DDR4 CAS Latency

[RazrLeaf]

Any ideas how high/low CAS Latency on 2133 CL15 DDR4 will be? I was looking at a some information on Wikipedia, and was wondering how DDR4 would compare.

 

[Frick]

CL15 = CAS latency 15.

As always, it will not be worth it at first (I assume, cadaveca will correct me if I'm wrong (I see you!)) but eventually it will.

 

[cadaveca]

CL15 = CAS latency 15.

As always, it will not be worth it at first (I assume, cadaveca will correct me if I'm wrong (I see you!)) but eventually it will.

All things in due time...

 

[EarthDog]

Due time... That is for those under NDA... oh wait... so am I...

Being more serious, one of the big take aways from DDR4 is their low voltage. You will see (check out the Gskill press release) 1.2v and CL15 at 2133 and CL16 1.35v at obnoxious 3200MHz speeds versus the 1.65v+ it takes to get there now (granted with lower latency in a lot of cases).

Bandwidth isn't an issue anyway so, unless they offer something more, it, like quad channel in most cases, is pretty MEH.

 

[RazrLeaf]

Well, I asked my question in a foolish way.
Will the latency in ns be the same as they are for respective DDR3 modules given the same clock speed and CL?

 

[LaytonJnr]

Well, I asked my question in a foolish way.
Will the latency in ns be the same as they are for respective DDR3 modules given the same clock speed and CL?

What you are referring to there is known as RAM 'access time', or more casually as RAM 'speed' or 'performance', which is basically the time it takes in nanoseconds for the RAM to locate a single piece of information and make it available to the processor (a very rough definition). Latency is a reference to CAS timings, which refers to the delay between the memory controller telling the RAM to 'find' the information, and the information being made available.

The general way to work out RAM access time is: ( CL / Frequency ) * 1000.

So, for 2133MHz CL15 DDR4 memory, ( 15 / 2133 ) * 1000 = 7.03ns

And for 1600MHz CL9 DDR3 memory, ( 9 / 1600 ) * 1000 = 5.63ns

So actually, DDR3 memory is faster in terms of access time between these two examples (as AFAIK 2133MHz is meant to be the standard DDR4 RAM frequency). As many people have already said, DDR4 RAM probably will eventually have better access time than DDR3 in due time.

Layton

 

[cadaveca]

What you are referring to there is known as RAM 'access time', or more casually as RAM 'speed' or 'performance', which is basically the time it takes in nanoseconds for the RAM to locate a single piece of information and make it available to the processor (a very rough definition). Latency is a reference to CAS timings, which refers to the delay between the memory controller telling the RAM to 'find' the information, and the information being made available.

The general way to work out RAM access time is: ( CL / Frequency ) * 1000.

So, for 2133MHz CL15 DDR4 memory, ( 15 / 2133 ) * 1000 = 7.03ns

And for 1600MHz CL9 DDR3 memory, ( 9 / 1600 ) * 1000 = 5.63ns

So actually, DDR3 memory is faster in terms of access time between these two examples (as AFAIK 2133MHz is meant to be the standard DDR4 RAM frequency). As many people have already said, DDR4 RAM probably will eventually have better access time than DDR3 in due time.

Layton

That's a good way to look at it, although, truly, yyou'd be better off comparing 1066 MHz or 1333 MHz DDR3 vs. 2133 DDR4, since these are "base speeds" each technology will have launched at. Comparing DDR3 1600 would be comparable to a higher DDR4 speed than 2133 MHz... perhaps DDR4-2666 or DDR4-3000 might be the more accurate compare.

If you are simply comparing what's available now, it would be more prudent, IMHO to maybe compare DDR3-2400 to DDR4-3000 or DDR4-3200. Or you could go by price... I'm not sure where the truly accurate comparison is at this point.

And no, EarthDog, I do not have an NDA on DDR4. However, I do feel that all info must be provided in proper context, and that context does have some limitations for me at this time.

 

[RazrLeaf]

I just wanted to know of they used the same units/measurements when reporting specs. That was most helpful (and educational). Thanks.

 

[EarthDog]

Yes.

They didn't switch the barometer(how they measure) in the middle of the game... That is like saying they measure an automobile's horsepower by using baseballs, LOL!


Dave, I meant the platform, not DDR4 specifically.

 

[cadaveca]

Yes.

They didn't switch the barometer(how they measure) in the middle of the game... That is like saying they measure an automobile's horsepower by using baseballs, LOL!


Dave, I meant the platform, not DDR4 specifically.

NO Intel NDA either...

 

[LaytonJnr]

That's a good way to look at it, although, truly, yyou'd be better off comparing 1066 MHz or 1333 MHz DDR3 vs. 2133 DDR4, since these are "base speeds" each technology will have launched at. Comparing DDR3 1600 would be comparable to a higher DDR4 speed than 2133 MHz... perhaps DDR4-2666 or DDR4-3000 might be the more accurate compare.

If you are simply comparing what's available now, it would be more prudent, IMHO to maybe compare DDR3-2400 to DDR4-3000 or DDR4-3200. Or you could go by price... I'm not sure where the truly accurate comparison is at this point.

And no, EarthDog, I do not have an NDA on DDR4. However, I do feel that all info must be provided in proper context, and that context does have some limitations for me at this time.

I did a little bit of digging, and the base level for DDR3 on launch was 1066MHz CL7 and 1333MHz CL8, and the base level for DDR4 is claimed to be to the tune of 2133MHz CL13. Time for some more math.

So for DDR3, the RAM access times are 6.57ns and 6.00ns respectively. And for DDR4, the RAM access time is 6.09ns. Hence, the entry level 'access times' of both technologies is comparably similar at the base level at launch.

Layton

 

[EarthDog]

NO Intel NDA either...

I guess TPU is missing the boat... or you are being an ass to me... Id bet on the later.

 

[cadaveca]

I guess TPU is missing the boat... or you are being an ass to me... Id bet on the later.

Nah, I will have my launch-day coverage. I just do not have NDAs for most of what the platform offers. I'm only not reporting anything worthwhile early out of respect for the companies involved.

I just manage to do this for every launch, get what I need without having to sign NDAs, and without having to use ES parts. Any site should be able to do the same, IMHO. If other sites are signing NDAs, and are restricted to ES parts for launch day reviews...it's THEM, and their readers...that are missing out.

Or perhaps I have a better relationship with Intel. Who knows. ROFL.

 

[EarthDog]

Looking forward to your 'launch day, no NDA, no ES' coverage!

 

[ZenZimZaliben]

Every new generation of ram does this. The difference in CAS rating is due adjustments in higher/lower bandwidth and manufacturing tolerances. I am totally excited for Haswell-E, X99, DDR4! Been waiting to upgrade and this will be it.

DDR: CAS 2, 3 - Command Rate: 166, 200
DDR2: CAS 4, 6 - Command Rate: 333, 400, 533
DDR3: CAS 5, 11 - Command Rate: 533, 666, 687, 800, 1066, 1200
DDR4: CAS 10, 22 - Command Rate: 1600, 1866, 2133, 2400, 2666, 3200, ???

 

[Hilux SSRG]

What you are referring to there is known as RAM 'access time', or more casually as RAM 'speed' or 'performance', which is basically the time it takes in nanoseconds for the RAM to locate a single piece of information and make it available to the processor (a very rough definition). Latency is a reference to CAS timings, which refers to the delay between the memory controller telling the RAM to 'find' the information, and the information being made available.

The general way to work out RAM access time is: ( CL / Frequency ) * 1000.

So, for 2133MHz CL15 DDR4 memory, ( 15 / 2133 ) * 1000 = 7.03ns

And for 1600MHz CL9 DDR3 memory, ( 9 / 1600 ) * 1000 = 5.63ns

So actually, DDR3 memory is faster in terms of access time between these two examples (as AFAIK 2133MHz is meant to be the standard DDR4 RAM frequency). As many people have already said, DDR4 RAM probably will eventually have better access time than DDR3 in due time.

Layton

That's an easy way to do quick comparisons, thanks for sharing.

 

[X71200]

Unless you're comparing shit DDR4 to good DDR3 expect DDR4 to be faster, not slower. It's the bandwidth that matters the most for performance and not CAS latency, as long as speeds and timings go in line. There are already high end DDR4 RAM such as Dominator Platinums made with extremely high frequencies. Here is an example of comparison, as you can see from the pic TridentX 2933 did better than it's 2600 version even though it has higher CAS latency.

 

[LAN_deRf_HA]

That pic isn't really going to sell anyone on speed before cas, in fact the review it comes from would seem to show the opposite. Especially if you want to do well benchmarking.

 

[cadaveca]

That pic isn't really going to sell anyone on speed before cas, in fact the review it comes from would seem to show the opposite. Especially if you want to do well benchmarking.

Way to hand-pick the benchmark that shows that single-sided sticks are not as good as dual-sided sticks in all apps to prove an unrelated point.

 

[X71200]

That pic isn't really going to sell anyone on speed before cas, in fact the review it comes from would seem to show the opposite. Especially if you want to do well benchmarking.

How come slower memory comes up with higher R/W results?

And yes, speed does become in front of CAS latency. If it didn't then DDR 400 CL2 would have performed better than DDR2 800 CL4 and DDR2 800 CL4 would have performed worse than DDR3 1600 CL8.

Way to hand-pick the benchmark that shows that single-sided sticks are not as good as dual-sided sticks in all apps to prove an unrelated point.

Care to elaborate?

 

[cadaveca]

How come slower memory comes up with higher R/W results? Care to elaborate?

I think you already did. Once you've played with a lot of DDR3, you learn to recognize these things for what they are.

 

[LAN_deRf_HA]

Way to hand-pick the benchmark that shows that single-sided sticks are not as good as dual-sided sticks in all apps to prove an unrelated point.

You're reading too much into it. He made a non-point and I threw it back at him, but since you bring it up that's not really a defense of the crappiness of high speed kits either. Unless you're saying they now have dual sided 2933 and higher kits? I haven't seen any yet.

And yes, speed does become in front of CAS latency. If it didn't then DDR 400 CL2 would have performed better than DDR2 800 CL4 and DDR2 800 CL4 would have performed worse than DDR3 1600 CL8.

We're speaking in more practical terms than your example. I'd take a 2400 c9 kit over any of the 2933 kits I've seen, and I'd save a ton of money and have the same or better performance for it. Those bandwidth or bust situations are hard to find.

 

[cadaveca]

You're reading too much into it. He made a non-point and I threw it back at him, but since you bring it up that's not really a defense of the crappiness of high speed kits either. Unless you're saying they now have dual sided 2933 and higher kits? I haven't seen any yet.

You don't pay attention to ram like I do. Not that you'd have any reason too... But I'm a ram geek.

so here you go:

http://www.newegg.com/Product/Produ...p=&AID=10446076&PID=6146846&SID=182eb60nw67z2

 

[X71200]

I think you already did. Once you've played with a lot of DDR3, you learn to recognize these things for what they are.

That's a non-answer.

You're reading too much into it. He made a non-point and I threw it back at him, but since you bring it up that's not really a defense of the crappiness of high speed kits either. Unless you're saying they now have dual sided 2933 and higher kits? I haven't seen any yet.

We're speaking in more practical terms than your example. I'd take a 2400 c9 kit over any of the 2933 kits I've seen, and I'd save a ton of money and have the same or better performance for it. Those bandwidth or bust situations are hard to find.

Non-point as in what? The OP is obviously wondering whether DDR4 will perform better or not in the end. There's no "crappiness" of high speed kits.

A 2933 kit WILL perform better than a 2400 kit regardless of it's higher timings. Period.

 

[JoshuaAJones]

What you are referring to there is known as RAM 'access time', or more casually as RAM 'speed' or 'performance', which is basically the time it takes in nanoseconds for the RAM to locate a single piece of information and make it available to the processor (a very rough definition). Latency is a reference to CAS timings, which refers to the delay between the memory controller telling the RAM to 'find' the information, and the information being made available.

The general way to work out RAM access time is: ( CL / Frequency ) * 1000.

So, for 2133MHz CL15 DDR4 memory, ( 15 / 2133 ) * 1000 = 7.03ns

And for 1600MHz CL9 DDR3 memory, ( 9 / 1600 ) * 1000 = 5.63ns

So actually, DDR3 memory is faster in terms of access time between these two examples (as AFAIK 2133MHz is meant to be the standard DDR4 RAM frequency). As many people have already said, DDR4 RAM probably will eventually have better access time than DDR3 in due time.

Layton

Your math is accurate for the old SDRAM but for all DDR SDRAM you must multiply by 2000.
Therefore, in your examples:
2133MHz CL15 DDR4 memory, ( 15 / 2133 ) * 2000 = 14.06ns
1600MHz CL9 DDR3 memory, ( 9 / 1600 ) * 2000 = 11.25ns

That's a non-answer.



Non-point as in what? The OP is obviously wondering whether DDR4 will perform better or not in the end. There's no "crappiness" of high speed kits.

A 2933 kit WILL perform better than a 2400 kit regardless of it's higher timings. Period.

Actually, 2400-CL9 is faster than 2933-CL12 (the only CL I can find)...
DDR3 2400-CL9 = 7.50ns (~$270US for a 16G Kit)
DDR3 2933-CL12 = 8.18ns (~$700US for a 16G Kit)
DDR4 3000-CL15 = 10.00ns (~$390US for a 16G Kit)

As for now, the only reason to go to DDR4 is for the low power draw (1.2V vs 1.65V).
Give it some time though and we'll see DDR4 prices drop and speeds soar.
Eventually, they'll get up to DDR4 4200-CL15 (7.14ns) using only 1.35V.