RAM Timings Table question

[ck_kent]

Hey Guys,

I have 8GB of DDR2 RAM installed on my pc, a 2GB Apacer, 2x 2GB Kingston, and 2GB Transcend. I've attached below a screenshot of CPU-Z and I noticed the Apacer has different frequencies compared to the other two brands. Though all of them are DDR2 800 (PC2-6400), the Apacer has different values in the Timings Table. What are the implications of this? I really don't put too much thought about RAMs, knowing that they're all DDR2 800, I just installed it. Thanks!

[Bizarre™]

@ck_kent:

Copy the timings of Apacer and set them in your BIOS. Your Apacer sure has better memory timings than Kingston and Transcend.

Better yet, research the memory timings of Apacer and set them in your BIOS. If your PC is still stable after the change, then leave them at that timing.

[ck_kent]

Thanks Bizarre™. So, having low numbers in the Timings Table is good then? All this time, I thought higher numbers are better. :lol: So, what benefits would I get if I get to match my BIOS' to Apacer's timing? Would I notice any difference?

[Bizarre™]

@ck_kent:

Think of it this way: the lower the timing, the faster data is processed.

The equivalent of OC when it comes to RAM is the lowering of memory timing.

[ck_kent]

Thanks again Bizarre™, for clearing things up. :) And I thought Apacer is the lesser model I have. :lol:

I also found this from another forum while browsing, courtesy of user stevecn70:

Timings are how many clock cycles it takes to access a memory location. This is a "rough" table I put together to somewhat explain what is going on.

400mhz memory (800 Dual channel but each module is only 400):

CAS 5

400,000,000 / 5 = 80,000,000 memory locations per second

CAS 4

400,000,000 / 4 = 100,000,000 memory locations per second

CAS 3

400,000,000 / 3 = 133,333,333 memory locations per second

333mhz memory (667 Dual Channel)

CAS 5

333,333,333 / 5 = 66,666,666 memory locations per second

CAS 4

333,333,333 / 4 = 83,333,333 memory locations per second

CAS 3

333,333,333 / 3 = 111,111,111 memory locations per second

266mhz memory (533 Dual Channel)

CAS 5

266,666,666 / 5 = 53,333,333 memory locations per second

CAS 4

266,666,666 / 4 = 66,666,666 memory locations per second

CAS 3

266,666,666 / 3 = 88,888,888 memory locations per second

So as you can see, the two values go hand in hand with figuring out the speed. Let's look at some popular high MHZ DD3 memory that is out now and see if it is really faster than DDR2:

Lets try 1600 DDR3 which is 800mhz per module. The most common CAS frequency is 7:

800,000,000 / 7 = 114,285,714 memory locations per second

So as you can see, even though DDR3 is higher Mhz, it comes in at being only slightly faster than DDR2 800 CAS 4 and is actually slower than DDR2 800 CAS 3. However, DDR2 800 CAS 3 is Rare and expensive usually

[LeetPirate]

With 4 ram slots in use it may be difficult for all of them to sync when you lower the timings. It might become unstable and cause BSOD. You might have to up the voltage to stabilize the ram if your system starts crashing.

Also note that with higher frequency operating ram, it still accesses data faster than slower frequency ram despite the seemingly larger latency. So for example CAS9 1600MHz DDR3 can probably access data with lower latency than CAS7 800MHz DDR3 even though cas latency 9 is greater than cas 7. I can't remember the exact details off the top of my head as to why it occurs but it has to do with the increased number of clock cycles and rising/falling edges.

Also note that tRC = tRP + tRAS so you have to make sure those value add up if you plan to mess with them. You can see it from your screenshots, the 5+18=23 and the 6+18=24.

RAS to CAS and CAS can be adjusted independently. The most theoretical performance boost can be obtained from lowering the CAS latency but honestly people have done all this before and there are reliable benchmarks to suggest that the performance gain from lowering your ram latency is tiny. The performance reducing delay between CPU and RAM transactions comes from the memory controller itself so it depends on what CPU you are running. In other words this is how you can get faster performance from DDR3 as opposed to DDR2 running at the same speed, when the cpu architecture and memory controller are improved.

Overall if you reduce your ram latency and you do the benchmarks you may not even notice anything significant. Also Kingston is bottom of the barrel low quality memory modules so don't expect much from them. Even though they may use similar brand name ram chips on their modules as other memory brands (such as samsung, hyundai, elpida etc.), Kingston's implementation is flawed. What they do is take many lower quality chips and stack them together, push the voltage higher and increase the latencies because more chips = greater chance of sync errors. Which is why you might find Kingston ram using ram chips filled on both sides of their modules, like 8 chips, whereas better quality ram manufacturers use less ram chips with higher density so that overall the module will use less power and be more stable with lower latencies. People always think Kingston is some great brand but they just have strong marketing, with strong marketing you can sell water to the ocean.

[Nima]

Yeah, that forum says the truth, timing and frequency are both very important in RAM but they are in contrast with each other.

Higher Frequency = Higher Timings

As you see very old RAMs have very low frequency along side a very low timings.

In my opinion, timing is a bit more important than frequency, but instead of upgrading to 8GB of DDR2 RAMs, I suggest you to upgrade to DDR3, something like 1333 @ CL 6 / 1600 @ CL 7 (but you may need to overclock them a bit, I have also heard of G.Skill DDR3 RAMs of 1600 @ CL 6).

And one point for brand, Crucial and after that, Corsair produce the very best RAMs, next time buy from one of them (BallistiX Tracer of Curcial and Dominator-GT of Corsair). ;)

[ck_kent]

I appreciate all your answers, guys. I'm not really planning on overclocking my RAM especially when performance benefit is not noticeable at all. I was just curious about it when I saw the differences on CPU-Z. All the information you guys gave me has contributed a lot to my knowledge about RAMs.

Kingston really is popular here. I guess they really have good marketing. They also have cheaper compatible replacement RAM for branded computers here as well. I guess that's another reason. We only have several brands available here, Kingston being the most popular one, Transcend, Apacer, Corsair, PQI, A-Data, are the common ones.

Among the brand names I mentioned above which, in your opinion or based from experience, is the best one in general (entry-level to gaming)?

[LeetPirate]

We only have several brands available here, Kingston being the most popular one, Transcend, Apacer, Corsair, PQI, A-Data, are the common ones.

Among the brand names I mentioned above which, in your opinion or based from experience, is the best one in general (entry-level to gaming)?

I read alot of reviews about ram on various sites, especially newegg. Based on my research and personal experience I can tell you that the best ram in terms of both system compatibility and performance is from Mushkin and A-Data. People often think A-Data is crappy ram probably because their brand name isn't as catchy or popular but I have used them often and the most significant reasons why I continue to use them is because they provide proper specifications. A-Data provides a proper operating voltage range and their ram operates close to spec voltage (1.5V for ddr3 etc.) whereas other brands of similar speed ddr3 tell you the voltage requirement of 1.6V and higher which is automatic sign of factory pushing lower quality chips to their limits. The A-Data ram is also priced well, I often buy higher speed ram from A-data at the same cost of the ram I actually need from other brands, so for example I buy A-Data gaming series ddr3 capable of 1600Mhz or 2000Mhz when my system only requires 1333Mhz and it costs around the same as buying 1333Mhz ram from another brand. The reason I do it is so that I know the ram can operate at a higher speed by default so it does not limit my overclocking abilities.

The first time though I ended up getting A-Data ddr2 with pink heatsinks but I figured nobody would really see it anyway and I bought them on a really good sale, lol. Seriously though their heatsinks are high quality. Mushkin is also high quality but their prices are higher, Mushkin warranty is one of the best in the world though. I have reduced the latency on both A-Data and Mushkin ram modules without adjusting the voltage so their modules are of high quality. You must always look out for the voltage specification of the ram you buy. The ones with a voltage range like 1.50V - 1.75V means that they can work at 1.5V for stock speeds and can tolerate up to 1.75V if you plan to OC, the ones that actually state a range are better quality ram than the ones that just print 1 voltage (especially if that 1 stated voltage is high). Usually it's good to operate stock ram between 1.5V to 1.55V, the little extra is to compensate for fluctuations of the motherboard supply, which affects stability.

Of the other brands you mentioned Apacer and Transcend are probably good performers as well but once A-Data is an option I will choose that. Corsair is one of the top brand names but I don't like their product categorization, it is vague and misleading and intentionally confuses the customer into thinking they need some higher more expensive edition. Take for example they make you think that if you have an i3 then you must use their i3 branded ram modules, and similar for i5 and AMD and so on. They also have too many variations of latencies and heatsink editions which leads me to think that they just want to justify adding on higher costs for giving you a minor adjusted latency ram with a different heatsink. While their ram modules are probably good performers, I don't agree with their business practices or their high prices. It's not like they make their own ram chips anyway so what justifies the added costs? Their brand name and their heatsink designs. There are only a few actual ram chip fabrication facilities, I think SpekTek, micron and crucial are the same company, there is Elpida, Hyundai, Samsung, Toshiba, and I believe Transcend makes their own chips but not entire sure.

[ck_kent]

Thank you very much, LeetPirate! As always, I get very well informed with your answers. :) I might consider A-Data considering it's one of the cheapest here of all the brands I mentioned. I guess cheaper doesn't always mean lesser quality. We so got used to the mentality of "more expensive means higher quality" or "you get what you paid for" but there are always some exception to the rule.

[DKT27]

I won't recommend Transcend or Kingston. Used Transcend on older computer and had some real problems in 1-2 years. Using Kingston now, lets say life is full of learning from mistakes. :)

Also, AIDA says the kingston ram is running at v1.55 to v1.565 with overclocking disabled. Not that it matters, just saying. :P

[teamacc]

The apacher ram could operate faster if it were on its own, but removing the other 6gb would slow your system down more than the lower timings would speed it up. Do not mess with the timings, as the timings are forced to all your memory banks, so all memory has to be that fast. If its not, it could become unstable, causing data-loss.

You wont notice the lower timings unless you are overclocking your pc to run super-pi under 10 seconds.