CT-7NJL6 w/ barton 3000+ only running at 1300

Bill, it appears you've got your cpu stable and at the correct speed
using FSB 333. I'd maintain that setting for now, but I mean as to
speed, not particular settings in the bios. So pay attention as to
how things change as you attempt to bump up the Memory bus.
You will likely have to set as expert to get adjustment of memory
frequency. Since auto isn't cutting it, or SPD, I'd bet you got some
misbadged RAM from ULTRA. Mine appears to be in the opposite
direction of yours in that the SPD says 400 when in reality a sticker
on it says 266.

Once you have the expert setting using 166/333 for the cpu, go to the
memory part and set your memory to a percentage of the FSB. You might
start with the percentage that puts you at that 100(200) you are
currently running it at. Should work without any hitch. This as a
confidence builder that you are approaching it correctly.
Then go back in and start bumping the percentage to get higher
standard memory speeds. I say this because I bumped mine to one of
the odd percentages yielding maybe 186 or something odd. System would
not boot.
Give it some time in windows to see if it is stable. Maybe run some
tests.
Once you hit that unstable point, I'd double check those part numbers
on the RAM chips. Do a search and find out what they are really
supposed to be. Could be they were pushed too far or maybe they
weren't up to par and intended to be run slower.
Also see my link to overclockers in previous post about RAM. Once you
hit that unstable point use utilities to try to locate which stick
might be problem or swap out/rotate sticks so you are running 2 or
maybe even 1, could be two bad sticks. Goal is to get up to the speed
it is supposed to be. But also could be current draw issue as pointed
out in link as I noted you said you were using a 350W PS. If you can
find data on sticks or RAM chips try to calculate the total required
power and check what the PS has available on the 5V line. Other parts
will also draw power, but I have no idea how much. Can only tell you
my PS is rated at max 32A on +5V line.
 >> Stay informed about: CT-7NJL6 w/ barton 3000+ only running at 1300 

I found some of those 2900 Barton's on ebay. Considering buying one.
Had a near major meltdown over the last few days. USB keyboard died.
USB2 hub lost bus-powered mode. System restore wouldn't work.
Something caused rewritting system.ini with garbage.
I was trying different timings of RAM to push it to FSB speed of 333.
Nothing I tried gave good stability.

Did the memory swap per mobo help?
 >> Stay informed about: CT-7NJL6 w/ barton 3000+ only running at 1300 

Nthums,

I did not try the memory swap, yet. Both kids had school projects and
these two systems were in use practically all weekend. I plan to
swap the chips this weekend.

The 2900 chip, with the Nforce2 chipset and PC 2700 memory, provides a
quality platform. This configuration is quite stable with an XFX
GeForce 5200 / 128 MB. I can play Halo and Civ IV on it with minimal
slowdowns, but I was not building a gaming box anyway....

In retrospect, buying the 2900 as a good decision and I am very
satisfied.

I noticed the supply of XP chips is drying up, and the cost increased
dramatically.
 >> Stay informed about: CT-7NJL6 w/ barton 3000+ only running at 1300 

Hi Guy,
Definitely correct on some chips disappearing, in fact all faster
socket A. Only 3K & 3.2K Mhz processors had 400 FSB per AMD.
These 2900 are some oddity, but if they work at that speed, a couple
of hundred MHz won't be missed much.
That is why I was wondering if you swapped that RAM and found out if
it was stable with RAM running at FSB of 400. It would be a
confidence booster and I'd definitely try one of these 2900 Bartons
at the price is $85 plus shipping. Way cheaper than the insane prices
on the 3200. I've seen used being bid into the $135-$175 range and
some "as is". New are over $200.
If it works out, that processor and some dual channel would max out
capabilities as far as what socket A has available. Plenty of slots
left to try other stuff, so only short coming for some time would be
SATA300, Firewire2, and Gigabit lan. Jumping into 64 bit with PCI-e
seems cost prohibitive at this point because that means buying
everything new, except ram.
 >> Stay informed about: CT-7NJL6 w/ barton 3000+ only running at 1300 

I agree with the cost assessment. back in September, Uddarts referred
me to the 754 Sempron chips. Before I shell out for a Socket A 3200,
I would strongly consider upgrading to a 754 system. I thought the
Sempron was just a Celeron-type version of the Athlon 64, but after
reading several reviews it seems to be a resaonable step up from
Socket A, despite the need for a new board. Of course, we already
see the end of the line for 754 as well....

I am taking Friday off, so if possible I will swap chips and send you
the assessment.
 >> Stay informed about: CT-7NJL6 w/ barton 3000+ only running at 1300 

delete
 >> Stay informed about: CT-7NJL6 w/ barton 3000+ only running at 1300 

I really haven't caught on to what the advantages of 754 are. It seems
maybe that dual channel is capable without having to have the matched
sticks? And faster bus speeds for RAM? Does that sound right? Fill me
in if you have time. I hadn't heard that 754 was coming to an end, but
something about that 939 socket coming to an end. There seems to be a
maximum processor speed they are bumping up against and hence the
dual core processors. And AMD intends to be releasing a quad in the
near future as the next step up and think that is where the 939 is
supposed to disappear.
Definitely getting hard to plan the building of a system and maybe opt
for a few less pricier parts hoping to upgrade as prices fall. That is
where I got caught with the Sempron thinking an AthlonXP down the road
as price drops. Only I got surprised. Just checked pricewatch.com.
Only four vendors left selling the 3200 socket A. Prices start at
$290.

The 2900 I mentioned, this guy has like 90 of them. They had been
posted for 4 days and only 4 buyers. I think I'm going to jump so I
have time to test and get another if I get a bad one before he sells
out. Pretty darned sure he will.

Please send me what you see as advantages of 754 & 939. What
little I knew, it seemed like the 939 might be better and there are a
few mobos that have the PCI-e with those cpu's. Although most mobo
vendors are now pushing the Sempron with Socket A, and it looks like
that cpu will be the alternative as 754 & 939 processors dry up,
it might be wise to rethink the part priority. In the past, it was
buy the mobo and then upgrade cpu and ram, video, etc. I would guess
that there is going to be a surplus of some top notch mobo's because
of the shortage of Athlons. In other words, buy as much processor as
you can afford and then work your way up the mobo ladder.
Unfortunately it will probably put a number of MOBO manufacturers
under as they get stuck with these surpluses.

I've been watching the prices on HDD's and the sudden upturn tells me
the glut of PATA drives is near an end. Sadly I've seen no real price
cuts on SATA. Circuit city doesn't even stock any yet. I'm not sold on
the bios thing. I think I mentioned I tried one from ebios and saw no
improvements. They claimed I needed xp to really see the advantage,
so maybe after I go to xp, I'll try again. (30 day love or leave) no
reinstall, except audio drivers.
 >> Stay informed about: CT-7NJL6 w/ barton 3000+ only running at 1300 

I do not believe a socket A Sempron is equivalent to any Athlon XP
chip. Since the supply of Athlons is exhausted, Socket A is the new
budget competition for the Celeron, and the socket 754 becomes the de
facto mid-range architecture for hobbyists and manufacturers. I cannot
speak details on the 754's advantages and disadvantages, but my gut
tells me it was an incremental improvement on the 462 to accommodate
the 64-bit architecture and reach that magic number (3.0 GHz). The
PCI-E and PCI-Express changes help to dramatically increase the 754's
capabilities over the 462, but the memory architecture is almost
unchanged from socket A. AMD is marketing 754 Semprons that are NOT
equivalent to a Celeron, but instead compete with mid range Pentiums
a full GHz higher.

Unless AMD finds a few million Athlons somewhere, the current stock of
socket A boards, equipped with Sempron 2500s are suitable for small
businesses, granma's email box and firewalls, and those of us who
just like building them. 754 is the ubiquitous AMD standard,
competing with stock Pentiums, while the 939's compete with the EE's
for high-end business.

This last few years has been pretty Darwinian; at least as bad as the
transition from 286 to 486. I have not followed the various socket
iterations, except to know that both AMD and Intel were trying to
keep up with the supposed expectations of Moore's Law -- to double
processor speed every two years. (That is not really what Moore's Law
states, but when is the public perception of anything ever right?)

I think that both mfrs have finally run into a hard, inviolate, law of
physics, and that is how small a silicon pipe can be before the
electrons stop moving freely.

If I had to guess, I would bet that in five years, we will see massive
multicore chips for the home PC market. I imagine there are mechanical
engineers at AMD and Intel right now figuring out how to develop a
vertical-core CPU with 4 - 16 cores and heatpipes interlaced for
cooling.

But for the present, I viewed the 754 and 775 as interim steps. It
shocks me to hear that AMD may abondon 939, and I expected something
beyond 775 from Intel. To me, this implies two radically different
strategic thrusts -- Intel to max out the 775, and AMD to come up
with yet another socket optimized for multi-core architecture. IMHO,
abandoning 939 in the mid-term is a mistake for AMD, but it implies
that their engineers are gaining confidence and that they plan to set
the standard for the next architecture, instead of reverse-engineering
another Intel design.

I think both companies stay with 64-bit architectures through 2015 and
rely on multiple cores to carry the fight.
 >> Stay informed about: CT-7NJL6 w/ barton 3000+ only running at 1300 

I have to wonder how pressing it will be to upgrade to 64 bit . I
haven't heard anything about xp-64 bit yet, as to is it worth it.

If I had to bet, I'd say the road block to going beyond 64 bit is OS,
and software. As you pointed out multi-cores could get large and I'm
sure they could easily parallell several of those cores to accomplish
128 bit or something similar to the 128 bit memory channel.

I saw a show where they have made processors in the lab working on
photons. Why they haven't made the jump to production is a mystery.
Likely cost for one. They'd certainly use less power and dissapate
less heat.

These rapid changes, such as dead ending so many sockets, will have to
put a strain on parts of the industry. And of concern will it cause
some loss of market share. I recently heard AMD may have surpassed
Intel, and brings some sense as to why there isn't a tremendous price
difference anymore.
Backward engineer? I can't point to any specific stories of AMD doing
this although there is one about Compaq and how it came into being.
Darwinian, with a healthy dose of natural selection. When AMD started
with the technological change allowing rating and true speed, Intel
probably fell off their chairs laughing when they first heard of it.
Laughing probably ceased quickly and they stupidly pushed foward with
RDRAM. AMD not wanting to pay exhorbitant prices for it and the
technology and that many RAM manufacturers were left out, they joined
forces and came up with DDR. Intel certainly came to that late. And
I'd bet that was the problem with one series of Nvidia cards. I saw
an interview with owner, he admitted they had memory problems and had
jumped ahead of technological capabilites. To get the product out the
door it sounded like they had to slow the GPU so the memory design
could keep up and were abandoning that design in the next series
release. This was at a time when Nvidia was in bed with MS, Intel.
This strongly suggests that architecture was similar to RDRAM.
Nvidia since jumped from that ship and ATI is back on, which is where
it was when Nvidia took the lead. And may have taken the lead again.
This pattern seems to suggest a fundamental problem with that
relationship of partnering. Now it is true that AMD has had that
relationship with MS but I bet they did it in a limited way, just to
ensure their processors were able to work with MS code.
A short time ago, I thought DDR was the next to die on the
evolutionary plan, to be replaced with DDR2. But the last time I
checked, a few weeks ago, only Intel platforms use it. I wish I knew
more about the differences, is it really better or was this some way
for Intel getting around paying fees to the developers of DDR, which
includes AMD. And are they hoping they reap profits from AMD with
DDR2 technology?
BTW, it should be interesting to see how much cache will be available
with these multicore's. Overcoming heat problems, we might see 32M of
cache in the next few years, which seems wild when thinking of how
much RAM was considered necessary to run W95, first 32bit. Of course
we could go way back before that. I still have a machine laying
around here that was considered a Lamborghini in its day. Very
impressive was it's boot time, it didn't use a hard drive and it's OS
booted from RAM. A whopping 384K if I remember correctly.
Unfortunately, it didn't get the added RAM which would make it
capable of running Wordperfect DOS.

Back on track for this MOBO, I got a couple of blue screens last
night. Pretty sure it is software issue and I'm off to see what I can
learn.
 >> Stay informed about: CT-7NJL6 w/ barton 3000+ only running at 1300 

Nthums:

I did the memory swap; I took the PNY 7ns PC2700 RAM (2X512) out of
the 2900 system and swapped it for the Viking 8ns PC 3200 RAM (2X512)
in the 3000/333 system.

It worked, but the performance drop on my 3000 was enough that I
swapped them back.

I retained a CAS latency of 2.5 on the overclocked 2900 (CPU frequency
overclocked to 200), but the 3000 system still would not stabilize
with CL 2.5, so I had to leave it at 3.
 >> Stay informed about: CT-7NJL6 w/ barton 3000+ only running at 1300 

Not related to this mobo, but just for info.
Guy, you mentioned interest in 754 and I as well relating to AMD64 and
also socket 939. I had observed that only Intel is using DDR2. I also
have been interested in PCI-e as possibly the future. I did some
searching and despite the myriad of boards with PCI-e listed at
pricewatch, few AMD mobos have them and only those with NForce4
chipset. Maybe, when the newer AMD cpu's with the new sockets show
up, more mobo's will have the PCI-e. And then maybe not. It would
appear that DDR2 is Intel's answer to the RDRAM mistake and not
having to pay fees to the AMD conglomeration that came up with DDR.
And maybe the reverse is true of PCI-e and AMD will come up with it's
own bussing, in cooperation with card makers, to avoid paying Intel.
Could be some drastic changes in near future with AMD putting memory
controller on processor, normally a function of chipset.
And Nvidia, normally GPU people, combining this with their NForce
chipsets which can come with lots of integrated functions, including
high end audio.
Somewhere in the future, we might see a Nvidia/AMD type merger putting
everything into one chip.

"Hello HAL" (Space Odyssey 2001)
 >> Stay informed about: CT-7NJL6 w/ barton 3000+ only running at 1300 

Hi Guy, thanks for trying.
Dang system lost video sync again as I was finishing a very lengthy
reply on this subject. Monitor goes black with square and error
showing Vertical out of range at 249Hz. Since frequency also jumps
around after hibernate but not this high, have to wonder is it card
or OS. Need to update drivers for start.

It not working out seems really weird and unexpected. I have to ask if
you went into the bios after the swap and tried auto or spd settings
and noted the timings produced. On the later bios, it still displays
the timings but in a shade indicating they are not individually
adjustable until you go into expert mode.

It seems unreasonable that the PC2700 running at 333 CAS 2.5 which you
had on FSB400 would not function as well when moved to a FSB333?
BTW, from what I've read that 2900 is actually a 400FSB, not
overclocked, so maybe something with the bios that it doesn't
automatically properly detect or the way it was derated from the
original 3200 it was making it hard for the bios to determine what it
is.
The only thing that makes sense with it not working is what others
were saying about taking a performance hit when not running FSB and
RAM bus at 1:1. Sort of like what I was told is called 'slack' when
trying to run a USB2 PCI add-in vs onboard controller. I was told
over at the USB-IF (implementers forum) that timing differences
between the controller and PCI bus introduce extra wait states to
stay in sync. So, when you have that ram set at 333 on the 400FSB
system, despite your settings, the cpu is negotiating(creating) wait
states by trying to stay synced. That would be plausible when others
are saying you take a performance hit if running the RAM faster than
the FSB. Their statement isn't very clear, that is do you still
actually see a slight boost running it 400 vs 333 on a 333 FSB, just
not the expected percentage, or does it actually slow down?

Again sorry to hear it did not make you or your XP3000 happy.
Logically, it would have seemed that the overall throughput of system
A + B synced with RAM would have been greater than XP2900(400FSB) +
XP3000(333FSB) unsynced to RAM.
 >> Stay informed about: CT-7NJL6 w/ barton 3000+ only running at 1300 

one of you ought to try setting your fsb to 1:1 166/166 and from the freq control area inch the fsb up 170, 175, etc to see if you can get better preformance out of the system. once you start losing stability start working with the timing to get a boost that way.


ud
 >> Stay informed about: CT-7NJL6 w/ barton 3000+ only running at 1300