Why my Socket 754 love affair is over...
Im at my wits end.
In my house, I have had FOUR socket 754 boards:
Biostar K8NHA Pro: nForce 150 (passed onto a friend in need)::
Barfed on WinXP installs, refused to function with ATA RAID card and was a horribly unstable motherboard. Awful BIOS, unable to change multi's, almost zero adjustment of any traditional features. Nasty PCB layout.
MSI K8T NEO: VIA KT800 ::
The only truly stable board of the four but had it's problems. Again, refused to work with ATA RAID card, glitchy with Audigy sound card, VIA SATA RAID corruption (resolved by using PROMISE controller or not using RAID at all).
Still, has served me well once the initial gripes were resolved and has been as solid as a brick since March.
Gigabyte K8NSNXP: nForce 250 ::
I really tried hard with this board. Really did. First the display wouldn't initialise (Radeon 9800xt). A flash of the VGA BIOS later, made it kick in but it took an age to kick in (often missing the BIOS altogether).
Windows simply refused to find a boot device. Even booting off four floppies failed to aid installation of SATA or IDE. Flashed BIOS several times. A dead one Jim (as they say).
MSI K8N NEO: nForce 250GB ::
The latest (and reportedly greatest) motherboard from MSI. Carrying the full version of the NF250 chipset. Unusuable.... why?
Memory. Even with just one stick of TwinMOS memory in the board, BSOD hell. I got puzzled and annoyed over the course of the last two days. This is BH5 memory, approved for use with just about any machine you can throw at it. It has run 11x220 in an NF-7 and cruised at Cas2 in the K8T.. yet apparently, TwinMOS performs terribly with this board (MSI words) but it's on their approved list. Interesting.
.. so in short....
I love the concept of socket 754, I love the ideal of the next generation computing platform. From reading many forums, boards and newsgroups with other experienced, knowledgable PC enthuiasts who have suffered the same hardware growing pains, Im far from alone. Really, with the best components money can buy.. you are still entering the great unknown. As Omega65 once said.. "bleeding edge is bleeding edge". It's a true statement.
When four "premier" motherboards don't work correctly (out of the box), it's a bad show of a promising technological ideal. Frustration is frustration, no matter how good you think or know you are at something.
I've come to a sad conclusion...
It's almost there, but needs more time. More time spent refining and testing chipsets, more time spent defining tighter controls. If an end user needs to know they can only use x memory from x manufacturer, then say so. If it is likely to be incompatible with x card.. say so! Loud and clear. Don't bury it in a .pdf on the slowest loading site from Taiwan.
There are just too many little glitches and hurdles to jump through before this is fully matured. We all know AMD want to hit the big time and penetrate the big world of enterprise.. but it's not all roses round the door with A64. Not yet. While AMD may have done their homework, others need to really think about what they are putting onto the shelves. Where is the NF7-S of the A64 world? It's not with us yet.
/me looks lovingly at NF7-S & Barton 2500+...
I might just sell the A64 lot and return to the Socket A platform that never let me down
... unlike socket 754, which is really starting to leave a sour taste in my mouth.
In my house, I have had FOUR socket 754 boards:
Biostar K8NHA Pro: nForce 150 (passed onto a friend in need)::
Barfed on WinXP installs, refused to function with ATA RAID card and was a horribly unstable motherboard. Awful BIOS, unable to change multi's, almost zero adjustment of any traditional features. Nasty PCB layout.
MSI K8T NEO: VIA KT800 ::
The only truly stable board of the four but had it's problems. Again, refused to work with ATA RAID card, glitchy with Audigy sound card, VIA SATA RAID corruption (resolved by using PROMISE controller or not using RAID at all).
Still, has served me well once the initial gripes were resolved and has been as solid as a brick since March.
Gigabyte K8NSNXP: nForce 250 ::
I really tried hard with this board. Really did. First the display wouldn't initialise (Radeon 9800xt). A flash of the VGA BIOS later, made it kick in but it took an age to kick in (often missing the BIOS altogether).
Windows simply refused to find a boot device. Even booting off four floppies failed to aid installation of SATA or IDE. Flashed BIOS several times. A dead one Jim (as they say).
MSI K8N NEO: nForce 250GB ::
The latest (and reportedly greatest) motherboard from MSI. Carrying the full version of the NF250 chipset. Unusuable.... why?
Memory. Even with just one stick of TwinMOS memory in the board, BSOD hell. I got puzzled and annoyed over the course of the last two days. This is BH5 memory, approved for use with just about any machine you can throw at it. It has run 11x220 in an NF-7 and cruised at Cas2 in the K8T.. yet apparently, TwinMOS performs terribly with this board (MSI words) but it's on their approved list. Interesting.
.. so in short....
I love the concept of socket 754, I love the ideal of the next generation computing platform. From reading many forums, boards and newsgroups with other experienced, knowledgable PC enthuiasts who have suffered the same hardware growing pains, Im far from alone. Really, with the best components money can buy.. you are still entering the great unknown. As Omega65 once said.. "bleeding edge is bleeding edge". It's a true statement.
When four "premier" motherboards don't work correctly (out of the box), it's a bad show of a promising technological ideal. Frustration is frustration, no matter how good you think or know you are at something.
I've come to a sad conclusion...
It's almost there, but needs more time. More time spent refining and testing chipsets, more time spent defining tighter controls. If an end user needs to know they can only use x memory from x manufacturer, then say so. If it is likely to be incompatible with x card.. say so! Loud and clear. Don't bury it in a .pdf on the slowest loading site from Taiwan.
There are just too many little glitches and hurdles to jump through before this is fully matured. We all know AMD want to hit the big time and penetrate the big world of enterprise.. but it's not all roses round the door with A64. Not yet. While AMD may have done their homework, others need to really think about what they are putting onto the shelves. Where is the NF7-S of the A64 world? It's not with us yet.
/me looks lovingly at NF7-S & Barton 2500+...
I might just sell the A64 lot and return to the Socket A platform that never let me down
... unlike socket 754, which is really starting to leave a sour taste in my mouth. 0
Comments
I hate to say it, but that's the sort of stuff I expected from A64 boards at this point in time. A whole new architecture needs much more time than the "new x86 chip" in the past needed to mature.
I imagine next year it will really take off, though.
The socket 754 is the msi neo also and its been flawless. One of the most stable boards I have ever owned and perfect in every way. It;s my daily driver and runs cool and silent 24/7
Tex
Now what was that saying about a bad workman always blames......
This is my feeling on it
As I said in my post matey, the K8T is and remains the most stable of the four. It's my daily workhorse, hell 95% of the new sites code was written and tested on it. It's the others Im more commenting on. and the general quirks of all of them and my honest thoughts on it all.
Opteron is a different animal altogether. My buddy Stuart has a 144 Opteron on an ASUS and has had zero problems. Opteron's are very nice.
The new nForce 250GB ASUS was <s>in</s> my next planned Socket 754. Im just very unlucky or I need to kick myself around the house more. I swear I even read the damn manuals!
I think Im just royally annoyed that I would have to go buy £200 worth of Corsair for my K8N to work
So many people had trouble with the dual AMD boards bssed on the old MPX chipset but 90 percent of those probs were from folks trying to run non reg ecc memory. I wonder how much of it is related ?
I just got lucky I guess on the memory in the msi neo? I don't know? I got a GB of cheap pc3200 Geil on eBay and just chunked it in there and it's never blinked. Its not their high end crap but their low end memory,
Tex
I've been recommended either Corsair Value or some of their ultra fast memory. Im not going to be overclocking the K8N (even though it has working PCI/AGP locks!).. so the value might make the difference. Or I could just sell the damn lot and get a dual opteron
I for one followed his advice. Not regretting it for a second.
Sticks of cheap registered pc2100 run me 20 to 30 bucks a stick on ebay. I boiught another pair of microns a couple weeks ago for 39 for the pair.
Tex
I have a CO stepping...
I got a 3200+ C0 stepping too. Do you have 256MB or 512MB TwinMOS sticks Dan? I got a pair of 512MB TwinMOS BH5 modules in my server which I can try in my A64 setup if you want. From experience I seriously doubt 2x512 BH5 will yield any decent OC results...I'm loosing 15-20MHz just going from 1x256MB to 2x256.
512mb's
Im not looking for overclocking at the moment, just getting this new MSI board to actually be stable. :banghead: ..
Check out:
http://forum.msi.com.tw/thread.php?sid=&postid=348417#post348417
There are plenty more. Apparently Toms Hardware's tests said that CO stepping failed with this RAM in their tests
Seems to me that your A64 RAM controller gets overloaded. I said I lost 15-20MHz in RAM OC when moving from one to two sticks. But I don't loose that in Memtest/Goldmemory stability...the problem manifests itself in CPU stability (fails Prime95 and so on).
Shorty - sell the Twinmos & CPU and buy some Corsair and a CG!
Cheers for that, I was starting to wonder if I was seeing things.. or if this hardware just had it in for me
Omega, man.. arrows?? I've got some rockets that have exploded in mine... from socket 754 hell
Corsair on order and .. anyone wanna buy a 3200+ Clawhammer core? CO Stepping... DONT USE TwinMOS!
One thing I do love about both the K8T and K8N... native Serial ATA. If the A64 chipsets have brought one good thing for us, it's native SATA (non RAID).
@ Isevald, I also found that in my DFI NF2 rig, that my TwinMos BH-5 (256mb sticks) ran more stable & overclocked higher when at cas 2.0 than 2.5. Mackanz helped me discover this when we were trying to get the board to run stable at any given speed.
It'll be a few days before I post any results with that tool. My scsi disc is being rma'd at present & I have no 32bit os installed on the rig, only 64bit. When it does come back & I've got windows installed again I'll be sure to post my results for you.
What's the S-M FAH Team Discounted Price?!?
Great
Will depend on what happens when my new memory arrives. If it still croaks, Il make sure you get the right price
But i am leaning towards selling my A64 and board and going back to socket A till next year then go S 939 when its more mature
It's not a timing parameter stored in the DIMM SPD. My guess is that with more than 2 banks populated, the BIOS should autodetect and enable 2T for greater stability. 1T greatly improve RAM performance, but it's affecting stability at high DIMM load. I believe most recent boards defaults to 2T if a CG is installed. I found some info here:
http://www.amd.com/us-en/assets/content_type/white_papers_and_tech_docs/31410.pdf
No, every time I try & set it to T2 enable it reverts back to disable.
my chip
default settings
with 2T set to 'disable'