot order. booting from CD with my HDD freshly formatted (#1 in order, CD #2) hangs after my NIC boot-on-lan ... possible issues with SCSI/RAID if orders are wrong? solution was to move the CDROM to #1 in order (+/- on numpad) until OS was installed. board seems a bit fussy. Might also want to make sure bootup virus detection is OFF since this is usually at an OS install.
feel free to add any comments and I'll edit them to show up top, I'll be adding a lot once I fiddle w/ these timings a bit more though.
this thread will be filled with lots of general help information and links as the day goes by and I (hopefully) manage to get this memory to run. This board has a lot of strange undocumented quirks but when it works, it is SOLID. (read: if you make it into windows, sandra & Q3 run ok... you're good) I felt the need to start it because I've read/replied to info in well over 10 threads in the past 2 days.
Full system specs:
Asus A7V333-raid (flashed to 1007 BIOS)
Athlon XP 1900+ w/ retail HSF & AS3 running cool @ 49C max load (or so the sensor says, 55 tops I'd say)
512MB Corsair XMS PC2700 @ cas 2.0/3/3/5 w/ 1T 4:5 turbo
MSI Geforce4 Ti4400 @ stock speeds w/ MSI drivers (ref 29.42's now)
Antec True430W PSU
3com nic, SB live value, USR 56k ext. (woot)
link to compare with my Sandra scores (which compare decent to this review) when you consider the timings. and yes, I don't like to read into these marks too much but they are the best measure for this kind of rapid-reboot testing. Basically this is for reference of what to try and what not to try. I found this thread which makes me believe it's probably the memory, as opposed to this board (which is quite nice!)
update: I setup 98SE and it was just as fussy as XP, sandra rebooted when I tried w/ fastest timings...gotta be the mem
for all settings, 4:5 is CPU/mem divider, Optimal/Turbo is for SPD settings (even when I'm not using it, it seems to do other changes as well) ..auto means dimm bank settings (4-bank,2-bank, auto), 1T/2T is precharge time. A indicates memory now in bank 1, not 3.
Sisoft Sandra Memory Benchmarks (sort of in order of speed/tested)
config div- SPD CAS PRE BANK INT FLOAT STABLE
num ider ------------------------------------------------ (y/n/?)
1 ----- 4:5 opti - 2/3/3/6 2T AUTO 1952 1848 y
2 ----- 4:5 opti 2.5/3/3/6 1T AUTO 1976 1881 y
3 ----- 4:5 opti - 2/3/3/5 1T AUTO 1992 1902 n not fully
4 ----- 4:5 turb - 2/3/3/6 1T AUTO 2032 1945 n bluescreen on 2nd test (IRQL NOT LESS OR EQUAL)
5 ----- 4:5 opti - 2/3/3/5 1T AUTO 1995 1900 n not fully, oddly it works better than the above but Q3 fails after 2 benchmarks. (very high ones!)
6A --- 4:5 opti - 2/3/3/5 2T AUTO 1950 1875 y fully
7A --- 4:5 turb - 2/3/3/5 1T AUTO 1975 1883 n no 3dmark
8A --- 4:5 turb - 2/3/3/5 1T AUTO 2016 1935 n no 3dmark
9A --- 4:5 turb - 2/3/3/5 2T AUTO 2002 1906 y fully (so far)
currently stuck @ 4:5 turbo 2/3/3/5 2T auto (manual not by SPD) ^^
10A-- 4:5 turb - 2/3/3/5 1T AUTO 2026 1945 y fully *
3dmark2001SE (1024x768 defaults) scores for setups:
#6A: 9616
#9A: 9852
that's about 250 3dmarks going from optimal to turbo (harder to get stable, but it is
#10A in 3dmark: up to 10064 (with 29.42's)
1007 BIOS and OPTIMAL/TURBO "System Performance" Setting
This BIOS flash adds this option, and turbo has a positive influence on performance and with the new BIOS I believe it was a lot easier to get stable at lower timings
setting lists that DO NOT WORK (erased, it's my mem) update: If anything, I am guessing this is a problem with my memory. Especially because neither Corsair's site, nor Mwave could tell me the exact specs of this mem. I'm going to email Corsair today. It isn't XP, 98SE is the same way.
Q3 numbers (v 1.31, 28.32 dets) more soon):
most surprising and identifiable hint of hidden power:
4:5 optimal 2/3/3/5 1T auto
1024x768 max settings
193.1 fps, 175.5 fps, crash
vs
4:5 optimal by SPD (2.5/3/3/6) 1T auto
167.5 fps, 168.2fps, 172.5fps (169.4 avg)
more Q3
Q3 (cont) same drivers/version/etc still, took 3 timedemos in a row and avg'd together
comparison of turbo to optimal...
both are 4:5 2/3/3/5 2T auto (not by SPD), in 1024 max settings
optimal || turbo
172.2 || 179.0
I'm not sure what these settings change besides SPD, perhaps AGP speed? /boggle ...finally some solid higher performance (it's just 10fps...but once I turn on 8-tap anisotropic filtering and/or quincunx FSAA that gives me more headroom)
update: SoFII has no benchmark yet, demos are broken. JKII all settings (even 8-tap aniso + quincunx @ 1280x1024 down to 640x480 w/o) get 70fps...very odd. vsync is definitely off, both in and out of game.
conclusions so far:
1) that 1T precharge helps a lot, as does (not surprisingly) the initial cas 2.0, and also the fourth number (SDRAM Active to Precharge Delay) when it is at 5T. 6T seems less stable than 5T. The middle two 3's (RAS to CAS, and RAS precharge) seem to be the least important. the final 1T is proving to be the most stubborn, and that I *KNOW* Corsair says it can do. update: works with Asus' DIMM overvoltage, but that makes my USB hub conk out
2) WinXP Pro and 98SE both act funky with low timings...and in the aftermath, XP handles memory issues a lot better
3)I did some serious DV camera w/ firewire editing on this box today (w/ the stable settings) and divx encoding speeds were unbelievable... near real-time... I can only imagine if this memory would work at spec and I setup RAID.
4) MSI vs 28.32's don't make a huge difference until the memory timings are really up there. I haven't posted 2 pages of Q3 benchmarks yet because most of them are near identical...except when you pimp out the 28.32's with my as of yet unstable config. I can't wait to try the new 29.42 WHQL's once my rig arrives here in Washington
5) Asus overvoltage to DIMMs is not a good thing. With it, USB acts strangely. Without it, mem won't work at full timings. If you don't have good mem, supposedly it could fry your DIMMs at default configuration. I have no other explanation, as my PSU should be MORE than sufficient, yet not on this board.
I've tried my DIMM in the third bank, and now I am keeping it first bank with slightly better results. I am starting to like this board a lot!
HTH