lo

I can see in the PCI-X version of this card that the
inductors are sourced from `COILCRAFT'

The actual components may be from the DO3316P
range but I cannot say if they are the correct size nor
whether their ratings are a match for your missing
component.

It is however somewhere for you to start looking and
the DO3316P-682_L_ device could be a replacement.

The do3316P product page is here:
http://www.coilcraft.com/do3316p.cfm

bye

Hmmm thanks. That does look promising.

I sent you high resolution images of my Matrox Parhelia AGP 128 and Matrox Parhelia PCI-X 256. Unfortunately, I have the same AGP 2/4/8X model shown on Wiki. Some of the component layout is a little different as well. If you can find another image your exact card which indicates a part number of the missing inductor, that would be best. On my card, not all the inductors are the same.

On my card, the inductor in the place of where your missing inductor is, is labeled LCD104-6R8A 0228. If the 104 is the inductance value code, I think that is a 100 mH inductor. 103 is a 10 mH inductor, and 102 is a 1 mH inductor. At least that is why my LCR meter shows for some spare inductors I have in my bin. I'd see if you can find the exact datasheet though.

Did you specifically not want a Parhelia keyed for 2X as well? I'm surprised that your Parhelia AGP 128 is keyed for 4X/8X only. Is it the 256 MB version? I thought all 128 MB versions were keyed for 2/4/8x.

It's 128MB since it has 8x 16MB DDR SDRAM chips. Apparently this is not only 4x/8x-only, but it is also a bug-fixed revision of the GPU, and it runs at higher clock speeds.

Every photo I find has a inductor labeled with only "682". Part number of the card is PH-A8X128.

http://how-to.wikia.com/wiki/How_to_identify_ … ductor_markings

101 = 10*10^1μH = 100μH
682 = 68*10^2µH = 6.8mH

BUT!!

There is more than this value to inductors. Operating frequency and current, to name two of them.

h-a-l-9000 wrote:

There is more than this value to inductors. Operating frequency and current, to name two of them.

Yes, that is why I would find the datasheet for the actual part used. If that cannot be found, you can try to determine the operating frequency using a scope and a working card at which point you could begin specing out an equivalent replacement.

swaaye wrote:

It's 128MB since it has 8x 16MB DDR SDRAM chips. Apparently this is not only 4x/8x-only, but it is also a bug-fixed revision of the GPU, and it runs at higher clock speeds.

Every photo I find has a inductor labeled with only "682". Part number of the card is PH-A8X128.

The easiest thing to do is to look for an exact part replacement from that number.

How did you determine that you have the bug-fixed revision? Are the 4x/8x-only cards the bug-fixed ones? What bug existed in the non-bug fixed revisions? My card has DDR SGRAM. How much higher clock speed does yours run at compared to the "buggy" version?

One other thought, aside from the missing inductor, have you determined if any other components on the PCB are faulty or missing?

feipoa wrote:

How did you determine that you have the bug-fixed revision? Are the 4x/8x-only cards the bug-fixed ones? What bug existed in the non-bug fixed revisions? My card has DDR SGRAM. How much higher clock speed does yours run at compared to the "buggy" version?

One other thought, aside from the missing inductor, have you determined if any other components on the PCB are faulty or missing?

Apparently the AGP 4x/8x-only cards indicate the newer chip because these cards were released sometime in 2004. I haven't been able to find thorough documentation of what was fixed (and it probably doesn't exist) but people have noted performance and image quality improvements. Clocks should be 250 MHz core 300 MHz RAM but apparently the chip has some headroom and people have been able to break 300 MHz on the core.

There is this big MURC forum thread with info, but also a lot of supposition.
http://www.murc.ws/showthread.php?t=47782&highlight=ph-a128r

I have no idea if other parts of this card are broken. The only thing missing from what I can tell is that one inductor. Two other inductors appear to have been reattached. I do wonder how these inductors were broken off in the first place...

This looks similar to the one you posted a pic of: http://www.ebay.com/itm/75-pcs-COOPER-UP1B-6R … =item20b72e75b9

Category Inductors, Coils, Chokes Family Fixed Inductance 6.8µH Tolerance ±20% Series Uni-Pac Package / Case 0.350" L x 0.240" W […]
Show full quote

Category
Inductors, Coils, Chokes
Family
Fixed
Inductance
6.8µH
Tolerance
±20%
Series
Uni-Pac
Package / Case
0.350" L x 0.240" W x 0.196" H (8.89mm x 6.10mm x 5.00mm)
Packaging
Tape & Reel (TR)
Type
Ferrite Core
Current
1.7A
Mounting Type
Surface Mount
Q @ Freq
-
Self Resonant Freq
-
DC Resistance (DCR)
89.70 mOhm Max

Seller also does smaller lot upon request...blah...blah 🤣

hi

it is probable that a reasonably high current
transfer is required. even as much as 5A
as the device is most likely within a circuit of
this type:

http://en.wikipedia.org/wiki/Buck_converter

bye!

Swaaye, what did you end up doing with this card?

I've decided to not bother with repairing it. It's not worth the time and effort, particularly when buying another would only be a bit more than $20 probably, and that I'm not really interested in using Parhelia after experimenting with the other one. Too noisy, buggy and slow.

What were the major buggy drawbacks you noticed? Is it noisy because the GPU fan on your unit is a bit worn out? I don't recall my card being very loud. Will you be able to add any 3D video captures using this card? I think you'll need to use W2K+ though.

I did post videos already and it's mentioned in the video thread.

The fans on both of my Parhelia cards are equally noisy but one does have some ball bearing grind to it when it starts. These cards are louder than G400 Max. It is a fairly common noise level for cards of the day though.

Do you have any plans to add this card (and others you've recently tested) to the comparison chart in 3D Accelerators Compared?

feipoa wrote:

Do you have any plans to add this card (and others you've recently tested) to the comparison chart in 3D Accelerators Compared?

Some of the cards are already on there. However some of the cards won't run 3dmark99 or Quake2 properly (or at all). Some, like Parhelia, 8500 and 5200 are too fast and would just show a CPU bottleneck in 3dmark99 and Quake2. Geforce 256 was as fast as I wanted to go there.

There might be a few that are slow enough and functional enough. Maybe I will look into the card pile again.....

Perhaps for the higher-end graphics cards, you could create a seperate chart using a P4, or faster? Alternately, if you want continuity between the cards, what about running them all, including the slower ones, on a P4, or faster? Just a thought.

Instead of me toiling away for days, I suggest this extremely impressive effort from ixbt. 😀
http://ixbtlabs.com/articles2/over2k4/index.html

There are some interesting results in there. It appears that in a few games the Parhelia's architecture performs somewhat competitively.

http://www.tomshardware.com/reviews/32-graphi … ltdown,127.html
Another nice roundup, this time from 1999 and with older cards.

It's historically interesting in an extra way because they used video card vendor drivers instead of the GPU manufacturer reference drivers. Video card vendors tended to neglect updates. So you see large differences between different TNT2 cards for example.

Did you notice any bugs with the Parhelia 128 in Unreal Tournament?

According to the fantastic work done by the guys at ixbt, the Matrox Parhelia 128 had about the performance of an ATI Radeon 9500, GeForce FX5600, and an S3 DeltaChrome S8 in Unreal Tournament (1024x768).

Unfortunately the GeForce 6200 was a bit too new to make this comparison. It would be nice to see where the budget GeForce 6200 made its way in there. I've settled upon the GeForce 6200 for my dual 1.4 GHz Tualatin and dual 850 MHz Coppermine machines.