Prev: How can I reinstall my mouse by keyboard?
Next: Very slow booting into windows from bios
From: LeeG on 19 Apr 2010 17:09
I see what you mean about the number crunching but at least you can compare
the gpu and memory speeds to get a rough idea how each card compares.

Regards to the new install - tried it. Nothing seems to work. The display
driver gets stuck in an infinite loop. To me that sounds like it is waiting
for a set of instructions that are not arriving. I had a similar problem
with some VB6 programs. It would get to a certain point and then go into a
continuous loop. In this case the correct parameters were not being set at
start up. inserting a delay at start up allowed time for the parameters to
be set. I'm not saying this is the cause but something is not performing as
it should and I know, for definite, that it is not a hardware problem. (e.g.
PSU, Graphics card, mobo etc.) If it is a display driver problem then it is
evident in a few versions because even earlier drivers do not cure the
problem.

"Paul" wrote:

> LeeG wrote:
> > Hi Paul.
> >
> > I found this site which I think is quite useful as a reference.
> >
> > http://www.gpureview.com/show_cards.php
> >
>
> I use that site quite a bit for comparisons, but I lack the skills
> to convert all the functional unit numbers, into a "card strength".
> Which is why I suggest a benchmark instead, as a safer rating method.
>
> In terms of your original problem, I'd probably take a spare disk
> and install a clean copy of WinXP and test again. I queue up all
> my software first on the disk (on a separate partition), disconnect
> from the Internet, and install the OS and SP3. Then give it a try and see
> if it runs OK or not. (That means I'd be missing all the security
> updates, but it should still work for hardware test purposes.)
>
> I don't see a lot of evidence for PCI Express compatibility issues.
> I think the design concepts used in PCI Express are cleaner than
> AGP. There is no "aperture". The cards look to the OS like they're
> on a PCI bus. The cards still do DMA for fast transfers. Interrupts
> are message based, so there isn't even an interrupt pin on the
> card. And in terms of compliance, there are very few issues with
> PCI Express. The only one that comes to mind, is when the first
> Revision 2 cards came out, and didn't negotiate down to Revision 1
> rates properly. And that was fixed by a video card firmware update.
> Other than that, the history of PCI Express, is amazingly clean.
>
> Any motherboard that has real problems, likely comes with a warning.
> There is at least one VIA chipset motherboard, where the motherboard
> manufacturer gives a list of "compliant" video cards. That is a
> warning that all is not right. But that particular case was
> caused by the video slot not having full x16 wiring. If you
> don't have a "bargain basement" motherboard (like the previous one
> I was using), chances are you'll be OK. My video slot was only
> wired x4, and that seemed to bother a few common video cards
> (according to the compatibility list). Regular motherboards have
> x16 wiring on at least one x16 slot, and then I would not expect
> surprises.
>
> Paul
> .
>
From: Paul on 19 Apr 2010 22:41
LeeG wrote:
> I see what you mean about the number crunching but at least you can compare
> the gpu and memory speeds to get a rough idea how each card compares.
>
> Regards to the new install - tried it. Nothing seems to work. The display
> driver gets stuck in an infinite loop. To me that sounds like it is waiting
> for a set of instructions that are not arriving. I had a similar problem
> with some VB6 programs. It would get to a certain point and then go into a
> continuous loop. In this case the correct parameters were not being set at
> start up. inserting a delay at start up allowed time for the parameters to
> be set. I'm not saying this is the cause but something is not performing as
> it should and I know, for definite, that it is not a hardware problem. (e.g.
> PSU, Graphics card, mobo etc.) If it is a display driver problem then it is
> evident in a few versions because even earlier drivers do not cure the
> problem.
>

Apparently, you can have hours of fun, trying to fix that problem.

http://www.nvnews.net/vbulletin/showthread.php?t=54242&page=1

At one time, when a video card driver crashed or froze, that was it.
The system was effectively dead, and you rebooted.

Then, someone got the clever idea, that if the video card and processor
stopped responding to one another for some period of time, they'd try
to restart the driver (called "VPU recover" ?). What you're seeing, could
be a lack of response from the card, within a certain timeout period,
the driver attempting to recover from the problem, and the problem happening
again immediately. That could account for the looping part of the fault, but
doesn't tell you why it stopped responding in the first place.

Paul
From: LeeG on 20 Apr 2010 03:48
You could well be right. One thing I have noticed, and I am thinking of
asking nVidia about it, according to the specifications provided by nVidia
this card with the GT216 chipset has DDR3 memory at 800Mhz. The two cards I
have dealt with have DDR2 memory on them at 400Mhz. Could it be this chipset
is optimised for DDR3 and having DDR2 on the card is contributing to the
problem?

"Paul" wrote:

> LeeG wrote:
> > I see what you mean about the number crunching but at least you can compare
> > the gpu and memory speeds to get a rough idea how each card compares.
> >
> > Regards to the new install - tried it. Nothing seems to work. The display
> > driver gets stuck in an infinite loop. To me that sounds like it is waiting
> > for a set of instructions that are not arriving. I had a similar problem
> > with some VB6 programs. It would get to a certain point and then go into a
> > continuous loop. In this case the correct parameters were not being set at
> > start up. inserting a delay at start up allowed time for the parameters to
> > be set. I'm not saying this is the cause but something is not performing as
> > it should and I know, for definite, that it is not a hardware problem. (e.g.
> > PSU, Graphics card, mobo etc.) If it is a display driver problem then it is
> > evident in a few versions because even earlier drivers do not cure the
> > problem.
> >
>
> Apparently, you can have hours of fun, trying to fix that problem.
>
> http://www.nvnews.net/vbulletin/showthread.php?t=54242&page=1
>
> At one time, when a video card driver crashed or froze, that was it.
> The system was effectively dead, and you rebooted.
>
> Then, someone got the clever idea, that if the video card and processor
> stopped responding to one another for some period of time, they'd try
> to restart the driver (called "VPU recover" ?). What you're seeing, could
> be a lack of response from the card, within a certain timeout period,
> the driver attempting to recover from the problem, and the problem happening
> again immediately. That could account for the looping part of the fault, but
> doesn't tell you why it stopped responding in the first place.
>
> Paul
> .
>
From: Paul on 20 Apr 2010 04:56
LeeG wrote:
> You could well be right. One thing I have noticed, and I am thinking of
> asking nVidia about it, according to the specifications provided by nVidia
> this card with the GT216 chipset has DDR3 memory at 800Mhz. The two cards I
> have dealt with have DDR2 memory on them at 400Mhz. Could it be this chipset
> is optimised for DDR3 and having DDR2 on the card is contributing to the
> problem?
>

The comments section here, mentions they use both DDR2 and DDR3. The
clock rate on the memories could be a bit different. I think the largest
spread I've ever seen in memory performance, on the same nominal model number
of card, is a factor of 4. That is the difference between the card
with the cheapest slowest memory, and the best memory. There is a
lot of latitude for the manufacturer, as to what they can stick on
a card. It is one thing you have to watch for, when looking though a
set of product offerings.

http://www.gpureview.com/GeForce-GT-220-card-617.html

I don't think the memory type makes that much difference from a
design integrity point of view. Before a video card can be shipped,
the video BIOS file must be modified, to contain the correct memory
timing settings for the memory selected for the card. So the card
is tuned up, before it is shipped. I don't recollect too many cases
where that was done in a clumsy fashion - they usually manage to get
it right.

Your GPU is designed in 40nm techology, and likely has pretty decent
memory I/O speeds on it. When they connect DDR2-800 to it, I doubt that
taxes its abilities at all. The only question that remains in my mind,
is whether the memories used are well tested, before the card ships.
I don't know whether they have a short burn-in process, with
the GPU doing the "memtest86" on the chips. It would make
sense to do it that way. With the power of the GPU, you should be
able to run a memory test pretty rapidly on the card. If the design
wasn't optimal, the dropout on the production line would draw attention
to it fairly rapidly. At our old factory, if bogus cards are coming
off a line, a huge pile of bad cards starts to build up, next to the
test stations. For anyone that cares (the management), they eventually
notice the mess :-) The people doing the testing, hardly ever care
to tell somebody, that a lot of bad stuff is coming off the line,
but the pile of bad cards is a pretty good means of saying "I've got
a problem".

Paul
From: LeeG on 20 Apr 2010 06:32
Fair enough.

With this infinite loop problem being such a long lasting problem, with
several different types of systems suffering, you would have thought that
someone would have found out the cause. There must be some common
denominator, however obscure, that could explain why this happens.

Concerning this card I wish I had access to the reviews I have recently
read. I would not have purchased this card for what I want. For a few
pounds more I could have purchased a far superior card.

One last question - Would you say that 1GB graphics memory on XP SP3 for
average gaming is a bit too much. When I upgrade to Win7 would it be prudent
to still stay with the 1GB or can you go to 512MB with no loss to
performance. At the moment it seems my FSX uses the most graphics memory but
this is below 512MB.

"Paul" wrote:

> LeeG wrote:
> > You could well be right. One thing I have noticed, and I am thinking of
> > asking nVidia about it, according to the specifications provided by nVidia
> > this card with the GT216 chipset has DDR3 memory at 800Mhz. The two cards I
> > have dealt with have DDR2 memory on them at 400Mhz. Could it be this chipset
> > is optimised for DDR3 and having DDR2 on the card is contributing to the
> > problem?
> >
>
> The comments section here, mentions they use both DDR2 and DDR3. The
> clock rate on the memories could be a bit different. I think the largest
> spread I've ever seen in memory performance, on the same nominal model number
> of card, is a factor of 4. That is the difference between the card
> with the cheapest slowest memory, and the best memory. There is a
> lot of latitude for the manufacturer, as to what they can stick on
> a card. It is one thing you have to watch for, when looking though a
> set of product offerings.
>
> http://www.gpureview.com/GeForce-GT-220-card-617.html
>
> I don't think the memory type makes that much difference from a
> design integrity point of view. Before a video card can be shipped,
> the video BIOS file must be modified, to contain the correct memory
> timing settings for the memory selected for the card. So the card
> is tuned up, before it is shipped. I don't recollect too many cases
> where that was done in a clumsy fashion - they usually manage to get
> it right.
>
> Your GPU is designed in 40nm techology, and likely has pretty decent
> memory I/O speeds on it. When they connect DDR2-800 to it, I doubt that
> taxes its abilities at all. The only question that remains in my mind,
> is whether the memories used are well tested, before the card ships.
> I don't know whether they have a short burn-in process, with
> the GPU doing the "memtest86" on the chips. It would make
> sense to do it that way. With the power of the GPU, you should be
> able to run a memory test pretty rapidly on the card. If the design
> wasn't optimal, the dropout on the production line would draw attention
> to it fairly rapidly. At our old factory, if bogus cards are coming
> off a line, a huge pile of bad cards starts to build up, next to the
> test stations. For anyone that cares (the management), they eventually
> notice the mess :-) The people doing the testing, hardly ever care
> to tell somebody, that a lot of bad stuff is coming off the line,
> but the pile of bad cards is a pretty good means of saying "I've got
> a problem".
>
> Paul
> .
>
First  |  Prev  |  Next  |  Last
Pages: 1 2 3
Prev: How can I reinstall my mouse by keyboard?
Next: Very slow booting into windows from bios