The HD3650 is made on a 55nm process, so it is a small die with modest power requirements compared to older high-end cards. It's not the fastest AGP card around, but I would say its about as good as a GeForce 7600 for DirectX 9 games.

Well depend on what brand you are looking at. For example the Sapphire cards from Radeon x1xxx series up to HD3xxx the cooler was total joke and scrub, i prefer every time other brands for the cooler. For example i have

That cards

https://imgur.com/a/3aZU1

Don`t mind the PCI ex cards they are only 2 of them all rest is AGP or regular PCI.

As you can see i have cards with superior cooling like ASUS HD3650 with that big cooler, HIS Radeon HD3850, MSI Radeon x1950 Pro. Only thing i can say from all of the 3 cards the ASUS is really poor, no matter it have copper heat pipe and big heavy cooler, still something is not right on that card, as i did have before Sapphire HD3650 AGP and the cooler was total joke, the fan is so thick (+ was broken and not spinning at all) the heatsink itself is some trash quality "metal" that have really bad feel when i touch it like plastic or so...
Now the interesting part is when i use the Sapphire 3650 and ASUS 3650 both have about the same temperatures in IDE they get hot at 60c while doing noting 🤣, most retarded thing was i did expect that ASUS cooler to be able to handle the card well without fans, but seems this is not the case.

This tests indicate the idle temperature of Sapphire HD 3650 AGP, one without fan and one with 80mm fan at not sure what speed was maybe 2-3K rpm. Take note the temperature did rise up after my shots up to 62c, also that temperature is identical to the ASUS HD3650 AGP as well....

Now about branding and cooler of AGP cards well i did have before also Sapphire x1950 Pro AGP and Sapphire HD3850 AGP, well the cooler was total joke i did sell them, after that i did find my precious MSI x1950 Pro AGP and HIS HD3850 AGP well they have great cooler on them, the cards never overheat no matter what, only - on HIS is the damn fan profile, as the fan is set waaay too low and the card gets hot outside windows, as the RPM is about maybe 1000 or even under that, but when i turn on the fan at more speed the card gets cold..

About why some cards, mobos and other have small heatsinks, or lack them, is the cost, as you can see the manufacturers have to lower the cost as they produce tons and tons of hardware and one simple thing like putting bad capacitors instead of japan ones, or using small heatsink and fan can result in really big impact on the manufacturing cost...

No you are wrong i believe Radeon and overall ATi cards since many many years was the one that did have superior image quality and better colors compared to nVIDIA cards. Also there is something very important the Radeon cards are all HD capable, that means even old cards like Radeon 9550 can run HD content like 1080P movies with the proper software and settings, while nVIDIA dies when you run HD content, nVIDIA did start to have HD support on the GEforce 8 series with nVIDIA Pure video, but that is far from the last AGP cards on GEforce 7 series.

I can play 30-40GB movies with the right codec, players and settings, will this on even GEforce 7950GT AGP will be impossible.....

user33331 wrote:

Hello Geforce's like:4800, 5900, 6800 are very large AGP cards... In 2007 ATI launched HD3650, HD4670 and such latest and last A […]
Show full quote

Hello
Geforce's like:4800, 5900, 6800 are very large AGP cards...
In 2007 ATI launched HD3650, HD4670 and such latest and last AGP cards.

So why ATI HD3650s like:Club 3D HD3650 is so tiny in size and equipped with a small unimpressive cooler ? Is it true that 2007 AGP HD3650 even when so tiny is better and more powerful than all those huge pre 2005 AGP Geforces 4,5,6,7 ?

Is there a poorer image quality in HD3650 ? It cost as little as 80 usd$ back in 2007 when it was new. When other big Geforces cost like 300-500$ 2 years earlier. What I'm missing here is the HD3650 a poor quality card or a good choice ?

Note: All cards are AGP-version not PCI-E.

The 4800, 5900 and 6800 were high-end graphics cards. The HD3650 and HD4670 were mid-range graphics cards, the HD3850 and HD4850 are substantially larger.
If you want tiny, try something like a Vanta.

user33331 wrote:

Hello Geforce's like:4800, 5900, 6800 are very large AGP cards... In 2007 ATI launched HD3650, HD4670 and such latest and last A […]
Show full quote

Hello
Geforce's like:4800, 5900, 6800 are very large AGP cards...
In 2007 ATI launched HD3650, HD4670 and such latest and last AGP cards.

So why ATI HD3650s like:Club 3D HD3650 is so tiny in size and equipped with a small unimpressive cooler ? Is it true that 2007 AGP HD3650 even when so tiny is better and more powerful than all those huge pre 2005 AGP Geforces 4,5,6,7 ?

Is there a poorer image quality in HD3650 ? It cost as little as 80 usd$ back in 2007 when it was new. When other big Geforces cost like 300-500$ 2 years earlier. What I'm missing here is the HD3650 a poor quality card or a good choice ?

Note: All cards are AGP-version not PCI-E.

this is like asking why new smart phones are faster than 10 year old desktops

the last agp cards to exist from ati are faster than pretty much everything agp before them , the 4670 being the fastest possible agp gpu.

the 4670 agp is simply a standard midrange 4670 pci-e card adapted to agp , midrange cards are usually always smaller than high end cards.

there was so much advancement between the old direct x 9 cards to direct x 10.1 that yes , midrange direct x 10.1 gpu's are smaller and outperform a lot of old direct x 9 gpu's.

Ahhhh the damn myth that Radeon HD4670 is faster then HD3850, but it is not, so many, many, many people believe in that and they are all wrong...
Can`t find the article where a guy was testing both AGP versions of that cards, and there was at least 10-20 FPS difference in favor of HD3850, but can`t find the site..
Also HD3850 is in the high mid range, while the 4670 is in the midrange, and no matter it is newer architecture it lacks raw power that 3850 have...

Radical Vision wrote:

Ahhhh the damn myth that Radeon HD4670 is faster then HD3850, but it is not, so many, many, many people believe in that and they are all wrong...

Ahhhh the damn myth that Radeon HD3850 is faster then HD4670, but it is not, so many, many, many people believe in that and they are all wrong...

There did me simply typing that make it true?

keeping in mind the 4670 agp is 1gb and most 3850 agps were only 512mb and that the 4670's gpu is faster , no , the 3850 wouldn't be faster.

the 3850 has faster and larger bandwidths in some areas of the design , but it's inefficiency is why it's overall performance isn't very good compared to the next generation of gpu's.

This is why the 3870x2 (the fastest 3000 series gpu that i actually own) isn't very good compared to the fastest 4000 series gpu's.

The 3850 can push more pixels at high res due to having twice the ROPs. The 4670 is dramatically faster with AA enabled and pretty much everywhere else.

Well at least i did find the link, take a good time reading this, as misleading and not knowing the things is never good..........

https://www.techpowerup.com/forums/threads/fa … -4#post-3533375

Radical Vision wrote:

Well at least i did find the link, take a good time reading this, as misleading and not knowing the things is never good..........

https://www.techpowerup.com/forums/threads/fa … -4#post-3533375

he was running it on an absolutely terrible asrock 4 core dual sata 2

one of the junkiest motherboards asrock ever made with more gpu based bios patches than any other board thanks to the terrible via chipset.

you couldn't have picked a more terrible board to conduct that test on

OH , btw , i owned and used this motherboard as my main rig for over 4 years. I wouldn't be surprised if the fast patchwork bios influenced those results heavily.

it's short sighted and bad judgement to use such a bad board based on only having an agp slot and support for a higher end c2d. those results have zero credibility as far as an owner of that board is concerned.

So @emosun, my question to You is this :
Can you beat My scores on bad ASRock board with QX9770 @ 3,84GHz while you use this "proper AGP board" ?

Here's easier test, Crysis with Low preset + Full HD resolution :

EDIT : Hello 😀
I'm glad I could help.

Hey nice to see you here as well agent_x007, im using your post from TPU every single time to show ppl they are wrong and HD3850 AGP is the true king of AGP cards and best out there...
Seems many, many, many ppl don`t know HD4670 is only mid range GPU and is not so powerful, while HD3850 is the card before HD3870 os is about the most powerful HD3xxx card, while HD4670 is far from the most powerful HD4xxx...
As you are one of the few ppl know that, will be interesting to disqus other things, maybe will be interesting for us..

emosun wrote:

he was running it on an absolutely terrible asrock 4 core dual sata 2 […]
Show full quote

Radical Vision wrote:

Well at least i did find the link, take a good time reading this, as misleading and not knowing the things is never good..........

https://www.techpowerup.com/forums/threads/fa … -4#post-3533375

he was running it on an absolutely terrible asrock 4 core dual sata 2

one of the junkiest motherboards asrock ever made with more gpu based bios patches than any other board thanks to the terrible via chipset.

you couldn't have picked a more terrible board to conduct that test on

OH , btw , i owned and used this motherboard as my main rig for over 4 years. I wouldn't be surprised if the fast patchwork bios influenced those results heavily.

it's short sighted and bad judgement to use such a bad board based on only having an agp slot and support for a higher end c2d. those results have zero credibility as far as an owner of that board is concerned.

What the hell man, how did "HD4670 is so much better then HD3850" become to your asrock board sucks so bad..........

Edit

Agent 007 i think you can also make a whole thread about this thing, as many people don`t know and think the HD4670 is the best, while is not the best is second, but is the last ever made AGP card..

I could do it in English, since I already have a thread about most stuff I own (AGP) on one of the Polish forum sites : LINK 😉

Well as me and you can see so many ppl don`t know the thing about 4670 and 3850, so i think you can copy-paste the things here, it will help to educate ppl.

Radical Vision wrote:

What the hell man, how did "HD4670 is so much better then HD3850" become to your asrock board sucks so bad..........

When you run a test comparing gpu's on a board that couldn't even support the 2000 series without the entire bios being patched only months after release it throws away all credibility to the test.

sorry but the board is a bad board , and the test only shows just how bad that 4coredualsata really was when they kept patching that via chipset over and over and over again just to keep it alive.

agent_x007 wrote:

Can you beat My scores on bad ASRock board with QX9770 @ 3,84GHz while you use this "proper AGP board" ?

No because I threw my asrock 4coredualsata in the garbage in 2010 , good riddance.

Not that I'd really want to , because if i have to scrape the bottom of the barrel that is slapped together early asrock boards just to have an upper end c2d with an agp slot , no way.

I did my time and suffered enough with that board.

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.

till then don't claim something as a fact based on bad testing methodology. it's just sloppy.

emosun wrote:

No because I threw my asrock 4coredualsata in the garbage in 2010 , good riddance. […]
Show full quote

agent_x007 wrote:

Can you beat My scores on bad ASRock board with QX9770 @ 3,84GHz while you use this "proper AGP board" ?

No because I threw my asrock 4coredualsata in the garbage in 2010 , good riddance.

Not that I'd really want to , because if i have to scrape the bottom of the barrel that is slapped together early asrock boards just to have an upper end c2d with an agp slot , no way.

I did my time and suffered enough with that board.

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.

till then don't claim something as a fact based on bad testing methodology. it's just sloppy.

I didn't meant for you to test ANY ASRock board.
I only wanted for You to find Me a NON-ASRock board with AGP support that has the chance of beating my setup.

I don't care how bad ASRock board are to you.
I'm interested in facts and numbers, so - show me board that can beat mine (in games/benchmarks).
Plain and simple. Links to 3DMark scores are in my thread on Polish forum I mentioned earlier.

Just FYI : I used latest BIOS available for my tests.

Ok i know ASrock sucks as brand i don`t like them as well, i avoid that brand, but the thing here was about the x2 cards

emosun wrote:

the last agp cards to exist from ati are faster than pretty much everything agp before them , the 4670 being the fastest possible agp gpu.

This is the thing im talking here about the cards, you did as well, now you started about asrock been trash, well many enthusiasts know asrock was total trash back in the days and today is not the same trash but is still bad, but you changed the subject and also you did call this guy have "those results have zero credibility" bcuz ehat, bcuz he show that the so called fastest AGP ever is not the fastest...

Well the thing on old ASRock boards is that thy was trash, and bad build quality, but they did have something unique none other brand did, that was features and compatibility, like upgrade path from K7 to K8 via only daughter board, later from K8 to K9 with the same, or the Conroe type of board like agent 007 did use for his tests, that is unique, but still i don`t like that brand..

Request to Moderator/Admin of this forum :
I would like to ask for splitting discussion on Asrock boards and performance of fastest AGP cards from here to new thread (to not derail it any further).

Thank you.

neither one of you are putting two and two together..... I know you aren't because you aren't bothering to read.

what's the point of even continuing if you won't actually read anything i type.

I could literally type why your testing methodology is flawed , why YOUR setup shows one card being faster than the other, and you'd skim it.

No point in trying to "discuss" a dang thing when for you it's a topic of pride rather than actual results.

so go ahead and keep using your flawed data and don't bother asking why it's flawed becuase you won't read it even if i took the time to type it.

Fine, riddle me this then - Crysis "High" 1280x1024 :

Why OC'ed Pentium 4 Extreme Edition on Gigabyte PGA 478 board "feels" the same as ASRock about 3850 vs. 4670 ?