DDR1 / DDR2 to coexist

Yay, my Level IIs will be able to move to any new system in the foreseeable future. Thanks for the info.
 
makes sense you know, ddr 1 really hasnt finished filling its full potential
 
This is of course, very predictable. Intel wants to avoid another RDRAM, with memory pricing them out of the market, IE Willamette.
 
Originally posted by scientificTHEgreat
and you think this because why?
Click to expand...
It's because DDR can be found at speeds up to 550 MHz rated, yet chipsets only support DDR at up to 400 MHz officially.
 
well i guess i see your point but ddr being overclocked a few mhz more is not that big of a deal to me

seeing more done per cycle and a few hundred higher mhz is something to actually talk about
 
DDR-II isn't doing any more per cycle, but it IS suffering from increased latency. You can see the effect of memory latency on benchmarks where the Athlon 64 has a significant advantage over the Pentium 4. This is because there is less latency between the CPU and memory, lending to higher performance. Now compare DDR-II, which is currently running at awful timings (3-3-7-3, if I recall correctly). Right now, at this point in the game, DDR-II looks to be at a disadvantage.

And DDR isn't being overclocked. The DDR 550 I referred to earlier is guaranteed to run at that speed, just like DDR 400 is now. It just needs a chipset or memory controller that can take advantage of the higher clock.
 
So your saying because the manufacturer says its guaranteed to run at the speed that its not being overclocked..hmm i question that

yup you are correct about the higher latencies

good info here about how DDRII works take a look and youll see that there IS actually more work done..



So per driven clock cycle on the memory controller, the DRAM device can work with two bits of data, twice per clock (DDR), using the mechanism outlined above. Here's how DDR-II takes a different approach with its data buffers, to offer double the data frequency (effective bandwidth) for the same driven clock.


DDR-II
If regular DDR DRAM devices can work two bits of data, twice per clock cycle, DDR-II offers double the effective bandwidth by essentially combining two regular DRAM cores on the same, new, DDR-II DRAM device. The data buffers on a DDR-II device are now double data rate themselves. With DDR and a 100MHz driven clock, the data buffers on the DDR DRAM device also run at 100MHz. The DDR principle then makes that 200MHz. With DDR-II the data buffers will run at 200MHz at the same driven clock. This allows them to essentially work on four bits of data per clock cycle, either read or written. Finally, apply the DDR concept to that, where it's done twice per clock, once on the rising edge and once on the falling edge, and you get an effective data frequency of 400MHz for that 100MHz initial clock frequency.

That's the essence of DDR-II. Here's it layed out simply.

DDR : 100MHz driven clock -> 100MHz data buffers -> DDR applied -> 200MHz final data frequency
DDR-II: 100MHz driven clock -> 200MHz data buffers -> DDR applied -> 400MHz final data frequency

The same essential principles still apply, memory timings are still in effect, but will take a different form from the current timings we've just become happy talking about with DDR DRAMs and memory devices.


got that from hexus.net

http://www.hexus.net/content/reviews/review.php?dXJsX3Jldmlld19JRD03MDImdXJsX2hpc3Rvcnk9LTI

enjoy
 
Here's a quick DDR vs. DDR-II comparison:

DDR = 184-pins 2.5v 2-2-2
DDR200\PC1600 = 2x 100MHz = 1.6 GB\sec 10ns latency
DDR266\PC2100 = 2x 133MHz = 2.1 GB\sec 7.5ns latency
DDR333\PC2700 = 2x 166MHz = 2.7 GB\sec 6ns latency
DDR400\PC3200 = 2x 200MHz = 3.2 GB\sec 5ns latency
DDR500\PC4000 = 2x 250MHz = 4.0 GB\sec 4.5ns latency

DDR-II = 232-pins 1.8v 4-4-4
DDR-II\400 = 4x 100MHz = 3.2 GB\sec 10ns latency
DDR-II\533 = 4x 133MHz = 4.3 GB\sec 7.5ns latency
DDR-II\600 = 4x 150MHz = 4.8 GB\sec 6.5ns latency
DDR-II\800 = 4x 200MHz = 6.4 GB\sec 5ns latency
DDR-II\1066 = 4x 266MHz = 8.6 GB\sec 4ns latency

I figured, what the heck, lets extrapolate the odd frequencies here, too...

SDR = 172-pins 3.3v 1-1-1
SDR\PC66 = 66MHz = 0.53 GB\sec 15ns latency
SDR\PC75 = 75MHz = 0.60 GB\sec 13.5ns latency
SDR\PC83 = 83MHz = 0.67 GB\sec 12ns latency
SDR\PC100 = 100MHz = 0.80 GB\sec 10ns latency
SDR\PC112 = 112MHz = 0.93 GB\sec 9ns latency
SDR\PC124 = 124MHz = 1.00 GB\sec 8ns latency
SDR\PC133 = 133MHz = 1.05 GB\sec 7.5ns latency
SDR\PC150 = 150MHz = 1.20 GB\sec 6.7ns latency
SDR\PC166 = 166MHz = 1.35 GB\sec 6ns latency
SDR\PC175 = 175MHz = 1.40 GB\sec 5.7ns latency
SDR\PC183 = 183MHz = 1.50 GB\sec 5.5ns latency
SDR\PC200 = 200MHz = 1.63 GB\sec 5ns latency
SDR\PC212 = 212MHz = 1.75 GB\sec 4.7ns latency
SDR\PC224 = 224MHz = 1.80 GB\sec 4.5ns latency
SDR\PC233 = 233MHz = 1.85 GB\sec 4.3ns latency
SDR\PC250 = 250MHz = 2.00 GB\sec 4ns latency
SDR\PC266 = 266MHz = 2.15 GB\sec 3.8ns latency
SDR\PC275 = 275MHz = 2.20 GB\sec 3.6ns latency
SDR\PC283 = 283MHz = 2.30 GB\sec 3.5ns latency
SDR\PC300 = 300MHz = 2.40 GB\sec 3.3ns latency

DDR = 184-pins 2.5v 2-2-2
DDR133\PC1000 = 2x 66 = 1.05 GB\sec 15ns latency
DDR150\PC1200 = 2x 75 = 1.20 GB\sec 13.5ns latency
DDR166\PC1300 = 2x 83 = 1.35 GB\sec 12ns latency
DDR200\PC1600 = 2x 100 = 1.60 GB\sec 10ns latency
DDR233\PC1800 = 2x 112 = 1.85 GB\sec 9ns latency
DDR250\PC2000 = 2x 124 = 2.00 GB\sec 8ns latency
DDR266\PC2100 = 2x 133 = 2.10 GB\sec 7.5ns latency
DDR300\PC2400 = 2x 150 = 2.40 GB\sec 6.7ns latency
DDR333\PC2700 = 2x 166 = 2.70 GB\sec 6ns latency
DDR350\PC2800 = 2x 175 = 2.80 GB\sec 5.7ns latency
DDR366\PC3000 = 2x 183 = 3.00 GB\sec 5.5ns latency
DDR400\PC3200 = 2x 200 = 3.25 GB\sec 5ns latency
DDR433\PC3500 = 2x 212 = 3.50 GB\sec 4.7ns latency
DDR450\PC3600 = 2x 224 = 3.70 GB\sec 4.5ns latency
DDR466\PC3700 = 2x 233 = 3.75 GB\sec 4.3ns latency
DDR500\PC4000 = 2x 250 = 4.00 GB\sec 4ns latency
DDR533\PC4300 = 2x 266 = 4.30 GB\sec 3.8ns latency
DDR550\PC4500 = 2x 275 = 4.40 GB\sec 3.6ns latency
DDR566\PC4600 = 2x 283 = 4.60 GB\sec 3.5ns latency
DDR600\PC4800 = 2x 300 = 4.80 GB\sec 3.3ns latency

DDR-II = 232-pins 1.8v 4-4-4
DDR-II\266 = 4x 66MHz = 2.1 GB\sec 15ns latency
DDR-II\300 = 4x 75MHz = 2.4 GB\sec 13.5ns latency
DDR-II\333 = 4x 83MHz = 2.7 GB\sec 12ns latency
DDR-II\400 = 4x 100MHz = 3.2 GB\sec 10ns latency
DDR-II\466 = 4x 112MHz = 3.6 GB\sec 9ns latency
DDR-II\500 = 4x 124MHz = 4.0 GB\sec 8ns latency
DDR-II\533 = 4x 133MHz = 4.3 GB\sec 7.5ns latency
DDR-II\600 = 4x 150MHz = 4.8 GB\sec 6.7ns latency
DDR-II\667 = 4x 166MHz = 5.4 GB\sec 6ns latency
DDR-II\700 = 4x 175MHz = 5.6 GB\sec 5.7ns latency
DDR-II\733 = 4x 183MHz = 6.0 GB\sec 5.5ns latency
DDR-II\800 = 4x 200MHz = 6.4 GB\sec 5ns latency
DDR-II\866 = 4x 212MHz = 7.0 GB\sec 4.7ns latency
DDR-II\900 = 4x 224MHz = 7.2 GB\sec 4.5ns latency
DDR-II\933 = 4x 233MHz = 7.4 GB\sec 4.3ns latency
DDR-II\1000 = 4x 250MHz = 8.0 GB\sec 4ns latency
DDR-II\1066 = 4x 266MHz = 8.6 GB\sec 3.8ns latency
DDR-II\1100 = 4x 275MHz = 8.8 GB\sec 3.6ns latency
DDR-II\1133 = 4x 283MHz = 9.2 GB\sec 3.5ns latency
DDR-II\1200 = 4x 300MHz = 9.6 GB\sec 3.3ns latency

It was interesting to see DDR-II was supposed to start at 83MHz, a 333MHz SDR-equivalent, back before AMD launched its DDR166 support. Its as if the memory industry decided "thats more bandwidth than they'll ever need..." and they completely forgot about the memory latency. I couldn't imagine what would have happened had DDR been released at either the sluggish PC1000 or PC1300 levels, if the public would have laughed it off as a joke. So what if that memory is equivalent in bandwidth to SDR memory twice its speed, it sooooooooo slow to get transfers started... Its still a joke to even consider releasing DDR-II below an 800MHz SDR-equivalent.
 
Originally posted by scientificTHEgreat
So your saying because the manufacturer says its guaranteed to run at the speed that its not being overclocked..hmm i question that

yup you are correct about the higher latencies

good info here about how DDRII works take a look and youll see that there IS actually more work done..
[...]
Click to expand...
Define overclocked and you may be right or you may be wrong. If the memory controller can handle the RAM at a given frequency, and if the RAM is rated at that frequency, it's not overclocked, in my definition. What's yours?

As you can see from that quoted material, DDR and DDR-II of the same effective clock do the same work, and from MadRat's quoted numbers, provide the same amount of bandwidth per effective clock. The only problem is, DDR-II at this point has higher latency as a whole.
 
Those are all theoretical numbers. In real world terms the 300MHz rate for SDR, 150MHz DDR300, and 75MHz DDR-II/300 will be along the lines of 1.8 GB/sec, 1.05 GB/sec, and .6 GB/sec. But run them all at 300MHz clock and you get a realistic 1.5 GB/sec, 1.96 GB/sec, and 3.4 GB/sec performance.

You probably figure, 3.4 GB/sec only for DDR-II/1200!? It may be rated at 9.6 GB/sec but it won't perform anywhere near that unless the trace lengths from memory to controller are nil.
 
DDR1
=============================================
the DRAM device can work with two bits of data, twice per clock (DDR)


DDR2
=============================================
Here's how DDR-II takes a different approach with its data buffers, to offer double the data frequency (effective bandwidth) for the same driven clock.

[IMPORTANT]This allows them to essentially work on four bits of data per clock cycle, either read or written.
 
If the DDR is rated to run at DDR500 and you run it at that speed on, say, an IC7, it's the bus that's being O/Ced. If you run DDR400 @ DDR500 you are overclocking both the memory and the bus.
 
It depends on individual chipset implementations AND the almighty marketing decisions. Intel's own chipsets appear to be exclusively AGP or exclusively PCI Express, but other manufacturers may not necessarily be limited in this regard. It is a lot more of a hassle from an engineering standpoint, routing more traces than what is really needed.
 
You must log in or register to reply here.
Back
Top