DDR II as it pertains to Geforce FX

[MUYA]

Guys just read an article over at lostcircuits about DDR II. Now I am not so technically adept as understand what they said entirely but one thing did grasp my eye, the technical adpet ppl probably know about this but at this page;
http://www.lostcircuits.com/memory/ddrii/2.shtml
The article says; To sum this up, DDR II 400 will feature a 100 MHz (10 ns clock cycle; tCK) core and I/O buffers that are running at twice the frequency while outputting data using DDR mode. This results in output of four bits / clock which is equivalent to a quad pumped interface in terms of bandwidth.

What was interesting to me was the last bit...quad pumped? So if I understood my what lil tech knowledge I have, then in my lay man logic it would appear to me that at 128bit ddr II (on the GeforceFX) would allow to compete to a 256 bit DDR mem( on the 9700pro etc?) ie 128bit times 4 (quad pumped) of DDR II as compared to the 256 times 2(double pumped) of the ddrI? Therefore in terms of bandwidth? there are about equal etc?
Pls people who are technically knowledgable please comment! I would like to if this is why nV decided to opt for 128 bit ddr II ratehr than ddrI because they could compete with 256bit ddr I memory and its bandwidth?

MUYA

[SurfMonkey]

I think it you'll find that the FX is still limited to 128bits of data per cycle, the only difference being that it can access the memory faster e.g. a higher clock speed. It all depends on how good the FX is at optimising the data stored and retrieved from local mem.

I guess that would also imply a 4x32 memory crossbar!?

[StealthHawk]

uh, WTF? i thought everyone agreed that ddrII was double pumped...

[-=DVS=-]

Uhh who cares what hardware spec are only real world Performance is important and MUYA how can DDR 2 for GF FX double its performance on 128bit bus when Nvidia stated they raw memory output is only 16GB/s at 500mhz (DDR2=1000) compared to 19.840GB/s of Radeon 9700Pro 310 (DDR1=620) :rolleyes:

[PreservedSwine]

If the NV30 had a core that was rwice as fast as the RAM, then, yes, it would have twice the bandwidth of DDRI.
Here's (http://www4.tomshardware.com/business/20021120/comdex_day_2-01.html#nvidia_geforce_fx_memory_bandwidth) an article hitting a few points, and here's (http://www.jedec.org/DOWNLOAD/pub21/HotDDR/DDR_II_evolution.pdf) the actual J-dec whitpapers....check out page 6...

However, unti core speeds are ramped up to twice the speed of the memory, the memory will operate as DDRI....unless I've mis-understood....:)

[Gar]

MUYA is just trying to convince himself that a 128bit bus is more advanced than it once was thought. The fact is the GFFX is going to be a fine FPS producing card at high res and with good image quality. Now would it be significantly faster if it was tweaked to use a 256 bit bus? Most definitely, but the price would be through the roof. It really doesn't matter because once the R350 comes out its going head to head with whatever that brings to the table.

[MUYA]

Originally posted by Gar
MUYA is just trying to convince himself that a 128bit bus is more advanced than it once was thought.

Dude u got me wrong there I was not trying convince anyone let alone mysef..I was trying to understand. Because from what I had breifly read in articles and also forum postings, I was under the impression that DDR II was double pumped as Stealthawk put it...now when I read that article at the URL given above, it said it was quad pumped. Which seemed to be surprising and then I thought quad pumped vs double pumped..hmm then 128 bit quad pumped vs 256 double pumped? Now for me, I though that data would be pushed 4 times a clock cycle in DDR II (quad pumped) vs 2 timesin DDR I. And on that train of thought, could it be that maybe (maybe being the operative word) 128 bit quadpumped memory (DDR II) offers the same performance as the 256 bit doubled pumped memory( DRRI). I seems logical but I know none of these matters as as clear cut and easy to understand...thats whY I was asking the technophiles and tech-adept ppl to comment as to if that was possible? Seems not as most return posts seem to suggest!

Anyway thanks guys ...although i am still not 100% clear. I am an Analytical Chemist..so forgive my 3d card/memory misconceptions !! :)

Thanks for replies!

MUYA

[Joe DeFuria]

Here's the deal. Use the illustration below as a guide.

http://www.lostcircuits.com/memory/ddrii/timing2.gif

Note that you must think of DDR and DDR-II as having three different frequencies. Memory Cell, I/O Clock, Data Rate. What's important to know is that the I/O CLOCK is the actual MHz of the ram chip....NOT THE memory cell array. 400 Mhz DDR-I has an I/O clock of 400 Mhz, and 400 Mhz DDR-II has an I/O clock of 400 Mhz.

(so in the above illustration, you're looking at a 100 Mhz DDR-I module, and a 200 Mhz DDR-II module. NOT a 100 Mhz DDR-II module!)

All DDR-II modules are in fact "quad-pumped" relative to the memory cell array. However, memory cell array frequencies have NOTHING to do with the stated frequency of the ram chip! It's an "internal" clock.

What does this mean?

It means that the "500 Mhz DDR-II" on the GeForceFX is running at an i/o clock frequency of 500 Mhz, for a data frequency of 1 GHz. This is exactly the same that a 500 Mhz DDR-I module would be. The difference is, the 500 Mhz DDRII module has in internal cell array frequency of 250 Mhz....HALF that of a 500 MHz DDR-I module.

So in short, it's helpful to think of it this way: DDR-I and DDR-II are both effectively double pupmped with respect to the reported chip clock frequency. And that is what is "most important" in terms of determining the bandwidth. However, DDR-II's "internal" cell array frequency is one half of the chip clock.

Did that help?

[Joe DeFuria]

If the NV30 had a core that was rwice as fast as the RAM, then, yes, it would have twice the bandwidth of DDRI....unless I've mis-understood....

You misunderstood. ;)

See my above response. NV30 core has nothing to do with the bandwidth of DDR-II. GeForceFX's "500 Mhz DDR-II" ram has a 250 Mhz cell array clock, 500 Mhz I/O clock, and 1000 Mhz data frequency.

Again, the I/O clock is what determines the actual Mhz rating of the ram, because that is the actual "interface" frequency between the ram and the bus.

[Bigus Dickus]

Joe is correct. DDR-II runs at half the clock and twice the bandwidth internally relative to DDR-I, but the I/O data clock is the same.

DDR-II 500MHz = DDR-I 500MHz as far as theoretical bandwidth is concerned.

[PreservedSwine]

Thanks, it's nice of you to post the explanation, it helps greatly.....I should've known a little bit of knowledge is sometimes worse than none at all.....:)

[ChrisW]

DDRII run externally at the exact same speed at DDRI. Internally, DDRII pushes twice the data per pin as DDRI but only runs at half the internal speed as DDRI.

Advantate of DDRII over DDRI:
chips run at a lower temperature because it's running at half the internal speed as DDRI. Because of the lower temperature, they are able to reach higher clockspeeds.

Disadvantage of DDRII over DDRI:
DDRII has a higher latency

Other than that they are exactly the same.

[Kruno]

Higher latency can be easily compensated for with buffering.

[nutball]

Originally posted by K.I.L.E.R
Higher latency can be easily compensated for with buffering.

And speculative pre-fetch.

[druga runda]

Originally posted by Joe DeFuria
Here's the deal. Use the illustration below as a guide.

http://www.lostcircuits.com/memory/ddrii/timing2.gif

Note that you must think of DDR and DDR-II as having three different frequencies. Memory Cell, I/O Clock, Data Rate. What's important to know is that the I/O CLOCK is the actual MHz of the ram chip....NOT THE memory cell array. 400 Mhz DDR-I has an I/O clock of 400 Mhz, and 400 Mhz DDR-II has an I/O clock of 400 Mhz.

(so in the above illustration, you're looking at a 100 Mhz DDR-I module, and a 200 Mhz DDR-II module. NOT a 100 Mhz DDR-II module!)

All DDR-II modules are in fact "quad-pumped" relative to the memory cell array. However, memory cell array frequencies have NOTHING to do with the stated frequency of the ram chip! It's an "internal" clock.

What does this mean?

It means that the "500 Mhz DDR-II" on the GeForceFX is running at an i/o clock frequency of 500 Mhz, for a data frequency of 1 GHz. This is exactly the same that a 500 Mhz DDR-I module would be. The difference is, the 500 Mhz DDRII module has in internal cell array frequency of 250 Mhz....HALF that of a 500 MHz DDR-I module.

So in short, it's helpful to think of it this way: DDR-I and DDR-II are both effectively double pupmped with respect to the reported chip clock frequency. And that is what is "most important" in terms of determining the bandwidth. However, DDR-II's "internal" cell array frequency is one half of the chip clock.

Did that help?

Now Joe... that was a quality explanation even though I never thought about this I understood :beer:

[Joe DeFuria]

Heh, well, "how DDR-ii" works confused the hell out of me initially :)

I realized that my basic problem was assuming that the memory clock was derived from the "internal core" clock. It's an easy mistake to make, because with SDR and DDR-I, memory clock just happens to equal the core clock.

So I figured that most people are confused for the same reason I was. So that's why I explained it the way I did. :thumbsup:

[druga runda]

And that might mean since we have DDR 400 already ... that might mean we could shoot for DDR II 800 soon with similar heat problems as DDR 400 only :eek:

Now that would be something.