Lost the auction because I was out getting a haircut, my car windows retinted and inspections done. You know...real life. No big loss. The listing was "Untested" anyway. I COULD get the X850xt but for $80 (new). I also see someone just put up a Sapphire X850XT for only $20 - used but tested working, but I still think that would be overkill for this system, even with a Barton core.

EDIT: Oh, nevermind. That $20 is for a PCI-Express version. No wonder it's cheaper - stupid AGP price gouging...

Found another one - it's a Radeon x850xt (NOT PE version) AGP for $25+shipping. It's a Sapphire brand. Good card? Buy it now? Yes? No? Same price as the auction would have been if no one else had bid and it's more than double the performance of the 9800xt, not accounting for a CPU bottleneck (think Barton 3000+ upgrade)

Seems like a fair price to me. If it is working of course.

I can't tell from experience about that particular graphics card, but the Sapphires I've used so far tended to be quite good.

Listing says it was working when it was pulled from an old system a year prior. No accessories, cables, disks, nothing, but I don't need those anyway. I only plan to hook up a monitor and I can download any needed drivers from the web. The OTHER one I found is a fully-new original-in-box all-accessories-included x850xt Platinum Edition (basically the same card with a 20 MHz overclock on the GPU and a 100 MHz overclock on the memory), for $80 and free shipping. Prices are in USA$.

With second hand stuff it is always a gamble and I think everyone here who has bought lots of parts that way, have gotten burned sooner or later.

The trick is to make the gamble worth it (my apologies, I can hardly be any more vague I guess..).

And another trick is to see if the seller has anymore items for sale that you might be interested in (or at least somewhat), but are very cheap or just plain useful. That way you can combine shipping and minimize the risk that you get a package of something that will end up in the bin. Also minimizes that let-down feel if you get a package and at least a part of the contents is in working condition.

Btw (fyi), the X850 series are powered mostly by 12v instead of the 5v of the 9?00 series, which can make picking a PSU a lot easier.

In the end it's your money, so we can't decide for you blah blah we all know the drill 😊

Good luck!

Rhuwyn wrote:

You could also find an unlocked Barton and simply run it the same clockspeed as a 3200+ or even possibly higher.

For that, he needs to get an early Barton.
If the manufacturing date after 43 calender week of 2003, it shall be locked.

@FFXIhealer
Try to get an Barton with manufacturing date < 43 calender week of 2003.

Example picture:
This 2800+ is from 27 week , year 2003

KT7AGuy wrote:

FireFox 28 was the last version that would work with P3 and Athlon XP systems.

https://ftp.mozilla.org/pub/firefox/releases/28.0/

Not quite. Firefox 48 was the last version that would work with PIII and A-XP systems. It was released back in...November, I think? So you'd still be fairly up to date.

My PIII-S auto-updated to 48.0.2 when I had it connected to the net a few days ago.

Standard Def Steve wrote:

KT7AGuy wrote:

FireFox 28 was the last version that would work with P3 and Athlon XP systems.

https://ftp.mozilla.org/pub/firefox/releases/28.0/

Not quite. Firefox 48 was the last version that would work with PIII and A-XP systems. It was released back in...November, I think? So you'd still be fairly up to date.

My PIII-S auto-updated to 48.0.2 when I had it connected to the net a few days ago.

Yep, you're right.

https://www.mozilla.org/en-US/firefox/48.0/releasenotes/

"After version 48, SSE2 CPU extensions are going to be required on Windows"

I don't know what I was thinking. For some reason, I thought that v28 was the last one that worked with CPUs that only support SSE.

I haven't used Firefox as a serious browser in a long time. I remember installing it on that P3 with WinXP a few years ago and v28 is what worked at the time. I've been a Pale Moon user for years now, but it hasn't played nice with SSE-only systems since v3.

Edit: Now I remember what I was thinking. Firefox v28 is the last version without Australis. That's why I chose it for my P3 WinXP system.

whats that? even wiki does not tell too much...

Opinions on "All-in-Wonder versions, please. I've found a number of 9800 Pro AiW AGP cards cheaper on the bay. If it has the same gaming performance as the non AiW, might just get one of hose and be done. Looking up the core specs of the x800xt and it's too close to my 7800GTX for my tastes.

FFXIhealer wrote:

Looking up the core specs of the x800xt and it's too close to my 7800GTX for my tastes.

Well, X850 XT PE (AGP) is between 7600 GT (AGP) and 7900 GS (AGP) with SM1.X/SM2.0 performance.
NV auto-wins in SM3.0 since Radeons from X8xx series doesn't support it.

PS. Oh, I should mention above statement is based on results I got with Pentium Dual-Core E5700 @ ~4GHz CPU...

Yeah, but the question was, how do you feel about All-in-Wonder versions of the Radeon cards as opposed to straight graphics cards?

The 9800 Pro has a DVI-I, S-Video, and D-SUB VGA output. The AiW version only has the DVI-V. The rest of the ports include a mutli-plug input, a multi-plug output, and an RF Cable input for TV capturing. Not that it matters anymore, because we stopped analog cable here in America a year or two ago and it's all digital now. Might be nice I guess if I hooked up an old school NES or SNES for gaming captures...but I'll never do that.

All I want is the 9800's triple performance improvement over the shitty 9550 that's in the system now. I mean, the 9550 is God-awful. I'm only getting 1,000 Million operations per second. The 9800 Pro is supposed to bump that to 3,040 MOps/sec, even with half the VRAM. It doubles the pixel shader and vertex shaders in the GPU as well as quadrupling the memory bandwidth from 64-bit to 256-bit.

I've also heard in some places that the AiW were clocked just a little bit lower in order to reduce temperatures due to all the other crap that was thrown onto the card. I dunno if that's true, but it's a concern. I'm trying to hit that 3GOps/sec mark as noted above.

For comparision, the X800XL hits 8 GOps/sec with 16 pixel shaders and 6 vertex shaders (same configuration as the 6800 Ultra @ 6,400 GOps/sec). And the 7800 GTX I have has 24 pixel shaders, 8 vertex shaders, and hits 10,320 GOps/sec.

Hmmm... I guess I could expand my options to the X800 GT, which also has 8:6:8:8 (like the 9800 Pro) and hits 3.4 GOps/sec. That would be around the same target area.

*pasting because I don't wanna close my tab again accidentally and have to retype all of this*

All-in-Wonder are OK, but they are bound to be a bit hotter since there is more things on PCB.

Are those "GOps/s" you write about simple multiplications (ROP/TMU and Pixel/Vertex by clocks) ?
Because those are really not good way to do this...

Theory :
Radeon X850 XT PE (540/1180) [AGP] vs. GeForce 6800 Ultra (400/1100) [AGP] :
They both have 16PS's, 6VS's, 16TMU's and 16ROP's, so by Your math :
x850 XT PE should be better by at least VRAM speed difference (~7%), and by GPU clock difference in best case scenario (that's 35%).

Guess what ?
It's never that easy 😀

Reality (with few additional GPU's) :

LGA 775 system is 4GHz Pentium Dual-Core E5700 (platform SS : LINK).
PGA 478 is Pentium 4 Extreme Edition 3,2GHz OC'ed to 3,5GHz (platform SS : LINK).

I really don't care what the numbers themselves mean individually. I'm only using them to COMPARE one card to the other in terms of how much processing power is available VS each other. You know GPU MHz speeds aren't a good comparison. I mean, just look at Haswell CPUs versus Sandy Bridge CPUs. Clock-for-clock, the Haswell does a hell of a lot more at the exact same clock speed.

Hmmmm. Love to see how all my cards compare in terms of Pixels/sec fill-rate.
MOps/s is Million Operations per second. MPix/s is Million Pixels per second. I notice by the time we get to the GTX 2xx series and higher, they stop reporting MOps and start using FLOPS - probably a much better way to compare GPU horsepower than straight instructions.

ATI Rage IIc - 60 MOps/s - 60 MPix/s
TNT2 - 250 MOps/s - 250 MPix/s (4.17x increase over the Rage IIc)
Radeon 7500 - 580 MOps/s - 580 MPix/s (2.32x the TNT2)
Radeon 9550 - 1,000 MOps/s - 1,000 MPix/s (1.72x the 7500)
Radeon X800 Pro - 5,700 MOps/s - 7,600 MPix/s (looking at on E-bay, 7.6x increase over the 9550)
Geforce 6800 Ultra - 6,400 MOps/s - 6,400 MPix/s (0.84x increase, actually less than x800 - slight increase in MOps)
GeForce 7800 GTX - 10,320 MOps/s - 6,880 MPix/s (0.9x the x800 Pro, still less, but about double the Ops/s)
GeForce GTX 480 - 1,344.96 GFLOPS - 33.60 GPix/s (4.9x the 7800GTX - crazy powerful)
GeForce GTX 980ti - 5,632 GFLOPS - 96 GPix/s (2.86x the GTX 480 - now it's getting ridiculous)

I threw in the 6800 Ultra because 1. I had it in my laptop first and 2. it's very close to the X800 so I can compare between nVidia and ATI for its time. The 6800 Ultra had Shader 3.0, as you guys pointed out, so it's a better graphics card I think. I mean, there's a reason that after I got the 7800 GTX replacement in my laptop, I've used nothing but nVidia since - but I wonder. I've been reading lately about some dirty shut-out practices by nVidia on game developers. Let me see if I can put the YouTube video about it... https://www.youtube.com/watch?v=ZcF36_qMd8M
That video made me want to get an AMD card for the next system I build for a friend....that and the new Ryzen intrigues me.

Take a look at that X800 Pro I linked above - it looks dirty and I've sent a question to the seller asking if that's dust/dirt or corrosion. He claimed on the listing that it works. If it's dirt, I can clean it thoroughly with alcohol and a toothbrush, but if it's corrosion, I'd rather not waste my money. But THAT listing would let me get this card AND the Barton CPU for roughly $30, a pretty sweet price point. Then I can test everything. It would drastically increase the gaming performance of this system even with the AGP 4x limitation, without actually stepping all on my Dell XPS laptop (7800GTX). This might actually be the one, I'm thinking... if it works. And as a bonus, the X800 should run Doom 3 at Medium or High fairly well (I don't care about going over 60FPS - it is a retro system after all)

And another thought - when I picked up my box copies of Half-Life 2...NOWHERE on the box does it say STEAM, but I'm pretty sure there's no such thing as HL2 without Steam...so weird. Any thoughts on that too?

agent_x007 wrote:

All-in-Wonder are OK, but they are bound to be a bit hotter since there is more things on PCB. […]
Show full quote

All-in-Wonder are OK, but they are bound to be a bit hotter since there is more things on PCB.

Are those "GOps/s" you write about simple multiplications (ROP/TMU and Pixel/Vertex by clocks) ?
Because those are really not good way to do this...

Theory :
Radeon X850 XT PE (540/1180) [AGP] vs. GeForce 6800 Ultra (400/1100) [AGP] :
They both have 16PS's, 6VS's, 16TMU's and 16ROP's, so by Your math :
x850 XT PE should be better by at least VRAM speed difference (~7%), and by GPU clock difference in best case scenario (that's 35%).

Guess what ?
It's never that easy 😀

Reality (with few additional GPU's) :

Quake III Arena (PGA 478).png

Doom3 (PGA 478).png

Crysis Low (PGA 478).png

LGA 775 system is 4GHz Pentium Dual-Core E5700 (platform SS : LINK).
PGA 478 is Pentium Extreme Edition 3,2GHz OC'ed to 3,5GHz (platform SS : LINK).

Nice charts. Looks like on those systems, everything UP TO the 9800xt cards were GPU limited and the CPU increase in performance did nothing, but then as the cards got a lot faster, the difference in CPU performance really starts to show between the (relatively) crappy Pentium 4 and the much nicer Pentium Dual-core. That right there is the reason I stay away from Pentium 4 systems. They almost universally run hot and are very inefficient with a stupid-long instruction pipeline. No wonder AMD Athlon chips beat the crap out of P4s clock-for-clock back in the day. And no wonder my laptop's Pentium M (runs at 2.1 GHz and runs circles around a 3.4GHz or faster P4).

@up Thanks, I have every result in screen shot form if anyone is interested 😀

I think "Ops" term (like FLOPS), is oversimplification to the point of it being 99% meaningless.
Why ?
Because those numbers have very little in common with real performance you actually see playing games or using programs.
Here's another chart (done on PDC 4GHz) :

^That is the reason why I think you should rely more on comparisons made in actual games you want to play.

As to Athlon vs. Pentium 4.
That Pentium 4 from my charts is A LOT more powerfull than Palomino you got or Barton you may upgrade in future.
Athlon 64... that's a story for another topic 😉

PS. FLOPS become THE horsepower measurement standard for GPU's since DirectX 10 and 8800 GTX (because both ATI and Nvidia used unified shader units in this generation of cards and 8800 GTX [from Tesla lineup] was marketed as compute card for scientist and they wanted to know FLOPS not "Operations").

agent_x007 wrote:

All-in-Wonder are OK, but they are bound to be a bit hotter since there is more things on PCB. […]
Show full quote

All-in-Wonder are OK, but they are bound to be a bit hotter since there is more things on PCB.

Are those "GOps/s" you write about simple multiplications (ROP/TMU and Pixel/Vertex by clocks) ?
Because those are really not good way to do this...

Theory :
Radeon X850 XT PE (540/1180) [AGP] vs. GeForce 6800 Ultra (400/1100) [AGP] :
They both have 16PS's, 6VS's, 16TMU's and 16ROP's, so by Your math :
x850 XT PE should be better by at least VRAM speed difference (~7%), and by GPU clock difference in best case scenario (that's 35%).

Guess what ?
It's never that easy 😀

Reality (with few additional GPU's) :

Quake III Arena (PGA 478).png

Doom3 (PGA 478).png

Crysis Low (PGA 478).png

LGA 775 system is 4GHz Pentium Dual-Core E5700 (platform SS : LINK).
PGA 478 is Pentium 4 Extreme Edition 3,2GHz OC'ed to 3,5GHz (platform SS : LINK).

those are interesting numbers, but I would mention that, OpenGL and id games were an Nvidia specialty at that time, and that I'm not sure that Crysis low is running the same on an" SM3.0" card and an "SM2.0" card and could be a reason for the difference (no reason for the 9800XT to be this close to the 6800U if that's not the case)

Doom 3 is a Nvidia game - 100% true.

As for Crysis - belive me, I know it should score more - but it just won't.
Maybe it's SM2.0/SM3.0 but 7000 series are not affected as much by this (and all NV cards utilise the same driver).

9800 XT :

6800 Ultra :

I redo this test with A LOT more powerfull CPU :

Still "no go" : 22,36FPS (Avg.)
Score for lower resolution (1280:1024) on 6800 Ultra : 37,37FPS

My "best guess" for what's going on here is either a crappy Pixel Shader (since it was redone in G70), or ROP/TMU bottleneck somewhere (maybe bad use of memory bandwidth on NV40 part ?).
In short : Radeon 9800 XT is simply more efficient with code that Crysis uses.

Well....thank the Lord I'm not trying to run Crysis. 🤣