A couple of questions about DDR-II

[tieros]

I just reread the ATi press release about DDR-II, as well as the JEDEC whitepaper, and I have a couple of questions (3 actually).

1) What was the performance difference on the 9700 using DDR-II vs. the standard DDR ? They didn't mention if it was faster, slower, or if it even made a difference.

2) Have any OEMs announced a DDR-II version of the 9700? It seems like it would be a good differentiator if it offers a speed boost.

3) Does DDR-II signalling increase the effects of bus skew and crosstalk over standard DDR at the same clock rate? If so, could that be a reason why nVidia opted for a narrower bus?

Ok, so that was actually 4 questions

TIA

 

[jvd]

okay i'm going to try here. I believe there is no big diffrence over normal ddr memory except that it scales higher and it has a higher lantancy ? I think i'm right in saying that.

 

[Tahir2]

You are right but didnt answer any of his questions.

3rd possibly answerable by some guys here.. the rest no one will talk about until there is more information.

DDR I has lower latency so at same bandwidth to DDR II - DDR I is faster.

Anyone else?

 

[arjan de lumens]

At the same clock rate, I would expect DDR-II to have slightly less skew/crosstalk problems; because all clock/strobe lines are differential pairs, they should be less vulnerable to noise than the single-ended strobes of DDR-I, thus giving slightly better timing margins.

Which one of DDR-I and DDR-II is faster at a given clock speed is a bit unclear: on one side, DDR-II has higher latency and a minimum burst length of 4 - on the other side, DDR-II has more memory banks (reducing the number of page breaks) and faster bus turnaround. So it will depend on the exact usage pattern.

 

[Joe DeFuria]

1) Performance difference should only really be directly related to clock rate. I think the slowest DDR-II modules available are 400 Mhz. (800 MHz effective). That may also be the fastest DDR modules available as well (400 Mhz). You'd probably be hard pressed to see any difference between a 400 Mhz DDR and a 400 Mhz DDR-II implementation.

2) No, no OEMs have opted for this yet. While it may be a differentiator in certain performance aspects (400 Mhz DDRII vs. 310 Mhz DDR), it would also be a pretty big cost difference. Unless ATI supplies the OEMs with a version of the R-300 with a "guaranteed clock" of about 400 Mhz to go with that memory, you probably won't see OEMs go with that memory. The performance delta without the core clock increase might not justify only upping the ram clock.

3) No idea.

 

[jvd]

well i figured i answered question 1. The only plus to ddr2 is it clocks higher so at the same speed they should both be the same mabye ddr1 being faster.

 

[Sage]

i suspect ATi will release a 400MHz version of the R300 late in the year, and have it on shelves along with DDR-II memory about the same time the GeForceFX comes out. After all, most R300s will do near 400 with decent cooling - hook up a funky GFFX-type cooler and WHAM nVidia takes it in the ass...

hmm but then we hear NV35 is about to come back from first tape-out. with the 100 days from tape-out to launch that could mean the NV30 will truely be the shortest living grafix card ever... but then ATi will be ready with a .13u version of the R300, or maybe even a R350 (which could very well be little more than a .13u R300) and by then PVR, Matrox, Trident, SIS, 3DLabs may very well have products ready to launch with .13u DX9 flooding the low end DX9 market, and taking a jab at the high end gaming and workstation markets leaving little room for the looser of ATi vs nVidia to maneuver in....

will 2003 see the death of another grafix company- low or high end? Probably not, nVidia and ATi have lots of gold in their coffers to draw on while they try and make a comback from a possible defeat and, the others mentioned all have unknown resources which could be quite large. But, just because a company has money to keep playing the game doesn't mean they wont decide to play in another playground where the neighbourhood bullies aren't so big.

 

[Dave Baumann]

hmm but then we hear NV35 is about to come back from first tape-out.

Where on earth did that one come from??

 

[Nappe1]

will 2003 see the death of another grafix company- low or high end? Probably not, nVidia and ATi have lots of gold in their coffers to draw on while they try and make a comback from a possible defeat and, the others mentioned all have unknown resources which could be quite large. But, just because a company has money to keep playing the game doesn't mean they wont decide to play in another playground where the neighbourhood bullies aren't so big.

as we finns say "kuules puhemies, selitäs ny uuestaan niin että tyhymempikin tajuaa." which translates about like this "how about another round of this chapter, but this time so that even the stupido can get it."

so are you stating that in 2003 some company dies, but it isn't ATI nor nVidia?

 

[DadUM]

so are you stating that in 2003 some company dies, but it isn't ATI nor nVidia?

There are 3 other "minor" players in the consumer graphics business. Intel with integrated chipsets accounts for a staggeringly large portion of all graphics chips and isn't going out of business in 2003, even if Itanic is taking on water that fast. SiS has the Xabre line and some integrated parts. There parts have always been low end and I assume Sis, while not getting rich, isn't losing money on this venture. Besides, they need something to use as a basis for integrated chipsets.

Then there is Matrox. Financials unkown as they are private, but bad rumors abound. They did also last year. Matrox appears to be a niche role, and as long as they haven't spent too much money on products which will never make back life-cycle costs, they will stay around. And judging by past history, they should manage to stay alive.

As for nVidia, they are the current market leader with lots of income from mid-low end designs. That won't dissappear overnight. And they have chipsets to pull in a little more cash. And this is all assuming GFFX is a complete failure (not just "dissappointing"). 2003 should not see the death of nVidia.

ATI lost a lot of ground in terms of overall shipments over the past couple years. They have the highground with the R300, but I personally still wonder how much of a money maker it is for them. The low end line up has a solid anchor and they need to get the mid covered. But overall, they have a good lineup and sales are on the rebound. They will not go out of business unless something catastrophic happens.

There you have it. No players are set to dissappear this year. Just my outsider analysis.

 

[tieros]

Thanks for all the great answers, everyone!

I've always been a big fan of Matrox, but I think the triple whammy of weak economy (especially Canada), even weaker IT expenditure, and the disappointing rollout of the Parhelia might just be too much to recover from.

We should know by the end of Q1 2003 if there's a substainable recovery possible. I'm seeing IT work ramping up substantially (I'm an Oracle consultant) over the past couple of weeks, and it looks like there will be some serious spending in January, but I have no idea how long it will last.