RAM - Memory Technology Overview
by Jarred Walton on September 28, 2004 12:05 AM EST- Posted in
- Memory
Memory Latencies Explained
One big question that remains is latency. All the bandwidth in the world will not help if you have to wait forever to get the needed data. It is important to note, however, that higher latencies can be compensated for. The Pentium 4, for example, has improved buffering, sophisticated prefetch logic, and the ability to have many outstanding memory requests. It loves bandwidth, and performance has been helped substantially by increasing the bus speeds, even with higher memory latencies. Graphics chips also tend to be more forgiving of higher latencies. Any design can be modified to work with higher or lower latencies, of course; it is but one facet of the overall goal which needs to be addressed. Still, the question remains, how does memory latency relate to timings and bandwidth?The simple answer is that it is directly related to the memory timings, but you cannot compare timings directly. The reason for this is that the memory timings are relative to the base clock speed of the RAM - they are the number of memory clock cycles that each operation requires. For DDR memory, this means that the cycle time is calculated using one half of the data transfer speed. PC3200 DDR memory has a 64-bit bus that transfers up to 3200 MB/s. Converting that to a clock speed means converting bytes to bits (multiply by eight), then divide by that bus width, and we get the effective clock speed; the base clock speed is half the effective clock speed.
PC3200:
3200 MB/s * 8 bits = 25600 Mb/s
25600 Mb/s / 64-bits = 400 MHz
400 MHz / 2 = 200 MHz base clock speed
Other memory types may use quad or even octal data rates, but if we convert those into the base clock speed, we can compare latencies. Where timings are listed in clock cycles, latency is listed in nanoseconds (ns). A CL of 2.0 sounds better than a CL of 5.0, but depending on the memory clock, it may actually be closer than we would at first expect. By converting all of the timings into nanoseconds, we can compare performance. We will save detailed comparisons for the next installment, but as an example, suppose we have two memory types - one with a CL of 4.0 and a base clock speed of 333 MHz, and the second with a CL of 2.5 and a base clock speed of 200 MHz.
| CL | Clock Speed | Cycle Time | Real Latency |
| 2.5 | 200 MHz | 5.0 ns | 12.5 ns |
| 4.0 | 333 MHz | 3.0 ns | 12.0 ns |
In this specific example, we see that even with a CL that's 60% higher, the effective latency can actually end up being slightly slower. This is something that we will examine further in the next article of this series.
Facebook
Twitter
RSS
22 Comments
View All Comments
Lynx516 - Tuesday, September 28, 2004 - link
Your description of how SDRAM works is wrong. you do not bust down the rows as your artcle implys but instead it bursts along the columns.The whole column is sent imeadiatly but the other columns in the burst are not and are sent sequencially (idealy not quite the case if you want to interleave them).
Comparing Banks to set associativity is probably counter productive as most of your reader wont fully under stand how it works. And infact comparing banks to set associativity is a bad annalogy. A better one would be just to say taht the memmory space in the chip is split up into banks.
On top of this you have referred to a detailed comparsion of DRAM types. Even though there are many different types of DRAM most are not that interesting or used that much in PCs. I also assume that as you have said this you will not be talking about SRAM or RDRAM in forth coming articles which highlight the different approaches that can be taken when designing a memory sub system. (SRAM the low latency, high bandwidth but low density, RDRAM the serial approach)
I assume you are going to talk abit about how a memory controller works as they are one of the most complex components in a PC (more complex than the exectution core of a CPU) but you have not refered to any plans to talk about memory controller and how the type of memory you are using affects the design of a memory controller.
All in all a pretty confusingly written article. If you want a DRAM for beginners arstechnica have two good articles (though one is fairly old but atleast correctly and CLEARLY describes how SDRAM works).
Resh - Tuesday, September 28, 2004 - link
I really think that some diagrams would help, especially for novices like #10. Other than that, great article and hope to see the follow-ups soon.Modal - Tuesday, September 28, 2004 - link
Great article, thanks. I like these "this is how the pieces of your computer work" articles... very interesting stuff, but it's usually written in far too complicated a manner for a relative novice like me. This was quite readable and understandable however; nice work.danidentity - Tuesday, September 28, 2004 - link
This is one of the best articles I've seen at Anandtech in a long while, keep up the good work.deathwalker - Tuesday, September 28, 2004 - link
I..for one, would rather have 1 GB of CL 2.5 high quality memory than 512 MB of CL 2 high quality memory. I'm conviced that in this instance quantity wins out over speed.AlphaFox - Tuesday, September 28, 2004 - link
where are the pictures? ;)Pollock - Tuesday, September 28, 2004 - link
Excellent read!mino - Tuesday, September 28, 2004 - link
Sry for triple post but some major typpos:"1) buy generic memory until your budget could afford no more than 512M DDR400"
should be:
"1) buy generic memory until your budget could afford more than 512M DDR400"
and
"Goog"(ROFL) should be "Good"
onother -> another
Hope that's all ;)
mino - Tuesday, September 28, 2004 - link
OK, 3 rules ;) - I added 3rd after some thought.mino - Tuesday, September 28, 2004 - link
#2 You are missing one important point. That is, unless You can(want) afford at least 512M high quality RAM, it makes NO SENSE to buy 256M DDR400 CL2 since there are 2 basic rules:1) buy generic memory until your budget could afford no more than 512M DDR400
2) then spend some aditional money for brand memory
3) then go 1G and only at this point spent all additional money for better latencies and so on.
Also do remember that at many shops(here in Slovakia) there is 3 or 4 yrs warranty for generic memory(like A-DATA) and also if you have major problems with compatibility they will usually allow you to choose different brand/type for your board for no additional cost except price difference. Also in case the memory works fine with onother board.
Also Twinmos parts have 99month warranty (for price 10% higher than generic). That speaks for itself.
Except this little missing part of reality,
Goog work Jarred.