Bummer... no 1.6ns here, but questions on DDR3 memory...

[godofgorks]

Bummer....

Well SuperNade if you see this, you LIE. Don't worry bud, it's all good, but I just realized the neigther of my 6600gt's have the 1.6ns memory. Since the system isn't running I pulled the cards and started going over them to get more comfortable for volt modding and ran the ram numbers through Samsung sight and found both cards to be running with 2.0ns memory.

Then I realized something, the same module, comes in many more flavors, 2.0, 1.6, 1.4ns. They are rated 500, 600, 700 MHz repectively. There is also a Pb free version that runs on .2v less, with in addition to the speeds already also has the option of 1.2ns @ 800Mhz.

Soooo... if some cards used the 1.6ns and others the 2.0ns, why can't others have the even 'better' stuff. Get some 1.2ns that'll run 800, think (1600) on only 1.8v. I mean that stuff would be the UBER 1337.

Linky to samsung site:
http://www.samsung.com/Products/Semiconductor/common/product_list.aspx?family_cd=GME1001

Thoughts?

EDIT: After a little more digging I found that one model of the Pb version runs at 1.8v and that's the 2.0ns. But the lower ns are rated for 2.0v. So card makers would have had to either A drop to the Pb-free version to stay with 1.8v or B. change the board layout to offer 2.0v to the Pb version chip, or lastly C. undervolt the 1.6ns Pb ram

[crimedog]

you didn't mention what card but i assume 6800gs
1.6ns was put on the ultras, 2ns was put on the GT's. faster than that is only going on the new cards.

ps mickeymouse is supposedly releasing the 6800gs vmod at vrzone tomorrow

[godofgorks]

you didn't mention what card but i assume 6800gs
1.6ns was put on the ultras, 2ns was put on the GT's. faster than that is only going on the new cards.

ps mickeymouse is supposedly releasing the 6800gs vmod at vrzone tomorrow

whoops, sorry got ahead of myself...

Nope not 68XX here, still back in the days of 6600GT..

editing first post now

[Kiriakos]

Speed and quality parts goes only in the top range, thats why there is always a big price gap .

Your question is not that wise.

I do not flame the 6600 as model , but why some one to use high quality memory on a model with an( low to mid scale) GPU engine ?!!

From another point of view , you get what you payed for .
You did not pay for high quality ram !! I did .. :D

370 Euros , not dollars !!

[godofgorks]

Speed and quality parts goes only in the top range, thats why there is always a big price gap .

Your question is not that wise.

I do not flame the 6600 as model , but why some one to use high quality memory on a model with an( low to mid scale) GPU engine ?!!

From another point of view , you get what you payed for .
You did not pay for high quality ram !! I did .. :D

370 Euros , not dollars !!

Considering these memory modules where available at the time the 6600gt WAS top of the line, your are misunderstanding my whole question. I understand the price gap and it's reason, but I'm looking at this as a why didn't they then, as opposed to why don't they now. It also goes to question what manufactures choose to do when they used 1.6ns ram. Which route did they take?

[Kiriakos]

Well my good friend, it looks that NVIDIA at that time was not feel any pressure from any opponent, for a faster product .

If you feel the danger, that some one becomes faster , only then you develop or use better parts , simple as that .

By the way, the better clocking memory, is not a mast have, in modern GPU’s , a good example is the 6800GS with DDR II .

Personally I prefer the GT over the GS .
The reason is that the NV40 was build with the DDR 3 in mind , and the constructional specs weight more , than what you get as FPS in some of the gaming titles.

I do not like to confuse you with in depth technical stuff , I will just say what I read from a game developer , ( The modern games uses the GPU not only for pixels mapping , but for geometric calculations too , plus that the card has to render graphics that the user does not see ) .
The last sounds like hocus pocus, but the truth is that one GPU needs equal fast memory , enough to cover the needs of the GPU , in order to get the max of millions of calculations per second, that the GPU is designed for .

If the memory becomes faster than the GPU calculations speed , all that will happen is the GPU, it will NOT take any advantage from it, there is no benefit from that .

In praxis , I did see more gain from the GPU core speed than the memory .

Last and very worthy as info is the memory latency effect.
The ultra speedy rams most of the times they run at low latency ( GT Ultra) , from the other side (GT flavours ) run at lower speed , BUT with smaller latencies(faster) .
There is no gain from the plane memory speed , as I have all ready say , the GT Ultra core , needs speedy memory to make the amount of calculations that it had got designed for , if at that point we use slower memory modules , the GPU will never bring the performance levels that we expect .

Simple, as that.

[godofgorks]

I understand everything you're saying...

Well my good friend, it looks that NVIDIA at that time was not feel any pressure from any opponent, for a faster product .

If you feel the danger, that some one becomes faster , only then you develop or use better parts , simple as that .
But who is the determining body though, NVIDIA, or the card manufacturer who decides what ram gets put on the card. I'm sure NVIDIA designs their cores tailored to certain rams, as you stated with the NV40, and most manufactures use a certain line, but who ultimately picks the final product.

By the way, the better clocking memory, is not a mast have, in modern GPU’s , a good example is the 6800GS with DDR II .
I must have overlooked that spec, very interesting. And it makes sense, that insane fast ram would be overkill for some GPU's. There always has to be balance.

I do not like to confuse you with in depth technical stuff , I will just say what I read from a game developer , ( The modern games uses the GPU not only for pixels mapping , but for geometric calculations too , plus that the card has to render graphics that the user does not see ) .
The last sounds like hocus pocus, but the truth is that one GPU needs equal fast memory , enough to cover the needs of the GPU , in order to get the max of millions of calculations per second, that the GPU is designed for .
Don't worry technical details don't scare me, I just read down the spec sheet of the samsung DDR3 ram... but if they are not needed, they do make things a lot simpler. The part about the render graphics you don't see I understand because wouldn't they render layers and some layers over lap others and in the end you as the end user don't actually see some of the work. I'm assuming that's what he was implying.

But mainly my biggest topic of interest was these companies, such as XFX that put out cards with both the 2.0ns and the 1.6ns ram. Due to the specs, did they have to alter their card design to specially accomodate each ram, or did they run them out of spec?

This really isn't a I don't know what's going on and just keep asking questions thread, it's more of a discussion and as I'm researching info online, I bring my thoughts and comments here to be bounced around. I'm looking towards to having a fairly indepth understanding of memory, and not just the video ram, but ram in general, as the stuff really intrigues me. Thanks for the info so far....

[Kiriakos]

If the memory becomes faster than the GPU calculations speed , all that will happen is the GPU, it will NOT take any advantage from it, there is no benefit from that .

The above is the all secret ....



Now , the 2.0ms or the 1.6ms memory , it does not make the big bang, in the 3D benchmarks.

Now about the VGA makers, they do not have lots of choices for parts, specialy for mem modules.
So , the worst part can only be 4% slower that the fastet one. :)

[Kiriakos]

Ok .. :) thats a bumb, but I think that this info must become sticky.

[godofgorks]

Ok .. :) thats a bumb, but I think that this info must become sticky.

Cleaned out the subscriptions and this one went out by mistake.

Now with the DDR4 coming out, it seems memory for graphic card memory is advancing leaps and bounds ahead of the cores. At least we can be comfortable that the cores won't be being held back by memory. At least for awhile. And also since newer cards won't require complete top of the line memory, it will help keep costs down.

[Kiriakos]

I will agree, that core speed (or cores workload capacity ), goes hand to hand, with the necessary bandwidth, that a memory module have to offer.

Larger core = more memory bandwidth.
From the other hand, the memory capacity of the graphic card is not related with their performance levels.

Its only have to do with higher resolutions, more resolution = more memory.

So if its true, the story about the DDR4, then we will see more faster cards real soon.

But be careful, the gain from the DDR to DDR2, was not like night and day, same story with the DDR3.

If the DDR4 has something to offer, it will be a gain of 15% at most.