Best video card for DFI K6bv+/66

[radman3d2]

I have been trying to figure out what is the best AGP video card for this MB that will be complient with it. I know it is old and all that, but it is my #2 machine and I would like to upgrade it to a better video card. Any ideas?

[Direct1]

Without knowing your system specs and what you use the PC for, I like the 9000 Pro 128MB (http://www.newegg.com/app/ViewProductDesc.asp?description=14-102-245&depa=0&section=2). Good luck! :t

[iceblue]

What video card does it currently use?

[Peter M]

You need something that is still backward compatible to 3.3V signalling for the AGP 1x/2x modes your board supports; and you need to aim low on the card's power consumption.

With the K6-2 not having too much rendering power anyway, I'd say you're well served with a Radeon 9200SE or a plain 9200, depending on your budget and connectivity wishes.

[radman3d2]

Hi Peter, Thanks for the reply. I do not know if you remember me or not, it was about 3 years ago you recommended this MB and I would like to thank you for that, it has served me very well over the years. I will look into the ATI 9200 and see what the reviews say about it. Do you know where I could read more about that video card? System specs are below, sorry I didn't include them before.

AMD K6III+ 450@600 2.2v

DFI K6BV3+/66 Rev. B+ 1MB

256 MB PC133 CL2

Voodoo3 2000@157Mhz

SB512

Seagate 15GB 7200 RPM ATA66

Maxtor 4GB 5400 RPM ATA66

[Peter M]

Well, there are plenty of reviews of the 9200 all over the 'net. The 9200SE is a stripdown, with half as many RAM chips on and a slower core clock. This is a budget solution ($50 and below), while the 9200 is for you if you want your CPU to be the bottleneck.

Oh, and thanks for the flowers ... good to hear that my advice made sense ;)

[iceblue]

Budget-Showdown (http://www.hardocp.com/article.html?art=NDkw)

[Peter M]

Well tested, there. But given the 5200's quite a bit higher power consumption, and also given that with a lowly K6-III processor the graphics card won't be the bottleneck, I'd still put a Radeon in there.

[iceblue]

The 5200 non-ultra has no additional molex connector and most likely does not draw that much power, so he should be fine.

[Peter M]

That just indicates it fits into AGP 2.0 power budget. AGP 1.0, as seen on this old socket-7 baby AT form factor board, has a significantly lower limit.

Given that in midrange and budget cards, the Radeon chips have HALF the power consumption of the GeForces - at comparable performance - putting a 9200 in is much less of a gamble. Get a 9200pro if you're after a tad more performance.

[radman3d2]

I like the looks of the FX 5200, but for the life of me I cannot find the system recommendations. ATI gives the system recommendations right up front. It lists my AGP port voltage (3.3) and multiplier (1x/2x). I would have to say I think I would have very little troubles with the ATI, but if I could find something that shows the system recommendations for the FX 5200 I would like to get it. I e-mailed ATI about the system recommendations for that video card and so we will see if they answer.

[iceblue]

Here are specifications of the Chaintech A-FX20 (http://www.newegg.com/app/ViewProductDesc.asp?description=14-145-042&depa=0&section=2). As you can see, 1X/2X/4X/8X bus speeds are supported.

edit: these (http://www.asus.com/products/vga/v9520m/overview.htm) do too, so i don't think you will run into any problems.

[Peter M]

I can only repeat: Signalling voltage and transfer mode is only ONE criteria. Power consumption is the other, and noone's going to tell you on their datasheet.

So if we go Newegg, this is your minimum solution:

http://www.newegg.com/app/ViewProductDesc.asp?description=14-131-226&catalog=48&manufactory=BROWSE&depa=0

9000pro (same as 9200pro, only w/o 8x mode support)

http://www.newegg.com/app/ViewProductDesc.asp?description=14-131-189&catalog=48&manufactory=BROWSE&depa=0

btw, the pictured GF5200 is _not_ using 128-bit memory. See the large picture - half of the RAM chips are not present. This one is going to stink just as badly as the 9200SE, performance-wise - only without being as cheap and quiet as the latter.

[bob05]

Or you could not worry about AGP at all and get the 5200 PCI like I did: http://www.newegg.com/app/ViewProductDesc.asp?description=14-145-055&catalog=48&manufactory=BROWSE&depa=0
:t

[Peter M]

... which also uses only 64-bit wide RAM, and is even lower clocked than the AGP version (core/RAM is 250/333 rather than 250/400) - once more because of the power budget, which is just as low on PCI as it is on AGP 1.0. Combine that with the much lower throughput on PCI, and you'll see what the performance is going to be like.

Have we now collected enough evidence about why NOT to use a GF5200 in this machine here?

[radman3d2]

I know that I do not want to go PCI when I can use AGP. The ATI 9200 does look pretty good from everything I have read so far. It is only a bit below the NVidia over all, but if the NVidia causes lock-ups due to drawing to much power, what fun is that. Not to mention what else it could fry pulling so much power. I think I would need NVidia to tell me straight out that it well work in my machine with no problems before I would buy it. Peter, my hope is that I will be able to play BF 1942 with it. What do you think? System specs for that game are P3 500, 64 Meg of Ram, and a video card with 32 Meg of Ram. With this card I would beat the specs, but not by much.

[iceblue]

People have ran their 5200s at 2X speeds and without any "power consumption" problems.

Btw, it has 128-bit memory, not 64-bit.

[Peter M]

Look at the pictures, my friend. The Chaintech GF5200 cards offered at newegg are 64-bit.

Besides, there are plenty of more recent boards that still are 1x/2x-only but AGP 2.0 compliant, providing more power to the slot. This DFI board here isn't. I've said that before.

Acknowledge the facts before claiming I'm wrong!

mushroom mushroom

[iceblue]

Ok, just checked both the PCI and AGP Chaintech 5200s a 2nd time. AGP one says 128-bit right in the title and in the Detailed Specs. Then there's the PCI Chaintech 5200. It too says 128-bit right in the title and in the Detailed Specs. :t

Please note : due to manufacturer version updates, pictures may vary from the actual item. Newer versions of this item may be shipped before the picture shown is updated.

[Peter M]

*sigh*

Just as well, there's a "specifications may vary" disclaimer, so the text may be right or wrong too. Chaintech's web site confirms there are 64-bit versions of this card, in fact, a total of four versions regarding the RAM alone - 64- or 128-bit, 332 or 400 MHz. Would you take the risk of a random draw?

Besides, for the 3rd and last time, the GF cards consume too much power for this old mainboard. So why bother droning on and on about their RAM bandwidth? This or that way, they are not suitable for this task here.

[radman3d2]

I have looked around on the net and I cannot find a 9200pro, but I can find a 9000pro w/128 of ram. So the 9000pro would work too? I like the 9000 because it has DVI out, plus a fan on the cpu. Thank you both for all your help so far.
NVidia returned my e-mail, but they did not answer my direct question about power consumption. They basically only told me it is backwards compatible.

[Peter M]

The 9000 and 9200 chips are identical in performance and features - the 9200 just has added 8x AGP support which essentially does nothing, but looks nice on the box.

So the 9000pro you found will be just as fine. Consider using a fanless non-pro 9000 or 9200 card though - your slow CPU won't be able to use the extra performance from the pro version, so why deal with the extra noise and heat?

[iceblue]

If a "specifications may vary" disclaimer even exists, there can't be a "actual products recieved may differ" disclaimer or else there wouldn't be many owners of the card. If it's 64-bit, it'll say 64-bit. There's nothing to lie about. Many people are actual owners of the card. They know what they got.

[Peter M]

Hardly. Less than one percent of customers are even capable of figuring this out. The vast majority won't ever notice.
newegg really would be the first shop ever to not have a "typographic and other errors in product specification" weasel disclaimer in their terms and conditions.
Sure, if you actually get a 64-bit card from them you can complain and return, and my bet is they're going to be friendly - but why take the fuzz?

And listen, no matter whether it's 64 or 128 bit, it's no use for solving this thread. Put it to rest already.

[iceblue]

I seriously doubt it's that low. When it turns out to be an actual 64-bit card, the rating and reviews reflect it. So yes, the vast majority does notice.

[Peter M]

You just can't stand not having the last word on that, can you?

[lptech]

radman3d2-

Peter is correct to point out about the AGP slot's inability to provide enough power which would lead to various problems besides the lockup problem. I've seen this happen first hand with videocards that draws power beyond the capability of the AGP slot.

I know this to be true because I have one machine similarly spec like yours with 2MB cache on the mobo and same CPU although not overclocked. It is also a DFI brand mobo like yours. So yes, Peter is correct to point out the "Achilles Heel" of the mobo design and that his advice should be taken as a logical advice given his vast experience in such matters.

LPTECH

[iceblue]

Well isn't that odd. Other people managed to do it. Guess it's a DFI thing.

[lptech]

iceblue-

It is actually a problem with some of the Super Socket 7 mobo design. When they first came out, it was one of the very first thing that became obvious when people tried really high end videocards only to find out about the lockups and reboot problems since there was inadequate current/voltage provided to the AGP slot. You can research this from the Toms Hardware Guide and you will see in their past articles that this was one of the problems with some of the initial Super Socket 7 mobos!

LPTECH

[iceblue]

That was very informative. However, adding 0.1V to AGP bus in system BIOS may solve the problem. It is also possible that a BIOS update would fix it, if your motherboard ran into the incompatibility. In addition, some chipsets had incomplete support by Windows which would also cause the system to lockup.

[Peter M]

Wake up to reality.

It's not a DFI thing. As I said waaaaaay further up this thread, this power budget rise is the most important change from AGP specification 1.0 to 2.0. The other one being the introduction of 4x mode as an _optional_ feature.

Also, baby AT form factor boards cannot take the 3.3V supply from the PSU directly because BAT PSUs don't have that. So consequently, the 3.3V for the AGP slot come from an onboard voltage regulator - and of course this regulator is designed to meet the specification that was current at the time the thing was made.

Voltages programmable from BIOS? Hello? We're talking 1998 standard technology here, not 2003's overclocking toys.
Besides, even if it were there, raising the voltage makes current limitations even worse, because you'll raise the power draw even further. Brilliant move.

Things became a lot better when ATX form factor gained speed - although plenty of initial ATX board designs still didn't use the 3.3V input (idiots!), it soon got better. Not much later, AGP 2.0 raised the required 3.3V power supply strength substantially, and all was calm since ... until NVidia and ATi chose to bypass the problem altogether by putting power plugs onto the graphics cards - rather than using the extra power rails provided by the AGPpro 50/110 specifications that were introduced FOR EXACTLY THAT PURPOSE.

So, with an old AGP 1.0 compliant board LIKE THIS ONE HERE, you either choose your card carefully, or you put a Radeon 9800XT in it and pass the blame onto your power supply unit.

[Peter M]

lptech, just for the sake of history completeness: It wasn't like the SS7 boards had too weak a power supply, it was more like the then-popular Riva TNT2 cards exceeded the allowed power draw, particularly the Ultra ones.
AGP 2.0 specification followed soon after, but NVidia got a tad ahead of themselves there, much to the annoyance of many many users. I've been in this business for long enough to remember that mess very very well.

[iceblue]

Originally posted by Peter M
Voltages programmable from BIOS? Hello? We're talking 1998 standard technology here, not 2003's overclocking toys.
Besides, even if it were there, raising the voltage makes current limitations even worse, because you'll raise the power draw even further. Brilliant move.


Yep. Technical support pages (from that year) said just that. If you don't think it's such a good idea, argue with them. I'm sure the manufacturer knows a lot less than you. LOL.

[Peter M]

Did you ever have hands on this particular board, the DFI K6BV3+/66? No? Well, I did. No such function there.

Now stop trying to be the smartass and start acknowledging the facts.

[iceblue]

That's real mature of you tossing remarks like that. I won't even try to bring myself down to your level.

[Peter M]

Then how about raising yourself to the level of acknowledging the technical facts spread out before you?

As I said before, all you're about is having the last word, and you've already become rather desperate about it about five posts ago. You're wasting my time as well as the original poster's.

[iceblue]

Desperate? Not even. You shouldn't be saying that when you're in denial of even the most apparent facts. You seem to have the time to bounce on every comment on this thread, so i'm rather surprised you're concerned about your time.

[Peter M]

Not. Enough. Power. To. The. AGP. Slot.

Hence: No GeForce here. No matter how much missionary work you put into it.

End.

Everyone else got the message first time - the remaining five iterations of me explaining and you coming back "but he can use a GeForce too" were ENTIRELY redundant.

What's so difficult? What facts did you bring? Befoire you quote DFI's "raise the AGP voltage" bit again, go read the mainboard's manual to see for your self that this trick doesn't apply here. It's in four languages, there should be one you understand amongst them.

http://www.dfi.com.tw/Upload/Manual/41610020.PDF

If you think the manual doesn't describe the product, I'm sure you want to go argue with DFI. Come on, you KNOW you can't be wrong :rolleyes:

And once again, even if it were there, raising the voltage makes the current climb further, worsening the stress on the voltage regulator. Exactly how does that keep it from being overstressed? That's basic EE, you even tried to argue that for the sake of arguing.

To me it all boils down to a rabid NVidia fanboy trying to prevent an ATi purchase no matter what.

Any other "most apparent facts"?

[lptech]

Peter M-

I concur with you on the facts that you've presented. Having been on the manufacturing end of things, I've seen many such anomalies as you've described on the SS7 mobos as well as the early ATX mobos and you're correct about the Riva TNT cards, they did draw an extraordinary amount of power that made them flaky (operationwise).

Those early AGP slots just didn't have the ooomph that they're supposed to, henceforth the AGP Pro slot was next new thing that came about from this oversight. Also, as you pointed out about the Power Supply Units not having support for the 3.3volt output on the AT and the earlier ATX versions is correct. It seems that during that period in time that the PC industry was still vague on formalizing their standards or did not know which direction to move forward to. We live and learn as the saying goes! Thanks again for your input.

LPTECH