Sign in with
Sign up | Sign in
Your question

GeForce 8500 GT 8500 against ATI Radeon™ HD 2600 Pro

Last response: in Graphics & Displays
Share
September 9, 2007 10:21:02 AM

Hello!!

My videocard : GV-RX26P512H - 100$ has gone

Powered by ATI Radeon™ HD 2600 Pro GPU

Supports PCI Express and 120 stream processing units
Microsoft DirectX 10 and OpenGL 2.0 support
Integrated with the 512MB GDDR2 memory and 128-bit memory interface
Supports native CrossFire, Avivo™ HD video and displady technology
Features dual DVI-I / D-sub (by adapter) / HDTV /
Supports Dual Dual-link @ high resolution up to dual 2560x1600
HDMI and 5.1 surround audio(by optional adapter)

http://www.giga-byte.com.tw/Product...me=GV-RX26P512H

For the same money my dealer had following:

GV-NX85T512HP -109$
1. Powered by GeForce 8500 GT GPU
2. Supports PCI Express
3. Microsoft DirectX 10 and OpenGL 2.0 support
4. Integrated with the industry's best 512MB GDDR2 memory and 128-bit memory interface
5. Supports SLI and PureVideo technology
6. Features dual DVI-I / D-sub / HDTV
7. Supports Dual-link @ high resolution up to 2560x1600

http://www.giga-byte.com.tw/Product...?ProductID=2574

Any opinions??? Should I wait till the dealer gets the previous card HD 2600 Pro .
GeForce or ATI (8500 or 2600), both are 128bit and 512Mb.

I'm not a gamer.
Motherboard GA-P35-DS3L , E6550, 1Gb

September 9, 2007 10:39:11 AM

get 8600gt minimum for modern gaming
Related resources
September 9, 2007 11:07:12 AM

I would recommend this. http://www.zipzoomfly.com/jsp/ProductDetail.jsp?Product...
It is $10 more, but it has free shipping. It is also DDR3 and has 256 bit bus, makes up for having half the ram.

If your motherboard has an Intel chip set then go with the ATI, if not go with the NVidia. (You should have an Intel chip set)

2nd Edit:
Just looking and saw this, http://www.zipzoomfly.com/jsp/ProductDetail.jsp?Product... if you want to go with the 8500, this is a cheaper, and better one. http://www.zipzoomfly.com/jsp/ProductDetail.jsp?Product... (Same with 2600 Pro)
September 9, 2007 12:13:20 PM

if u have to choose between 8500gt and 2600pro, then get a 2600pro as the performance is nearly double.
September 9, 2007 1:15:34 PM

The 8500GT has got to be the most gimped "mainstream" video card ever released by Nvidia. If those are your only two options, you must go with the ATI card. I'm not saying the 2600 is your best option... but if it's 8500 or 2600... well, the choice is obvious.
September 9, 2007 1:46:47 PM

Thank You, guys!

I have to go with my dealer.
So it's 2600?
September 9, 2007 6:57:39 PM

Some corrections (I've spoke with my dealer)

1. Palit GeForce 8600GT Super – 149$
• Memory: 512MB DDR3 Memory Support
• Memory Interface: 128 bit
2.Palit Radeon HD 2600XT Super – 139$
• Memory Support: 512MB GDDR3
• Memory Interface: 128bit
or remain with
3.GV-RX26P512H – 99$
Integrated with the 512MB GDDR2 memory and 128-bit memory
interface

Please advice, I'm not a gamer.
Actually NVidia or ATI
a c 164 U Graphics card
September 9, 2007 7:35:59 PM

If you truly play no games, then get the cheapest one. (in this case, the third option.) If you like to play the occasional game, then you should get the 2600XT. It's faster then the 8600GT.
a c 130 U Graphics card
September 9, 2007 7:43:21 PM

I would say get the 2600xt its the best card there but if you are not worried about games then as has been said get the cheapest.
Mactronix
September 9, 2007 8:15:29 PM

Since you are not a gamer, there is no need to spend even $100 on a video card. Look at the lower down models. Spending more on a gaming oriented card (even a $100 model) just makes more noise, more heat, needs more power and becomes a more frequent maintenance issue to clean dust out and replace fan later.
September 10, 2007 2:07:32 AM

If you seriously think you need to upgrade 2600XT, faster and cheaper. I understand that you say that you are not a gamer, but do you by chance do some kind of graphic design(3D, photoshop)? If you do then you will see the performance gain from the bigger bus, and faster ram speed in the 2600XT. If not you probably would not have any problems with the 2600Pro. I don't understand why people assume that high end graphics are used only for games. Well without knowing anything more about what you would use it for, I would recommend the 2600Pro, it is the cheapest. Also I would see if your dealer has anything with a 256-bit bus with 256MB ram, usually cheaper since the amount of ram sells the card better. I know that I may get some people who don't think the same, but take the Ram * Bus Speed * Bus Width(bits) to get a rough estimate of the performance difference among the same chip set.
Example:
We will use the 2600XT for example (simply because I know it comes in 2 bus "widths")
2600XT = 512 * 700 * 128 = 45875200
Where as If you had only 256Mb Ram but with 256 bit bus width
2600XT = 256 * 700 * 256 = 45875200
I know you can use smaller numbers but rough example.
September 10, 2007 2:20:44 AM

F34R1355 said:
If you seriously think you need to upgrade 2600XT, faster and cheaper. I understand that you say that you are not a gamer, but do you by chance do some kind of graphic design(3D, photoshop)? If you do then you will see the performance gain from the bigger bus, and faster ram speed in the 2600XT. If not you probably would not have any problems with the 2600Pro. I don't understand why people assume that high end graphics are used only for games. Well without knowing anything more about what you would use it for, I would recommend the 2600Pro, it is the cheapest. Also I would see if your dealer has anything with a 256-bit bus with 256MB ram, usually cheaper since the amount of ram sells the card better. I know that I may get some people who don't think the same, but take the Ram * Bus Speed * Bus Width(bits) to get a rough estimate of the performance difference among the same chip set.
Example:
We will use the 2600XT for example (simply because I know it comes in 2 bus "widths")
2600XT = 512 * 700 * 128 = 45875200
Where as If you had only 256Mb Ram but with 256 bit bus width
2600XT = 256 * 700 * 256 = 45875200
I know you can use smaller numbers but rough example.


What are those numbers representing? What kind of equation is that and what? Radeon 2600xt doesn't come with 2 different memory bus.
September 10, 2007 2:44:43 AM

Were you trying to find effective memory bandwidth of the card? If you were. This is the correct equation for 2600xt.

128 / 8 x 700 x 2 = 32000

That's 32 GB/sec

September 10, 2007 3:14:03 AM

I know I'll get a hit for this from some people but I use an 8500 with 512 megs and I play lots of "modern" games (Fear, Quake4, Doom3, Oblivion,Bioshock, WOW, etc..). Most of the time the config utilities set up the games with everything set on high.(I don't have a lot of cash to aford the "high" end cards and here in Canada it was almost $200.00 when I bought it) I can't say I have much to complain about though it works fine for me!!:) 
September 10, 2007 3:16:06 AM

MickyDee said:
I can't say I have much to complain about though it works fine for me!!:) 


That's all that matters. :) 
a c 164 U Graphics card
September 10, 2007 5:30:26 AM

F34R1355 said:

Example:
We will use the 2600XT for example (simply because I know it comes in 2 bus "widths")
2600XT = 512 * 700 * 128 = 45875200
Where as If you had only 256Mb Ram but with 256 bit bus width
2600XT = 256 * 700 * 256 = 45875200
I know you can use smaller numbers but rough example.


LOL, did anyone else see a problem with this? Now that I've put the results in bold, is it easier to see? (and yes, as was also pointed out, the 2600XT only comes in the 128bit variety.)

While the video card might matter for some programs, I don't believe it play any part in photoshop. Here is a nice (and short) article about increase PS performance.

http://www.farrarfocus.com/ffdd/performance.htm

No where in the article or linked Extremetech article does it say to get a better 3D video card to handle the 2D riggers of PS. PS needs a fast CPU, lots of ram and fast harddrives, not a screaming video card.
a b U Graphics card
September 10, 2007 7:00:53 AM

marvelous211 said:
Were you trying to find effective memory bandwidth of the card? If you were. This is the correct equation for 2600xt.

No he wasn't, he was using a formula he invented to calculate a number that has no relevance to anything.

F34R1355 said:
I know that I may get some people who don't think the same, but take the Ram * Bus Speed * Bus Width(bits) to get a rough estimate of the performance difference among the same chip set.

He mentioned the amount of RAM in the formula, which is irrelevant. Another problem is that him talking about RAM has nothing to do with the performance difference among chipsets. The chips performs identically, but the cards (if there was a 256-bit 2600XT which I don't believe there is or it would be a semi-decent card) will not. And yes, he did get some people who don't think the same.

EDIT: Oh and those numbers are the same (that he came up with) so according to that, the mysterious 256-bit 2600XT would perform the same as the 128-bit. Whoops, can't see that happening.

September 10, 2007 8:32:26 AM

Yeah sorry. No I was just trying to show that a bus width of 256 gives with half the ram give the same ram performance as a bus width of 128 with twice the ram. The speed has no real factor in these equations. The amount of ram is relevant due to the fact that 256MB at 256-bit means faster transfer but at the same time you have less available so more time would be spent writing. I am not saying that a 256-bit 2600XT would be the same a a 128-bit. I am saying that with half the ram the performance would be comparable. After rereading what I put down I don't understand what I was doing there. ( 3 days up will do that I guess) And as far as photoshop I was aiming more for the Illustrator side, and the 3d Rendering effects, I know that it does not require to do 2d work. I also realize that a 256bit 2600XT (sadly) does not exist. I was thinking of the 1650XT. If you think that what I say is wrong then test it and show the results. I know that the bus makes a difference, and have known this since the MX440, ordered 2 128MB ones at the same time, but they performed different. Same Brand, Ram, Clock, the difference was one had a bigger bus. As far as the equation that I made is concerned, it is useless, unless you want a big number. Still I think that he would be better off with the 2600Pro unless he is doing 3D modeling.
If anybody figure out a use for that garbled number please let me know... lol
a b U Graphics card
September 10, 2007 8:46:55 AM

The bus speed is important, far more than the amount of memory. The 256-bit card with 256Mb RAM (if it existed) would beat the 512Mb card with only a 128-bit bus.
If we leave RAM speed the same (so we can remove it from your equation :kaola: ) but we change some of the values and extrapolate further:

1024MB with 64-bit bus: 1024 * 64 = 65536
256Mb with 256-bit bus: 256 * 256 = 65536

I assure you the 256Mb card would absolutely destroy the 1gb card.
a c 130 U Graphics card
September 10, 2007 9:46:52 AM

Marvelous is using the correct equation but i think randomiser has explaining it pretty well i know cars ar not a good annalogy for performance(Engines etc) but i would liken it to trying to get 256 cars down 64 lanes or 256 ie bandwidth then it would take 4 journeys to get them down the 64 lanes but only 1 down the 256 lanes.
So it should be obvious where the better performance lies.
Mactronix
September 10, 2007 12:03:30 PM

F34R1355 said:
Yeah sorry. No I was just trying to show that a bus width of 256 gives with half the ram give the same ram performance as a bus width of 128 with twice the ram. The speed has no real factor in these equations. The amount of ram is relevant due to the fact that 256MB at 256-bit means faster transfer but at the same time you have less available so more time would be spent writing. I am not saying that a 256-bit 2600XT would be the same a a 128-bit. I am saying that with half the ram the performance would be comparable. After rereading what I put down I don't understand what I was doing there. ( 3 days up will do that I guess) And as far as photoshop I was aiming more for the Illustrator side, and the 3d Rendering effects, I know that it does not require to do 2d work. I also realize that a 256bit 2600XT (sadly) does not exist. I was thinking of the 1650XT. If you think that what I say is wrong then test it and show the results. I know that the bus makes a difference, and have known this since the MX440, ordered 2 128MB ones at the same time, but they performed different. Same Brand, Ram, Clock, the difference was one had a bigger bus. As far as the equation that I made is concerned, it is useless, unless you want a big number. Still I think that he would be better off with the 2600Pro unless he is doing 3D modeling.
If anybody figure out a use for that garbled number please let me know... lol


Wrong... Do yourself a favor. Don't do anymore calculations for us. Some of us actually know what we are talking about. :kaola: 

And FYI there is no 256bit 1650xt either. :kaola: 
September 10, 2007 4:17:54 PM

Randomizer is saying the same thing I am. As far as a X1650XT 256-bit, it was a special release (Asus I believe), I know the I saw the thing and almost ordered one. To summerize what I am trying to say, same amount of ram, same speed wider bus = faster card. As far as cars on the road, can't say I have thought of it that way. I think of pipes, more specifically drains. It was X1650XT Xtreme Gamer, come to think of it, I saw the card on new egg. That was the only place I saw it. It was a useless card being that it was AGP, but I know that I have seen it. At the time I thought that there where only two bus sizes for the 1600 series (128, and 64). If you want to waste your time at some point you can try to find it. Why do you quote the whole post?
I don't see why to argue about this on the ram, the bottleneck is usually the core. Go ahead and tell me why I wrong at saying that.
September 10, 2007 4:58:33 PM

F34R1355 said:
Randomizer is saying the same thing I am. As far as a X1650XT 256-bit, it was a special release (Asus I believe), I know the I saw the thing and almost ordered one. To summerize what I am trying to say, same amount of ram, same speed wider bus = faster card. As far as cars on the road, can't say I have thought of it that way. I think of pipes, more specifically drains. It was X1650XT Xtreme Gamer, come to think of it, I saw the card on new egg. That was the only place I saw it. It was a useless card being that it was AGP, but I know that I have seen it. At the time I thought that there where only two bus sizes for the 1600 series (128, and 64). If you want to waste your time at some point you can try to find it. Why do you quote the whole post?
I don't see why to argue about this on the ram, the bottleneck is usually the core. Go ahead and tell me why I wrong at saying that.


Just stop while you are ahead. Although you were never ahead to begin with. :non: 

Ramdomizer is not saying the same thing. He made you look clueless with your equation and logic. :lol: 

You got a link to 1650xt having 256bit memory bus? You will never find it because 3rd party can't change memory bus on their cards except what ATI intended. They can use faster memory however and overclock their cards.

Fact that you give your opinion is great but you need to know what you are talking about first.
a c 130 U Graphics card
September 10, 2007 5:09:06 PM

HIS made a big thing of their 256bit internal memory ring bus on a X1650XT maybe that is where the confusion is coming from.
Mactronix :) 
September 10, 2007 5:13:42 PM

mactronix said:
HIS made a big thing of their 256bit internal memory ring bus on a X1650XT maybe that is where the confusion is coming from.
Mactronix :) 


If that's the case all x1000 series are 256 bit. LOL
a c 130 U Graphics card
September 10, 2007 5:32:50 PM

Sorry marvelous but you are wrong not all X1000 series cards came with an internal memory ring bus of 256.
This is taken from a review of a x1300
The power circuit is visibly simplified, and the wiring is different around the memory chips. Unlike on the RADEON X1600, GDDR2 memory is installed here, and on both sides of the PCB. Infineon’s HYB18T256161AF-25 chips have a 16Mx16 structure, so eight such chips were required for the 128-bit memory bus..
Mactronix
September 10, 2007 5:41:31 PM

mactronix said:
Sorry marvelous but you are wrong not all X1000 series cards came with an internal memory ring bus of 256.
This is taken from a review of a x1300
The power circuit is visibly simplified, and the wiring is different around the memory chips. Unlike on the RADEON X1600, GDDR2 memory is installed here, and on both sides of the PCB. Infineon%u2019s HYB18T256161AF-25 chips have a 16Mx16 structure, so eight such chips were required for the 128-bit memory bus..
Mactronix


You don't need to be sorry because I know. It was a joke but you didn't get it. :whistle:  You just wanted to say I was wrong. LOL

Screw internal bus. It's like 3dfx saying voodoo 3 had 32bit color internally but only spit out 16bits.

It's useless point to make an argument. The final result is the same.
a c 130 U Graphics card
September 10, 2007 6:07:38 PM

Yes i does seem i missed the joke,Dont supose you would care to explain it to me?
No argument from me i was just suggesting a reason for f34r1355 thinking there was a 256bit card.
Mactronix
September 10, 2007 6:25:39 PM

No point trying to explain my humor to someone with no sense of humor.

Anyways there is no 256bit 1650xt and internal bus or not the result is same gb/s in the end.
a c 130 U Graphics card
September 10, 2007 6:45:02 PM

Oh dear seems we hit a nerve there then most of us don't like being wrong but then again most of us are big enough to admit it :lol: 
Mactronix
September 10, 2007 6:48:09 PM

mactronix said:
Oh dear seems we hit a nerve there then most of us don't like being wrong but then again most of us are big enough to admit it :lol: 
Mactronix


Oh gee. You are still upset about last week when you got owned and I proved my original point. :lol: 

Weren't you the one who linked to firingsquad and trying to prove 8600gts was slower when it really wasn't. You proved yourself wrong. :lol: 
September 10, 2007 6:49:09 PM

It happens when you get overwhelmed by knowledge and logic when you were WRONG. :lol: 

That's the reason why you went out of your way to say I was wrong in this thread when I was joking with a lol.

Some people hold grudges or agendas against someone over a argument.

Do you really think I didn't know anything about internal memory bus when I can make a proper formula to calculate memory bandwidth? gimme a break.


edit:
Do you even remember what my original point was before you started being a follower put your useless 2 cents in?

Let me remind you what my original point was. I said 8600gts and 1950pro is neck and neck but 8600gts had better image quality. They are about equal in performance and cost about same.

We already know Geforce 8 has better image quality because I linked you to a techreport article that explains this. http://techreport.com/articles.x/11211/6

I linked to $130 8600gts which is in line with 1950pro internet prices and you linked to the firingsquad benchmark that tells you it's pretty much neck and neck. :lol: 



a c 130 U Graphics card
September 10, 2007 8:16:57 PM

Oh gee its got nothing to do with anything that you seem to think happened last week on a different thread some of us are bigger than that, but since you felt the need to bring it up i checked as i felt fairly sure that your assertion that i had posted links to you from firingsquad was again wrong.

marvelous2 11
Don't rape me!
Profile: enthusiast
Posted on 09-03-2007 at 05:04:59 AM

kamrooz wrote :

http://firingsquad.com/hardware/nv [...] efault.asp
Keep in mind we are talking about the 8600 GTS...Not the insane OC revisions such as the OC2 which retails between 210-250...Also keep in mind the x1950 pro outperforms most of these more expensive cards. If you want to talk about OC2 which retails over 200 the x1950xt is fair game as well...

In case you don't recognise it its what you posted back to Kamrooz who was the poster you seem to have me confused with.
So obviously your reasoning that this was why according to you i went out of my way to prove you wrong doesn't hold water either.

Don't see how posting a single link that showed your statement to be erroneous can be classed as "going out of my way" either.

Mactronix :lol: 

September 10, 2007 8:31:10 PM

mactronix said:
Oh gee its got nothing to do with anything that you seem to think happened last week on a different thread some of us are bigger than that, but since you felt the need to bring it up i checked as i felt fairly sure that your assertion that i had posted links to you from firingsquad was again wrong.

marvelous2 11
Don't rape me!
Profile: enthusiast
Posted on 09-03-2007 at 05:04:59 AM

kamrooz wrote :

http://firingsquad.com/hardware/nv [...] efault.asp
Keep in mind we are talking about the 8600 GTS...Not the insane OC revisions such as the OC2 which retails between 210-250...Also keep in mind the x1950 pro outperforms most of these more expensive cards. If you want to talk about OC2 which retails over 200 the x1950xt is fair game as well...

In case you don't recognise it its what you posted back to Kamrooz who was the poster you seem to have me confused with.
So obviously your reasoning that this was why according to you i went out of my way to prove you wrong doesn't hold water either.

Don't see how posting a single link that showed your statement to be erroneous can be classed as "going out of my way" either.

Mactronix :lol: 


I remember you were one of the less informed who joined in who put your useless 2 cents in and was very upset. LOL Either case I proved you along with bunch of mindless followers of this forum. :lol: 

What does that whole firingsquad article prove? that it's neck and neck. The card I linked to that was $130 last week was EVGA OC version. :lol: 

mactronix said:
To marvelous211 you are definatly wrong my friend the 1950 pro is a lot better in gaming than any of the mid range HD cards you say you have had both and the performance was similar and i wonder if your psu was up to running the 1950 pro?is there any chance it was under powered and clocked itself down?
Mactronix

a b U Graphics card
September 10, 2007 8:36:20 PM

Usually the order of important runs bitwidth>=speed >> Vram size>=latency.

hat's because the jump for bitwidth usually is huge and in most cases a crippled bitwidth can't be made up for by speed (see the 9800SE as a great example), think about the 128bit GDDR4/GDDR3 speeds on HD2600Pro/XT and think of what you'd need for speed to make up for it with 64 bit interface.

Now VRAM does matter, it's not completely out of the picture, but it matters most when the cards start to top out, so the impact is usually less. If we were talking about a 512Bit 32MB card versus a 128bit 256MB card you might see better examples of those limits, but since half of the limits we see are artificial (cards don't expose ultra setting unless 512MB is present yet some would perform worse than their 256MB counterparts due to the underlying processing and differece. So only for ridiculous constraints does it move further up the chain.

F34R1355 said:
Still I think that he would be better off with the 2600Pro unless he is doing 3D modeling.


Well I don't know about that, the recent 3Dproffesor results leads me to think may be the HD2600Pro has alot of untapped potential with it's 120 processors, especially since many modeling apps aren't going to be affected by the low ROP count, and they are one of the few places where memory size does matter alot.

Anywhoo, in the case of this thread the HD2600Pro>GF8500 and the HD2600XT will win/lose to the GF8600GT both being good at different areas / apps.

Considering that gaming is a mid-to-low priority I'd stick with the HD2600Pro, but if you are concerned about new games and higher settings then looking at the GF8600GT and HD2600XT makes sense too.

But for Photoshop and such, alot of the previous generations differences have been nulified and until they can actually make use of more GPGPU style co-processing, the graphics cards won't play a big role in either photo or video editing/manipulation, the focus will remain on CPU, RAM, HDD.
a c 130 U Graphics card
September 10, 2007 8:54:08 PM

As i have shown you the firingsquad link wasnt me i didnt read it so i couldnt coment.
What i do know is that it has nothing to do with you being WRONG about the memory bus size on the x1000 cards. And as for your assertion that it follows you would know about the subject bacause you can make the proper formula to calculate bandwidth :lol:  Do me a favour i have one thing to say to that. Google-Copy and paste :lol: 
Thats what you did through a lot of last weeks post that you keep banging on about trying to change the subject,i read most of the reviews on new egg and it all looked very familiar. I dont think its good practice to try and pass off joe blogs opinion as fact :non:  If i did have a problem with you i would have brought this up at the time but i am of the view that every body has an opinion and is entitled to air it right or wrong thats what the forums are about,i have learnt a lot from this forum and im guessing you may not care but the only reason i knew your statement was wrong was becaue i spent about a month researching the topic with a view to having half an idea before i upgraded my old 9800pro to a X1650XT.
So you see its not a witch hunt, just a coincidance that i was here and had the knowledge about the subject.
Mactronix :kaola: 

September 10, 2007 9:14:53 PM

Either way I proved my original point along with you and the mindless followers of this forum last week. :lol: 

What you didn't know is that I was making a joke with you and since you were upset from last week you wanted to say that I was WRONG. :cry:  Why not google the exact same formula and copy and paste and show me the source and the joe blogs? You can't because you don't even know how to search the internet and it took you a whole month just to figure out 1650xt had internal 256 bit memory bus~ :lol:  Internal or not it has the same 128 bit memory bandwidth as shown by my formula. :kaola: 

I could have told you x1900xt had internal 512 bit memory bus but it's pointless since it behaves like 256bit memory controller. :pt1cable: 

edit:
Talk about immaturity. I was trying to wipe our slate clean by joking but you couldn't even do that. You wanted to test me and insult my intelligence like this.

mactronix said:
Oh dear seems we hit a nerve there then most of us don't like being wrong but then again most of us are big enough to admit it :lol: 
Mactronix


I thought you were bigger than that? I guess not.

mactronix said:
some of us are bigger than that


I guess that some isn't you. I know how to put 1 + 1 together too. You aren't the only one... That includes YOU TOO.
a b U Graphics card
September 10, 2007 11:30:21 PM

marvelous211 said:

I could have told you x1900xt had internal 512 bit memory bus but it's pointless since it behaves like 256bit memory controller. :pt1cable: 


I don't care about the other stuff going on, you guys can fight on, but the above statement underlies your misunderstanding of the ringbus versus crossbar benefits for both communication to the DDR memory and the internall bit flow.

That people get the 2 way bus and interface numbers confused is one thing, but if in a thread where there is already confusion about which card has what bitwidth you add to it by a misleading joke, then that's not helping, and it also doesn't show whether your joke was misinformed sarcasm in reply or just a very poorly worded and poorly delivered joke. :heink: 

So I can't blame anyone for jumping on that, had I seen it before you two went at each other I would've mentioned it too.

Either way, it didn't contribute much to clearing up the bad formula that preceded it, if anything you're simply going to confuse people like f34 even more. :pfff: 

I think the point's been made and I think it's time to either focus on the main issue or move on.
September 10, 2007 11:56:48 PM

TheGreatGrapeApe said:
I don't care about the other stuff going on, you guys can fight on, but the above statement underlies your misunderstanding of the ringbus versus crossbar benefits for both communication to the DDR memory and the internall bit flow.


Why not show us if you have better understanding? I'm always willing to learn.


That people get the 2 way bus and interface numbers confused is one thing, but if in a thread where there is already confusion about which card has what bitwidth you add to it by a misleading joke, then that's not helping, and it also doesn't show whether your joke was misinformed sarcasm in reply or just a very poorly worded and poorly delivered joke. :heink: said:

That people get the 2 way bus and interface numbers confused is one thing, but if in a thread where there is already confusion about which card has what bitwidth you add to it by a misleading joke, then that's not helping, and it also doesn't show whether your joke was misinformed sarcasm in reply or just a very poorly worded and poorly delivered joke. :heink: 


English is my second language.



So I can't blame anyone for jumping on that, had I seen it before you two went at each other I would've mentioned it too.

Either way, it didn't contribute much to clearing up the bad formula that preceded it, if anything you're simply going to confuse people like f34 even more. :pfff:

I think the point's been made and I think it's time to either focus on the main issue or [b said:
move on.]
So I can't blame anyone for jumping on that, had I seen it before you two went at each other I would've mentioned it too.

Either way, it didn't contribute much to clearing up the bad formula that preceded it, if anything you're simply going to confuse people like f34 even more. :pfff: 

I think the point's been made and I think it's time to either focus on the main issue or move on.
[/b]

Then clear it up to your best knowledge.
a c 164 U Graphics card
September 11, 2007 12:07:04 AM

I'll leave the petty bickering to the ladies that are engaged in it. There is a reason why forums have the PM feature, learn to use it in the future please.

To the OP, you can pretty much safely ignore most of the lower half of the first page. I haven't seen you post in awhile, so I'm wondering what you are thinking about. Have you made a decision? Do you have any other questions?
September 11, 2007 12:15:39 AM

I openly admit that my first formula, and comment about card performance in general was wrong. It may be possible that I saw the internal bus number, and mistook it for the bus width on the X1650xt.
Thanks TheGrateGrapeApe for explaining that. That was what I was trying to say, just couldn't seem to get it out right. I understand the ram does not matter much unless you are working with big textures, and I also understand the memory size alone does not make up for poor bandwidth.
I don't mind being wrong. But I will ask that if you are going to tell me I am, please show me what I did wrong. For example if I said something like "1 + 1 *1 =3", then I would not mind "your wrong, "1+1*1= 2". But if you would like to continue telling me why I am wrong please start another thread.
I second the seeing of what TheGrateGrapeApes understanding for the workings of the memory bus on the card.
To marvelous211, if english is your secondary language, what is your primary. I could not tell english was your second.
To BorisAro, I apologize for the "argument" on your thread. For what it is worth I think that you would be good with the X2600Pro.
September 11, 2007 12:20:20 AM

I am Korean American as you can see by my avatar (seo taiji). He is Korean rock musician you can search on youtube if you'd like. Korean is my first language where our dialects have different meanings and spoken backwards.
a b U Graphics card
September 11, 2007 1:09:15 AM

To make it easy on everyone, here's the best description, from Wavey Dave, who now works for Ati;

http://www.beyond3d.com/content/reviews/27/5

Good enough for you?

Some of the things Dave didn't focus on in his article is to mention that the Crossbar designs increased wire length and number posses some problems for GDDR4 memory due to it's increased noise at higher frequency at the choke point, this may still be a major inhibitor for the G9x series in the near term trying to achieve higher bitwdith and not compromise speed, and they will likely have to move to a similar technique to push their next generation cards towards 512bit external bus.
Dave Orton already said that the push to 55nm won't pose a problem for the current 512/1024 ring bus and getting it to work on such a small package interface with the card, at some point you start running out of space to make those traces connect to the chip, and without the ring-bus that would come much sooner as the density at a single point it mich higher. So I suspect either nV goes with something similar or else they stay below 512bit external interface until after they move to multi-core and larger areas to work with again.

With the crossbar working at twice the bitwdith (or the same bitwidth in each direction) it can better match the external bandwidth provided by the DDR memory, wheres the memory crossbar with the bitwidth identical to the memory is stuck at half, it's not a huge deal, less so than the other factors listed above, but it's a nice little additional perk to maker the most of the architecture at hand, and allows it to maximize the throughput at the ring-stops and pass that through to/from the core.

The design also allows you greater flexability to disable portion without changing the chip architecture. They haven't used a-symetrical layouts, but odd number layouts (as in odd numnber groups, not an actual odd number which doesn't work well with memory) would be possible without disabling portions of the chip (as we see in the G80GTS and GTX). But the practicallity of this in the current lineup isn't there, and an R600 Pro makes more sense than something in between.

Main thing to focus on is that the internal ring bus would cause less confusion if it were (misrepresented) as pathways of equal bitwidth to the external memory interface. It simiplifies too much for people who know better, but I think that if people saw the internal ringbus as something like 2-way 128/256/512 bit interface instead of the 256/512/1024 example they would quickly get it, in the same way that people accept DDR400 as 200mhz in double data rate, and understand better 'effective' versus actual, with the ringbus acting like the 'effective' portion and the modules interface being the 'actual'. I think a similar explanation works better then trying to get people to appreciate the multiple ring stops and total bitwdith/bandwidth.

marvelous211 said:

English is my second language.


I didn't ask you if it was your first or second or third language or if you use BableFish to read/post. You say it as if it were an excuse after chiding someone else for not getting your joke or the meaning of your post.

So was there something else, or were you ready to move on now?
September 11, 2007 1:18:46 AM

TheGreatGrapeApe said:
To make it easy on everyone, here's the best description, from Wavey Dave, who now works for Ati;

http://www.beyond3d.com/content/reviews/27/5

Good enough for you?

Some of the things Dave didn't focus on in his article is to mention that the Crossbar designs increased wire length and number posses some problems for GDDR4 memory due to it's increased noise at higher frequency at the choke point, this may still be a major inhibitor for the G9x series in the near term trying to achieve higher bitwdith and not compromise speed, and they will likely have to move to a similar technique to push their next generation cards towards 512bit external bus.
Dave Orton already said that the push to 55nm won't pose a problem for the current 512/1024 ring bus and getting it to work on such a small package interface with the card, at some point you start running out of space to make those traces connect to the chip, and without the ring-bus that would come much sooner as the density at a single point it mich higher. So I suspect either nV goes with something similar or else they stay below 512bit external interface until after they move to multi-core and larger areas to work with again.

With the crossbar working at twice the bitwdith (or the same bitwidth in each direction) it can better match the external bandwidth provided by the DDR memory, wheres the memory crossbar with the bitwidth identical to the memory is stuck at half, it's not a huge deal, less so than the other factors listed above, but it's a nice little additional perk to maker the most of the architecture at hand, and allows it to maximize the throughput at the ring-stops and pass that through to/from the core.

The design also allows you greater flexability to disable portion without changing the chip architecture. They haven't used a-symetrical layouts, but odd number layouts (as in odd numnber groups, not an actual odd number which doesn't work well with memory) would be possible without disabling portions of the chip (as we see in the G80GTS and GTX). But the practicallity of this in the current lineup isn't there, and an R600 Pro makes more sense than something in between.

Main thing to focus on is that the internal ring bus would cause less confusion if it were (misrepresented) as pathways of equal bitwidth to the external memory interface. It simiplifies too much for people who know better, but I think that if people saw the internal ringbus as something like 2-way 128/256/512 bit interface instead of the 256/512/1024 example they would quickly get it, in the same way that people accept DDR400 as 200mhz in double data rate, and understand better 'effective' versus actual, with the ringbus acting like the 'effective' portion and the modules interface being the 'actual'. I think a similar explanation works better then trying to get people to appreciate the multiple ring stops and total bitwdith/bandwidth.


I had seen this article before. I know Dave works with ATI and he talks about stuff from their tech. So what is that really telling us in simple words so stupid people like me can understand that has anything to do with what you quoted on my post. So you are telling me it behaves like 512 bit memory bus?



I didn't ask you if it was your first or second or third language or if you use BableFish to read/post. You say it as if it were an excuse after chiding someone else for not getting your joke or the meaning of your post.

So was there something else, or were you ready to move on now? said:

I didn't ask you if it was your first or second or third language or if you use BableFish to read/post. You say it as if it were an excuse after chiding someone else for not getting your joke or the meaning of your post.

So was there something else, or were you ready to move on now?


Did you ever try using Babelfish? It's a horrible way to translate with missing words and grammar especially trying to translate from East Asian country to a Western country and vice versa.
a b U Graphics card
September 11, 2007 1:38:55 AM

The last part of that section does it quick an dirty if you can't get the more complex example. Think of it like traffic and how it would work if you were to have the data follow the wire length of a single crossbar versus the more evenly spread out and 'dual direction/channel' of the ringbus. All operating at the same frequencies. Then thing of choke points and how having it more spread out and having multiple on/off points going both ways make it quicker (ie less latency').

If you need it explained further I can make the effort after dinner, and after I'm done upgrading the memory on my new laptop.

I'll see if I can rustle up some of the old ATi R520 slides I posted here way back during the X1800 launch to help f34, you however claiming you're 'too stupid' to understand Wavey Dave after arguing with Mactronix is just laughable. :sarcasm:
September 11, 2007 1:39:06 AM

I think I get it... Not sure, but I think I get it. 512bit comes from the fact the it is 2 256-bit "pipes", one for writing, one for reading? Seems like it would have little effect on a card reading a few big objects(A shooter in a cube room, like some of the basic half life 2 maps?) from the memory, but would help a lot with writing and reading a bunch of small things(In a shooter, large map where the same texture is not used over and over, or maybe dynamic textures?). It increases memory efficiency, but does not do much for overall performance because you still only have 256-bit to the core. as far as seeing what this has to do with marvelous211's comment, I am confused. I mean yes it has 512 internal, but to get that out it is still cut down to 256 bit. I may be entirely mistaken, please let me know.
September 11, 2007 3:16:11 AM

TheGreatGrapeApe said:
The last part of that section does it quick an dirty if you can't get the more complex example. Think of it like traffic and how it would work if you were to have the data follow the wire length of a single crossbar versus the more evenly spread out and 'dual direction/channel' of the ringbus. All operating at the same frequencies. Then thing of choke points and how having it more spread out and having multiple on/off points going both ways make it quicker (ie less latency').

If you need it explained further I can make the effort after dinner, and after I'm done upgrading the memory on my new laptop.

I'll see if I can rustle up some of the old ATi R520 slides I posted here way back during the X1800 launch to help f34, you however claiming you're 'too stupid' to understand Wavey Dave after arguing with Mactronix is just laughable. :sarcasm:


Maybe stupid isn't the word I should have used. More like %uD575%uAC08%uB9B0 people. There's no correct word for this in English. Laugh all you want but I think I've explained it to you that English is not my first language. Seems you want to pick sides and verbally attack me for it. Not acting like a moderator at all.

I understand all this but what does it have any revelation with my the post you quoted? The simplest for people to understand is that it's in the beginning stages of being a 512bit memory controller. It still doesn't behave like 512bit memory controller either way which I correctly used on previous post. The end result is same memory bandwidth.
a b U Graphics card
September 11, 2007 5:34:08 AM

All right you guys asked for it.

a c 130 U Graphics card
September 11, 2007 6:29:32 AM

In closing i would just like to apologize to the op about this going of in to a point less argument but i felt aggrieved by being accused of posting stuff i clearly didn't.
To TheGreatGrapeApe
Sorry i went of on a tangent there seems I'm now guilty of some thing i pulled our fav mod up on previously
:D  so that makes us 1-1 now :lol: 
Mactronix
!