Incompatibilites with Skt 939 AMD 64 and Memory?

[Kate]

Greets all,
Well, we finally decided on preferred vendor for our new Athlon 64 Skt 939
CPU and Asus A8V Wireless M'Board system and today visited the shop to
order.

To our astonishment the, helpful and apparently knowledgeable, sales person
advised us not to buy AMD or Asus at all but to go for Intel Skt 775 instead
and MSI M'board. He said there have been countless problems with
incompatibilities with the AMD and memory chips -- even with the expensive
Corsair Modules -- so much so that he thinks it may take a coupla months
before these can be reliable and stable.

This has thrown us right off course --- we came home with a recommendation
list but no new puter :-(
Can anyone comment on what he's going on about, I've read and read on this
issue and heard nothing before about this problem. What have I been
missing?

TIA

 

[Ed]

Greets all,
Well, we finally decided on preferred vendor for our new Athlon 64 Skt 939
CPU and Asus A8V Wireless M'Board system and today visited the shop to
order.

To our astonishment the, helpful and apparently knowledgeable, sales person
advised us not to buy AMD or Asus at all but to go for Intel Skt 775 instead
and MSI M'board. He said there have been countless problems with
incompatibilities with the AMD and memory chips -- even with the expensive
Corsair Modules -- so much so that he thinks it may take a coupla months
before these can be reliable and stable.

This has thrown us right off course --- we came home with a recommendation
list but no new puter :-(
Can anyone comment on what he's going on about, I've read and read on this
issue and heard nothing before about this problem. What have I been
missing?

TIA

There's a list of ram modules for the A8V Deluxe on Crucial's site, there ram always works for me.
http://www.crucial.com/store/listpa...s&mfr=ASUS&cat=RAM&model=A8V+Deluxe&submit=Go

Why not just build the system yourself? www.newegg.com

Ed

 

[Tony Hill]

Greets all,
Well, we finally decided on preferred vendor for our new Athlon 64 Skt 939
CPU and Asus A8V Wireless M'Board system and today visited the shop to
order.

To our astonishment the, helpful and apparently knowledgeable, sales person
advised us not to buy AMD or Asus at all but to go for Intel Skt 775 instead
and MSI M'board. He said there have been countless problems with
incompatibilities with the AMD and memory chips -- even with the expensive
Corsair Modules -- so much so that he thinks it may take a coupla months
before these can be reliable and stable.

This has thrown us right off course --- we came home with a recommendation
list but no new puter :-(
Can anyone comment on what he's going on about, I've read and read on this
issue and heard nothing before about this problem. What have I been
missing?

If you look at Asus' spec page for this product, you'll notice that
there is a list of certified memory modules for their A8V motherboard,
down towards the bottom of the page:

http://usa.asus.com/products/mb/socket939/a8v-d/overview.htm


Note that this is nothing special for the A8V, Asus has some similar
lists for other boards. It's also a fairly long list and should not
be at all tough to find a module from that list at any local store.
And finally, it's not the only source of memory guaranteed for use
with that system, if you go to:

http://www.crucial.com/store/listpa...s&mfr=ASUS&cat=RAM&model=A8V+Deluxe&submit=Go

You can find a list of the memory modules that Crucial has certified
for this board. Similarly if you go to:

http://www.ec.kingston.com/ecom/configurator/modelsinfo.asp?SysID=19150&distributor=0&submit1=Search

You can get Kingston's list of certified memory for this board.
Corsair probably has a list as well, though their website seems to be
down at the moment.


Long story short, it should be no trouble at all finding memory for
this system. My guess is either a.) the store you were buying from
uses low-end generic memory and has had trouble with that, or b.) they
have a lot of stock of Intel Socket 775 processors/motherboards and
are trying to get rid of them. Having worked in retail sales before,
I know that often times the salespeople are, umm... "encouraged" to
sell certain products, regardless of whether they are the best choice.

 

[George Macdonald]

Greets all,
Well, we finally decided on preferred vendor for our new Athlon 64 Skt 939
CPU and Asus A8V Wireless M'Board system and today visited the shop to
order.

To our astonishment the, helpful and apparently knowledgeable, sales person
advised us not to buy AMD or Asus at all but to go for Intel Skt 775 instead
and MSI M'board. He said there have been countless problems with
incompatibilities with the AMD and memory chips -- even with the expensive
Corsair Modules -- so much so that he thinks it may take a coupla months
before these can be reliable and stable.

That's a gross exaggeration and I've been following all this quite closely.

This has thrown us right off course --- we came home with a recommendation
list but no new puter :-(
Can anyone comment on what he's going on about, I've read and read on this
issue and heard nothing before about this problem. What have I been
missing?

Have you visited any Forum for the AMD mbrd in question? Start by looking
at the recommended memory modules for the mbrd at the mfr Web site. Much
of the "problems" I've seen are stories from disappointed overclockers;
there's also a "story" (check the Forum at www.msi.com.tw) that the AMD
memory channel(s) needs slightly looser timings than what is generally
spec'd in the module SPDs... in particular the Trcd, so that instead of say
2-2-2-6, the BIOS Setup needs to be tweaked to 2-3-2-6 for *some* modules
and especially the performance ones like this with such tight timing.
Apparently overclockers get very peeved at having to back off timing on
their high priced "fashion" DIMMs.

I can relate my experience: I've done a couple of Skt 754 systems with the
MSI K8T Neo-FSR and noticed that in the recommended modules list, the
Kingmax ones were consistently mentioned by MSI for several different
mbrds. They're well priced so I gave them a go and found they work fine
with standard SPD timings - checked that two 512MB modules (double sided)
worked together too. I went with the Kingmax PC3500, running at standard
DDR400 in the hope that it'd give a bit of extra headroom, so I can't say
definitely if the Kingmax PC3200s would do as well.

I just did a MSI K8T Neo2-F Skt 939 system with a 3500+ Newcastle and used
the Kingmax PC3500 modules again - two 256MBs this time, double sided.
Those are spec'd at the same 2.5-4-4-8 and they work fine in the Skt 939
dual channel config too. I played a bit more with the timings on this
system to see what'd work and not: turns out that they need 2T command
rate to work, the 1T would pass memtest86+ but caused minor probs in
installing WinXP and would fail within minutes of starting Prime95.
Kingmax does not specify a command rate so don't know what is to be
expected here, nor whether the 2T is due to the modules, mbrd or CPU.

I also tried reducing the other timings... just to see: here I found that
those modules work fine at 2.5-3-3-7-2T with DDR400 clocking. When I say
"work fine", I've really pounded on this system at those settings by
running Prime95 for hours and 3DMark2001se and 3DMark03. I'm convinced
it's as stable as any I've ever seen and it scores ~5GB/s with Sandra's
memory benchmark. Now I suspect that if I loaded up both channels with
DIMMs, I might have to back off to 2.5-4-4-8-2T but I'm not about to try
that any time soon.

I'm just getting ready to do a MSI K8N Neo2 Platinum with a 3500+
Winchester - already received the CPU, memory and some other stuff; still
waiting on the mbrd and case. I decided to try Crucial DIMMs (regular
ones, not the Ballistix) this time, since I've never had trouble with
Crucial memory modules and I've found they are often underspec'd - I want
to see what I can get out of them in the way of timings but will *not*
overclock. I'll report back when it's working... in a couple of days.

As for your "sales person", it's possible he has a better mark-up on Intel
based systems or he may not be clued up on AMD systems in general and just
be parroting the usual FUD. Another thought: socket 939 mbrds are in short
supply right now - the nForce3 ones in particular are just flying out of
the dealers and e-tailers.

Finally, a word of warning: if you install WinXP SP2, either as an SP2
update or streamlined in, on a VIA K8 mbrd, do *NOT* use the VIA AGP driver
- IOW from the VIA 4-in-1 install, refuse the AGP selection. In SP2, M$
has new "unified AGP drivers" for AGP 3.0 and 3.5 compliant chipsets and
you *must* use them to avoid trouble. In particular they have two drivers:
for AGP 3.0 and 3.5, the first specific to K8 systems; funny thing is that
for me, it did not install automatically with the WinXP SP2 install and I
had to go to "Update Driver" for the VIA CPU to AGP Bridge and manually
tell it to use the "VIA AGP driver (Microsoft)" version. This caused it to
install a driver called GAGP30KX.SYS and it seems to work fine - without it
the SBA and Fast Writes don't get enabled; with a AGP 3.5 chipset the
corresponding driver is UAGP35.SYS.

Rgds, George Macdonald

"Just because they're paranoid doesn't mean you're not psychotic" - Who, me??

 

[Kate]

Thank you all so much for these encouraging comments esp George and Tony.
As a complete novice my brain needs some time to compute what you're saying
and follow up the linx you suggest.

But already I feel a little more confident in our AMD choice ----- and a
little *less* confident in our chosen vendor :-0
----
Kate

George Macdonald wrote :

That's a gross exaggeration and I've been following all this quite

closely.

 

[keith]

Thank you all so much for these encouraging comments esp George and Tony.
As a complete novice my brain needs some time to compute what you're saying
and follow up the linx you suggest.

But already I feel a little more confident in our AMD choice ----- and a
little *less* confident in our chosen vendor :-0

Tony and Goerge have it all together here (as usual). Ditch your "chosen"
ventor. They do not have *your* interests at heart. At best, they're
trying to hide the baloney.

 

[Yousuf Khan]

Thank you all so much for these encouraging comments esp George and Tony.
As a complete novice my brain needs some time to compute what you're saying
and follow up the linx you suggest.

But already I feel a little more confident in our AMD choice ----- and a
little *less* confident in our chosen vendor :-0

I went down to a computer store for the first time in a long time
recently. Went to spec out a system for a cousin of mine, to make sure
he didn't end up picking up the most expensive piece of crap system at
the urging of the salesman. When we got there the salesman tried to
direct us towards a P4 system, which seemed a good deal at first, but it
was really shortchanging you on memory, hard drive, and video. I
immediately put a stop to that, and told him to nix any further talk
about any P4 systems (we were there for value systems only). The guy
smiled knowingly and immediately began behaving. Picked out a basic
Cdn$299 AMD Sempron system, and started optioning it out.

We optioned out a pretty competent system for Cdn$550. Sempron 2200,
Asrock mobo, DVD burner, 512MB DDR, 100GB HD, ATI Radeon 9200 economy
video card. This was cheap enough so that there was enough money left
over in the budget to buy an LCD 17" monitor rather than a CRT monitor.
If we removed the ATI card, and went with only onboard video, that
would've knocked another $80 off of the price.

My cousin was glad I took him, because afterward he said he had no idea
what to ask the salesman. He said he would've probably bought the wrong
system just based on brandnames. You got to put a firm hand on the
salesmen and tell them exactly what you're looking for and not to budge
from your price ranges and features you had in mind.

Yousuf Khan

 

[Johannes H Andersen]

]

Have you visited any Forum for the AMD mbrd in question? Start by looking
at the recommended memory modules for the mbrd at the mfr Web site. Much
of the "problems" I've seen are stories from disappointed overclockers;
there's also a "story" (check the Forum at www.msi.com.tw) that the AMD
memory channel(s) needs slightly looser timings than what is generally
spec'd in the module SPDs... in particular the Trcd, so that instead of say
2-2-2-6, the BIOS Setup needs to be tweaked to 2-3-2-6 for *some* modules
and especially the performance ones like this with such tight timing.
Apparently overclockers get very peeved at having to back off timing on
their high priced "fashion" DIMMs.

I can relate my experience: I've done a couple of Skt 754 systems with the
MSI K8T Neo-FSR and noticed that in the recommended modules list, the
Kingmax ones were consistently mentioned by MSI for several different
mbrds. They're well priced so I gave them a go and found they work fine
with standard SPD timings - checked that two 512MB modules (double sided)
worked together too. I went with the Kingmax PC3500, running at standard
DDR400 in the hope that it'd give a bit of extra headroom, so I can't say
definitely if the Kingmax PC3200s would do as well.

I just did a MSI K8T Neo2-F Skt 939 system with a 3500+ Newcastle and used
the Kingmax PC3500 modules again - two 256MBs this time, double sided.
Those are spec'd at the same 2.5-4-4-8 and they work fine in the Skt 939
dual channel config too. I played a bit more with the timings on this
system to see what'd work and not: turns out that they need 2T command
rate to work, the 1T would pass memtest86+ but caused minor probs in
installing WinXP and would fail within minutes of starting Prime95.
Kingmax does not specify a command rate so don't know what is to be
expected here, nor whether the 2T is due to the modules, mbrd or CPU.

I also tried reducing the other timings... just to see: here I found that
those modules work fine at 2.5-3-3-7-2T with DDR400 clocking. When I say
"work fine", I've really pounded on this system at those settings by
running Prime95 for hours and 3DMark2001se and 3DMark03. I'm convinced
it's as stable as any I've ever seen and it scores ~5GB/s with Sandra's
memory benchmark. Now I suspect that if I loaded up both channels with
DIMMs, I might have to back off to 2.5-4-4-8-2T but I'm not about to try
that any time soon.

I'm just getting ready to do a MSI K8N Neo2 Platinum with a 3500+
Winchester - already received the CPU, memory and some other stuff; still
waiting on the mbrd and case. I decided to try Crucial DIMMs (regular
ones, not the Ballistix) this time, since I've never had trouble with
Crucial memory modules and I've found they are often underspec'd - I want
to see what I can get out of them in the way of timings but will *not*
overclock. I'll report back when it's working... in a couple of days.

As for your "sales person", it's possible he has a better mark-up on Intel
based systems or he may not be clued up on AMD systems in general and just
be parroting the usual FUD. Another thought: socket 939 mbrds are in short
supply right now - the nForce3 ones in particular are just flying out of
the dealers and e-tailers.

Finally, a word of warning: if you install WinXP SP2, either as an SP2
update or streamlined in, on a VIA K8 mbrd, do *NOT* use the VIA AGP driver
- IOW from the VIA 4-in-1 install, refuse the AGP selection. In SP2, M$
has new "unified AGP drivers" for AGP 3.0 and 3.5 compliant chipsets and
you *must* use them to avoid trouble. In particular they have two drivers:
for AGP 3.0 and 3.5, the first specific to K8 systems; funny thing is that
for me, it did not install automatically with the WinXP SP2 install and I
had to go to "Update Driver" for the VIA CPU to AGP Bridge and manually
tell it to use the "VIA AGP driver (Microsoft)" version. This caused it to
install a driver called GAGP30KX.SYS and it seems to work fine - without it
the SBA and Fast Writes don't get enabled; with a AGP 3.5 chipset the
corresponding driver is UAGP35.SYS.

Rgds, George Macdonald

"Just because they're paranoid doesn't mean you're not psychotic" - Who, me??

All sounds a bit complicated. Happy I chose a Northwood P4.

 

[George Macdonald]

]

Have you visited any Forum for the AMD mbrd in question? Start by looking
at the recommended memory modules for the mbrd at the mfr Web site. Much
of the "problems" I've seen are stories from disappointed overclockers;
there's also a "story" (check the Forum at www.msi.com.tw) that the AMD
memory channel(s) needs slightly looser timings than what is generally
spec'd in the module SPDs... in particular the Trcd, so that instead of say
2-2-2-6, the BIOS Setup needs to be tweaked to 2-3-2-6 for *some* modules
and especially the performance ones like this with such tight timing.
Apparently overclockers get very peeved at having to back off timing on
their high priced "fashion" DIMMs.

I can relate my experience: I've done a couple of Skt 754 systems with the
MSI K8T Neo-FSR and noticed that in the recommended modules list, the
Kingmax ones were consistently mentioned by MSI for several different
mbrds. They're well priced so I gave them a go and found they work fine
with standard SPD timings - checked that two 512MB modules (double sided)
worked together too. I went with the Kingmax PC3500, running at standard
DDR400 in the hope that it'd give a bit of extra headroom, so I can't say
definitely if the Kingmax PC3200s would do as well.

I just did a MSI K8T Neo2-F Skt 939 system with a 3500+ Newcastle and used
the Kingmax PC3500 modules again - two 256MBs this time, double sided.
Those are spec'd at the same 2.5-4-4-8 and they work fine in the Skt 939
dual channel config too. I played a bit more with the timings on this
system to see what'd work and not: turns out that they need 2T command
rate to work, the 1T would pass memtest86+ but caused minor probs in
installing WinXP and would fail within minutes of starting Prime95.
Kingmax does not specify a command rate so don't know what is to be
expected here, nor whether the 2T is due to the modules, mbrd or CPU.

I also tried reducing the other timings... just to see: here I found that
those modules work fine at 2.5-3-3-7-2T with DDR400 clocking. When I say
"work fine", I've really pounded on this system at those settings by
running Prime95 for hours and 3DMark2001se and 3DMark03. I'm convinced
it's as stable as any I've ever seen and it scores ~5GB/s with Sandra's
memory benchmark. Now I suspect that if I loaded up both channels with
DIMMs, I might have to back off to 2.5-4-4-8-2T but I'm not about to try
that any time soon.

I'm just getting ready to do a MSI K8N Neo2 Platinum with a 3500+
Winchester - already received the CPU, memory and some other stuff; still
waiting on the mbrd and case. I decided to try Crucial DIMMs (regular
ones, not the Ballistix) this time, since I've never had trouble with
Crucial memory modules and I've found they are often underspec'd - I want
to see what I can get out of them in the way of timings but will *not*
overclock. I'll report back when it's working... in a couple of days.

As for your "sales person", it's possible he has a better mark-up on Intel
based systems or he may not be clued up on AMD systems in general and just
be parroting the usual FUD. Another thought: socket 939 mbrds are in short
supply right now - the nForce3 ones in particular are just flying out of
the dealers and e-tailers.

Finally, a word of warning: if you install WinXP SP2, either as an SP2
update or streamlined in, on a VIA K8 mbrd, do *NOT* use the VIA AGP driver
- IOW from the VIA 4-in-1 install, refuse the AGP selection. In SP2, M$
has new "unified AGP drivers" for AGP 3.0 and 3.5 compliant chipsets and
you *must* use them to avoid trouble. In particular they have two drivers:
for AGP 3.0 and 3.5, the first specific to K8 systems; funny thing is that
for me, it did not install automatically with the WinXP SP2 install and I
had to go to "Update Driver" for the VIA CPU to AGP Bridge and manually
tell it to use the "VIA AGP driver (Microsoft)" version. This caused it to
install a driver called GAGP30KX.SYS and it seems to work fine - without it
the SBA and Fast Writes don't get enabled; with a AGP 3.5 chipset the
corresponding driver is UAGP35.SYS.

All sounds a bit complicated. Happy I chose a Northwood P4.

I don't see what's complicated here!

No more complicated than a similar exercise of self-build for any Intel
system... and certainly not for a Prescott, which also depends on chipset
framework which post dates the OS. You have to choose a memory vendor and
a performance/price target for it either way.

As for unified AGP drivers they're coming for *everybody*... like it or
not. Uhh, you *do* know what "unified" means?? Maybe you haven't gone
through the pain of dealing with SP2 yet... especially on a CPU with DEP?
Apparently Intel has gotten DEP right... now.

Rgds, George Macdonald

"Just because they're paranoid doesn't mean you're not psychotic" - Who, me??

 

[George Macdonald]

All sounds a bit complicated. Happy I chose a Northwood P4.

Hey I just noticed where you've been working - say hello to Gautam for
me.;-)

Rgds, George Macdonald

"Just because they're paranoid doesn't mean you're not psychotic" - Who, me??

 

[ykhan]

I don't see what's complicated here!

Uggh, you guys really oughta consider trimming your quotes when
undergoing your pissing contest.

No more complicated than a similar exercise of self-build for any Intel
system... and certainly not for a Prescott, which also depends on chipset
framework which post dates the OS. You have to choose a memory vendor and
a performance/price target for it either way.

As for unified AGP drivers they're coming for *everybody*... like it or
not. Uhh, you *do* know what "unified" means?? Maybe you haven't gone
through the pain of dealing with SP2 yet... especially on a CPU with DEP?
Apparently Intel has gotten DEP right... now.

If Microsoft is going to be supplying unified AGP drivers from now on,
I wonder if the new PCIe are going to require the same drivers or
different drivers?

Yousuf Khan

 

[George Macdonald]

Uggh, you guys really oughta consider trimming your quotes when
undergoing your pissing contest.

I usually do but in this case I judged it correct to leave it in - it was
only a 2nd level and it was relevant... and it *was* more than a couple of
sentences.;-)

If Microsoft is going to be supplying unified AGP drivers from now on,
I wonder if the new PCIe are going to require the same drivers or
different drivers?

I'm not sure what the story is here; I'd think it's likely a goal that the
same video GPU drivers work, whether through an AGP or PCI-e interface. I
can't imagine that PCI-e supports DIME (as oposed to the "free" AGP specs,
PCI-e is paid mode only :-() given its lack of success, though AGP drivers
and GPUs still seem to implement it to the letter. It's generally
"thought" that DIME has never been used but I wonder... when I see it
mentioned in diagnostic programs as available and with main memory
apparently allocated through the aperture for texturing. Anybody really
know?

BTW AGP cards are drying up somewhat in the supply chain so we need the
nForce4 and K8T890 chipset mbrds rather urgently now. When I ordered my
new system a bit over a week ago, the AGP card I selected went out of stock
during the ordering process.

Rgds, George Macdonald

"Just because they're paranoid doesn't mean you're not psychotic" - Who, me??

 

[Johannes H Andersen]

Hey I just noticed where you've been working - say hello to Gautam for
me.;-)

Hehe. I was a mature postgrad there; Gautam was my supervisor. Then I did
research for a while in the engineering dept, but my CV is long and varied.

Rgds, Johannes