emosun wrote:

if you want to run that test again on a board who's results aren't dependent on how well the manufacturer patched in support , I'd be interested in THOSE results.

thats more like it , literally what i wanted to see

now there's another massive flaw but you don't care anyway so why bother pointing it out not like you'll read it

emosun wrote:

now there's another massive flaw but you don't care anyway so why bother pointing it out not like you'll read it

...EOT on my part, thank you.

agent_x007 wrote:

...EOT on my part, thank you.

🤣 yeah that's what i thought. wouldn't want anyone questioning why you used directx 9 or only tried crysis now would ya. one game , one OS , plenty to base an entire opinion on for you i'm sure.

I think for me , i'll stick with the 1,792 sample rate passmark runs to form a better opinion. Especially since the tests are far more in depth than an unoptimized video game from 2007 and actually use multiple dx version and test every aspect of the gpu.

I apologize for the possible derail, but I just wanted to mention that the whole ordeal about HD4670 or HD3850 being the fastest..don't both have their own niche anyway? And besides the vast majority of AGP motherboards that can actually take these cards will end up being too slow for both the HD4670 and the HD3850 anyway.

I do think it's an interesting discussion though 😊
Btw, there are also different kinds of HD3850 and HD4670 cards, my HD4670 having somewhat slower memory compared to which I think was the HD4670 AGP made by HIS?

So lumping all of these cards onto a single stack wouldn't really make a lot of sense I think since both cards perform so similar when compared to well...basically all other AGP cards out there 🤣

Lots of drama in here; feels like modern politics. My line of thought is similar to Tetrium's. 92 fps vs. 61 fps, and 33 fps vs. 26 fps in a game wouldn't matter to me either way. I haven't heard the argument that synthetic benchmarks are more credible than actual game play though. I think a broader range of games and platforms would be of interest though. I normally don't read into these 'more modern' topics, but I happen to use an Asrock board with a lowly ATI HD4350 AGP. Its an Asrock dual sata2 with Opteron 185 (like a dual-core Athlon64 X2 2.6 GHz). This is my most modern desktop. The board was pretty flakey until I recapped it and replaced the HDD with something modern (2TB hybrid drive). I'm not sure if this board falls into your junk category, but runs 24/7 without issue for my needs. I think the processor upgrade card idea was to keep the consumer using their products longer. Interesting concept, but I've never bothered finding the AM2 upgrade card.

I like this one...

Now, if boards, agp versions, cpu's etc matter so much why not look at the difference in cards in pci-e form? And since Techpowerup came up:

https://www.techpowerup.com/reviews/Powercolor/HD_4670/

That is the review of the hd4670 PCI-E, benched in many games on many settings. You can clearly see the strongpoints and weaknesses of both cards. On agp it should be fairly identical?

@emosun
Why would you want to use Windows Vista/7 for AGP GPU testing, when 95% of them can't even run DirectX 10 games with proper speed ?

I tested Quake III Arena and Doom 3 along side 3DMark 99-2006, but I guess you missed that part...
Well, here's table with results for two other games.

Pentium 4 (in Pink) can't handle Quake III Arena or Doom 3 with anything faster than X1950 Pro/7900 GS (it's simply too slow).
Also, in both games ATI drivers have higher CPU load (at least for fastest DX10 cabable models), that's why (with PGA 478), X1950 Pro can win against (in theory) newer and more powerful cousins.

@Tetrium : I own both 4670 and 3850 in IceQ variants (4670 model has 1GB VRAM 1600MHz DDR3) 😀

EDIT :
@meljor
PCI-e version of 4670 has 2000MHz (effective) RAM speed, and even my IceQ has "only" 1600MHz.
On top of that, GPU Core clock is 20MHz higher on PCI-e edition - at least from TPU test (my card has Core Clock of 750MHz.
For HD 3850 situation is inverted. My card is a overclocked version of PCI-e one (670MHz vs. 715MHz on my, and 1600MHz [effective] VRAM clock vs. 1818MHz on my IceQ).

I didn't underclock my 3850 to stock, because getting 4670 AGP with 1GB DDR3 memory isn't easy, and most 4670s for AGP has DDR2 memory instead of (G)DDR3 (which makes them A LOT worse for gaming).

Tetrium wrote:

I apologize for the possible derail, but I just wanted to mention that the whole ordeal about HD4670 or HD3850 being the fastest […]
Show full quote

I apologize for the possible derail, but I just wanted to mention that the whole ordeal about HD4670 or HD3850 being the fastest..don't both have their own niche anyway? And besides the vast majority of AGP motherboards that can actually take these cards will end up being too slow for both the HD4670 and the HD3850 anyway.

I do think it's an interesting discussion though 😊
Btw, there are also different kinds of HD3850 and HD4670 cards, my HD4670 having somewhat slower memory compared to which I think was the HD4670 AGP made by HIS?

So lumping all of these cards onto a single stack wouldn't really make a lot of sense I think since both cards perform so similar when compared to well...basically all other GAP cards out there 🤣

Everyone knows that most of AGP based systems, MB/CPU will be not able to use even an x1950 Pro AGP to its full potential. The thing here is only about the 2 cards, bcuz i have seen many people to talk that HD4670 is the fastest card, while it is not, and the HD3850 is the fastest AGP card, it was about that.
HIS cards seems to be one of the best build out there, also MSI on x1950 Pro is very well made, while Sapphire is lacking things, and the cooler is total trash...

Agent 007 seems you have the best cards from HIS nice, as i see them to be the better ones, too bad i did lost my chance to buy HIS HD4670 AGP...

And this is the MSI card...

feipoa wrote:

Lots of drama in here; feels like modern politics. My line of thought is similar to Tetrium's. 92 fps vs. 61 fps, and 33 fps vs. 26 fps in a game wouldn't matter to me either way. I haven't heard the argument that synthetic benchmarks are more credible than actual game play though. I think a broader range of games and platforms would be of interest though. I normally don't read into these 'more modern' topics, but I happen to use an Asrock board with a lowly ATI HD4350 AGP. Its an Asrock dual sata2 with Opteron 185 (like a dual-core Athlon64 X2 2.6 GHz). This is my most modern desktop. The board was pretty flakey until I recapped it and replaced the HDD with something modern (2TB hybrid drive). I'm not sure if this board falls into your junk category, but runs 24/7 without issue for my needs. I think the processor upgrade card idea was to keep the consumer using their products longer. Interesting concept, but I've never bothered finding the AM2 upgrade card.

Well if for you recapping of all capas on the board mean quality............ Meanwhile i will just buy ABIT based board that have 100% japan capacitors that will not need recap even in the next 20 or so years or even more. Trash brands means when the products does not have
1 quality of components
2 OC capabilities, components
3 BIOS with wide range of settings
4 nice visuals, style, looks
5 more connectors, RAID or other
A good product need to have at least, i repeat at least one of all that point, or even 2 of them, otherwise is just junk that someone did spend resources on it..

Yes ASRock have the upgrade cards thing, and that is the only positive thing on them, but if i will need to recap the whole damn board no thanks i will just blacklist them and send them to hell with their pile of crap..

Radical Vision wrote:

The thing here is only about the 2 cards, bcuz i have seen many people to talk that HD4670 is the fastest card, while it is not, and the HD3850 is the fastest AGP card, it was about that.

I appreciate your sense commitment to the bottom line; I am the same way, especially with older systems. I am curious, was there any benefit of the HD4000 series compared to the HD3000 series, like with video acceleration or compatibility thereof?

If the newest HD4xxx AGP card was HD4850 then was going to be no debates at all, but they did make this stupid thing to make one last card, at least was going to be nice to see HD4850 AGP but seems not..
In terms of HD content like movies and video clips, even old Radeon 9550 is capable of 720p/1080P with proper software and settings (not 100% sure about the 1080P but i think is running smoothly), so that is not interesting, but in terms of games, i think HD3850 have the upper hand, as it uses GDDR3, bigger bus to the memory, also have really powerful GPU as well, no matter the 4670 is improved, it is still only mid range card, not high end there is the big difference..

Seems like you are really big fan of CyriX, nice to see that, as i heard from many ppl to say "CyriX was junk, they was slow, they was trash" well they was slower then AMD, but still they was the third big x86 CPU company out there like 3DFx was in the GPU market, and now they are gone and this is not good for the consumer, but who of most ppl gives a f***...
Im planing to build great machine with all great dead brand of the old time like CyriX for CPU, 3Dfx for 2D and 3D, Quantum for HDD, AOpen or ABIT for MB, some great dead brand foe audio something like Aureal, Compaq keyboard and mice, it will be great dead brand build...

Radical Vision wrote:

Well if for you recapping of all capas on the board mean quality............ Meanwhile i will just buy ABIT based board that have 100% japan capacitors that will not need recap even in the next 20 or so years or even more. Trash brands means when the products does not have

Abit is not immune to bad capacitors. Their top of the line AN8 32x was plagued with bad caps. They made some great boards but their quality was extremely inconsistent. I love my NF7-S 2.0 but I went through three of them before getting one that would run correctly... the third one has run great for 15 years now.

I have an Abit IS7 (top of the line P4 board) that has some bad caps as well... all good brand name caps, but they are failing.

It's a little sad when you present some charts and data for something as silly as pushing older computers to the max and you get responses like " oh you didn't test with X motherboard and Y videogame, that's suspicious".

I really appreciate your data agent_x007, thanks for sharing with the rest of us 😀.

agent_x007 wrote:

@emosun
Why would you want to use Windows Vista/7 for AGP GPU testing

Because they are direct x 10.1 gpu's?

Your "data"

"the 3850 is faster provided you run only crysis in dx9 mode"

is barely even a test at all. it's a short sighted determination based on either laziness , or lack of a copy of windows 7 or ANY of the modern dx10 games that exist from the past few years. (and yea btw there's a handful of more recent dx10 games still)

AT LEAST you ran it on a board other than the asrock frankenstein nightmare. But your limited testing (heck your one test of one game in legacy mode) means quite literally nothing to me , or the 1,792 people who ran a more in depth test than you did.

agent_x007 wrote:

@emosun
Why would you want to use Windows Vista/7 for AGP GPU testing, when 95% of them can't even run DirectX 10 games with proper speed ?

btw i also love that you said this , while literally showing benchmarks of you using the agp system to run Crysis

Honestly I'm done discussing this , no amount of bar graphs can make your flawed testing look in any way legitimate when you cannot even see your own hypocrisy

Crysis uses mostly DX9(c) code, only in Very High quality setting you get some DX10 effects.
However, you want to tell me a OC'ed HD 3850 or HD 4670 can run Crysis in DX10 mode on Windows Vista/7 ?
Here's anadtech test for you : LINK, take a look at what HD 3870 scores. They didn't even used Very High Preset, they only used High Quality Preset and changed Shaders to Very HIgh quality.
Personally, I don't see the point of testing graphics setting you can't enjoy playing at.

Ozzuneoj wrote:

Radical Vision wrote:

Well if for you recapping of all capas on the board mean quality............ Meanwhile i will just buy ABIT based board that have 100% japan capacitors that will not need recap even in the next 20 or so years or even more. Trash brands means when the products does not have

Abit is not immune to bad capacitors. Their top of the line AN8 32x was plagued with bad caps. They made some great boards but their quality was extremely inconsistent. I love my NF7-S 2.0 but I went through three of them before getting one that would run correctly... the third one has run great for 15 years now.

I have an Abit IS7 (top of the line P4 board) that has some bad caps as well... all good brand name caps, but they are failing.

Where i did say they are ? Im talking about the new ABIT boards that are all orange colored like the ABIT NF7 V2 (V2 is the important thing as they did start to use 100% japan caps after that line of boards), AN7 and others, while the old yellow ABIT boards all was with junk caps.. I don`t see how your Abit IS7 can have caps problems if is from the second revision orange color and japan Rubycon ?!?!?

emosun wrote:

btw i also love that you said this , while literally showing benchmarks of you using the agp system to run Crysis

Honestly I'm done discussing this , no amount of bar graphs can make your flawed testing look in any way legitimate when you cannot even see your own hypocrisy

I think he did lost his time to convince people that don`t care at all. Also im sure even Jesus himself to show you the same results you will still be like "nah this is not legitimate"...
But as you can see at least some people appreciate what he did shown us.....
Im wondering what happened to "your asrock board is junk, so your tests are bad as well, or you forgot that ?

@_007 Kudos for sticking to the facts and keep on showing stuff that is relevant for the discussion. Btw i have the cib Club3d hd4670 agp 1gb gddr3 version so i think the 1gb versions aren't that uncommon? Also have the hd3850 agp but mine is the Sapphire. I guess both are normal clocked versions. Best system i can come up with is an Asrock 865PE board with a C2D@3ghz and both cards will be a bit bottlenecked on that.

@Emosun Could you make this less personal and just stick to the facts? What about doing some benchies on your own and show the results? What about commenting on what's given to you like benchmarks on sites like Anandtech or the fact that the pci-e version were also pretty close in the TPU review? Both cards have their strong points but the hd3850 seems to be a bit quicker on the highest settings with AA.

Haha only thing is there was going to be not such war here, if the damn HIS did make an HD4850 AGP but....

meljor wrote:

@_007 Kudos for sticking to the facts and keep on showing stuff that is relevant for the discussion. Btw i have the cib Club3d hd4670 agp 1gb gddr3 version so i think the 1gb versions aren't that uncommon? Also have the hd3850 agp but mine is the Sapphire. I guess both are normal clocked versions. Best system i can come up with is an Asrock 865PE board with a C2D@3ghz and both cards will be a bit bottlenecked on that.

Thank you.
What I meant by "isn't easy to get" is pricing and availability : LINK <= This is my card (1GB DDR3 1600MHz). Who spends that much $ on a AGP GPU ?
I bought my card few years back locally for ~70$ (that's with with shipping).
Does your Club3D have VRAM clocked at 1333MHz ?
EDIT : Here's ASUS version of 4670 1GB AGP, with DDR2 memory : LINK

Lol you really spend 70$ for 4670 damn...
I did find the absolute the same card HIS 4670 AGP, and did cost only 12 or 15 euro no shipping. I did not get it as the f**king seller did make me mad, he was some douchebag, too arrogant, this is the only reason i did not buy the card, but i will find other one, i just hate to buy things from arrogant and nasty people...

this is also how i did lost an IBM Model M gen 3 for only 7 euro, but the seller was some dumb **** gypsy, that did not know he have Model M and how much it cost, but he did want to sell me more things, or no deal at all, he did have like 4-5 more Cherry keyboards, but they was not mechanical, so for me that is equivalent of useless items and junk, damn arrogant people...

Don`t get upset, this guy is trolling i think, on make the whole thing on purpose to make wars...