1600 CAS 9 -- vs -- 1866 CAS 10

[WhipperSnapper]

So, which is faster:

1600 CAS 9 -- or -- 1866 CAS 10

Assume that they are running at their rated frequencies.

 

[Brunnis]

1866 CAS 10 is faster. The higher frequency produces higher bandwidth, while it also provides lower latency than 1600 CAS 9:

(1600/1866) * 10 = 8.6

In other words, running the RAM at 1866 CAS 10 gives equivalent CAS latency to running 1600 CAS 8.6 (if that was possible).

 

[WhipperSnapper]

I found a good article to explain it:

http://www.anandtech.com/show/6372/...333-to-ddr32400-on-ivy-bridge-igp-with-gskill

I guess a basic rule of thumb is to take the speed divided by the latency and compare, and whichever number is higher is better:

1866 / 10 = 186.6

1600 / 9 = 177.8

The article I linked to says the formula for determining the latency in terms of nanoseconds is:

2000 * (latency / speed)

OK, good to know!

 

[Burner27]

But is the 'real world' performance increase perceivable to justify the price difference between the two?

 

[myocardia]

But is the 'real world' performance increase perceivable to justify the price difference between the two?

Figure out the speed difference yourself, as shown in the second or third post, and decide whether 4/10 of a percent (in the above example) is worth the difference in price, or not. BTW, this applies 100% of the time with system RAM: the higher the speed, at the lower the latency, the better, at least as far as performance.

 

[Cerb]

But is the 'real world' performance increase perceivable to justify the price difference between the two?

Depends. With IGP, maybe (if even light gaming with an AMD APU, or Skylake's IGP, definitely). With simulation work, maybe (including games like Flight Simulator X and Dwarf Fortress, that I know of). With significant CPU overclocks, or running multiple GPUs in a dual-channel socket, faster RAM can definitely lead to better real world performance.

But, today, they are often the same price, or almost the same price. 1866MHz might be cheaper, on some given day, and 2133-2400MHz DDR3 kits often come up for mere ~5% more, so why not? Worst case, you can run them like I do, at 1600MHz and tight timings (6-6-6 :twisted.

 

[WhipperSnapper]

Worst case, you can run them like I do, at 1600MHz and tight timings (6-6-6 :twisted.

Seriously? You have RAM running at 1600 CL6?

That's 1600 / 6 = 266.7 which is the fastest specs I've seen available for purchase on DDR3.

 

[Cerb]

Seriously? You have RAM running at 1600 CL6?

That's 1600 / 6 = 266.7 which is the fastest specs I've seen available for purchase on DDR3.

It's 2133MHz, CL11 or CL10, but I have a B85 motherboard. I can't remember the exact cost, now, but it was definitely less than $10 more for the whole 32GB of it, compared to 1600MHz CL9, at the time.

 

[tweakboy]

NO real world difference. very nice...

 

[Makaveli]

Seriously? You have RAM running at 1600 CL6?

That's 1600 / 6 = 266.7 which is the fastest specs I've seen available for purchase on DDR3.

that is pretty good

1600 / 7 = 228.5 here.

 

[Magic Carpet]

you can run them like I do, at 1600MHz and tight timings (6-6-6 :twisted.

Voltage?

 

[aaksheytalwar]

They will be very close in performance with 1866-10 being very slightly faster.

2133-10 will be faster though. And 2133-11 will be comparable to 1600/9 and 1866/10 but again very very slightly faster than both.

 

[Cerb]

Voltage?

1.5V. It's these (I can't run higher than 1600MHz). One of the slower profiles on the ones I got is 6-6-6, which they run well at, but it won't pass one memtest run with 5-6-6, much less any tighter.

 

[BFG10K]

But is the 'real world' performance increase perceivable to justify the price difference between the two?

In the real-world you won't notice any difference between 1333 and 2133, much less this. Heck, many can't even tell if they're running in single channel mode.

The exception might be the IGP, but anyone running an IGP doesn't care about about performance to begin with.

 

[Dasa2]

In the real-world you won't notice any difference between 1333 and 2133, much less this. Heck, many can't even tell if they're running in single channel mode.

The exception might be the IGP, but anyone running an IGP doesn't care about about performance to begin with.

a myth spread by bad reviews like the one anandtech did

sure a gpu bottlnecked game will show no difference as faster ram only helps the cpu
even then it only seems to help when the software is overflowing the cpu cache


most people didnt believe a few user run tests done by myself and others in the past but now there is lots of examples of cpu limited games that benefit from faster ram
http://www.overclock.net/t/1487162/...affect-fps-during-high-cpu-overhead-scenarios

http://www.techspot.com/review/1089-fallout-4-benchmarks/page6.html

When benchmarking the Intel Haswell processors, the memory speed in the BIOS kept reverting back to 1333MHz (was an issue with the BIOS which has now been fixed) at every processor change rather than holding the DDR3-2400 XMP (Extreme Memory Profile). At first we accidently tested the Core i7-4770K with the memory clocked at 1333MHz and were perplexed by the much weaker than expected performance.

It was quickly spotted that the memory was running much slower than it ought to be, though we weren't entirely sure this was the cause of the massive deficit. Turns out it was, as moving the memory back up to 2400MHz resulted in a massive 67% increase in minimum frame rate for the Core i7-4770K.
A similar situation was found with the Core i3-4360, which went from a minimum of 41fps with DDR3-1333 memory to a minimum of 65fps with DDR3-2400.
Yes, we know there's a significant difference between the now obsolete 1333MHz memory speed and 2400MHz, but from memory I don't recall it having anything like that kind of impact on gaming performance. Interestingly the same variation in memory speeds had little impact on the AMD FX-8350's performance, as the processor was just 13% faster with the quicker memory.
Anyway, we found this interesting and it would suggest that Fallout 4 fans with Intel hardware will want to ensure they are running their memory as fast possible.

67% is massive 10-30% is more what i would expect so i wouldnt be surprised if there was something else wrong with here test system
it will be interesting to see more testing done on fallout4 which is clearly horribly optimised

http://translate.googleusercontent....4.html&usg=ALkJrhitpE4CwC0uG8dYOMAkcv8GkAafqw

http://www.hardocp.com/article/2015...76700k_ipc_overclocking_review/6#.VkkDoeJrsUl

http://www.hardware.fr/articles/940-5/cpu-ddr4-vs-ddr3-pratique.html

https://forums.bistudio.com/topic/1...mance-comparison-1600-2133-up-to-15-fps-gain/

http://forums.atomicmpc.com.au/index.php/topic/55771-cpuram-performance-in-thief/

http://forums.anandtech.com/showthread.php?t=2455994

http://www.eurogamer.net/articles/d...it-finally-time-to-upgrade-your-core-i5-2500k

 

[Magic Carpet]

1.5V. It's these (I can't run higher than 1600MHz). One of the slower profiles on the ones I got is 6-6-6, which they run well at, but it won't pass one memtest run with 5-6-6, much less any tighter.

1600 Mhz CL6 @ 1.5v is quite rare. Too bad you are handicapped by the B85 chipset, it's a shame to run such a great kit at this lower speed.

 

[Leyawiin]

AT has a handy little grid in an article they did on the subject.
http://www.anandtech.com/show/7364/memory-scaling-on-haswell/10

 

[Cerb]

1600 Mhz CL6 @ 1.5v is quite rare. Too bad you are handicapped by the B85 chipset, it's a shame to run such a great kit at this lower speed.

I just selected one of the other profiles, after the auto setting wanted to run it at the same timings as 2133MHz specs, rebooted, ran Windows' RAM test, and then went on about my game playing (until the urge to try even lower timings came). I expected to run it at lower than CAS 9, which is why I bought 32GB (then re-purposed the 2x8GB I started with), at the time; but yeah, I was expecting more like 8-8-8, or 7-8-8, from 10-12-12@1.6V to start.

I'm also "handicapped" by a non-overclockable CPU, so I wouldn't be seeing much of those gains Dasa2 is showing, anyway. I manage .

 

[dazelord]

http://www.techspot.com/review/1089-fallout-4-benchmarks/page6.html
67% is massive 10-30% is more what i would expect so i wouldnt be surprised if there was something else wrong with here test system
it will be interesting to see more testing done on fallout4 which is clearly horribly optimised

Something is very wrong in Fallout 4.
http://www.techspot.com/review/1089-fallout-4-benchmarks/page5.html

"However, reducing the clock speed as low as 2.5GHz had a catastrophic impact on performance, with the 6700K becoming almost 40% slower when going from 3GHz to 2.5GHz."

 

[DrMrLordX]

Seriously? You have RAM running at 1600 CL6?

That's 1600 / 6 = 266.7 which is the fastest specs I've seen available for purchase on DDR3.

My old Pi Blacks would do DDR3-1600 CL6 on K10.5 memory controllers. Actually they did 6-5-7 which was weird, but fun. They would do it at around 1.6v, possibly lower (they were specced for 1.65v so I rarely bothered with undervolts).

My newer DDR3-2400 only does CL7 on Kaveri. But there's 16 Gb of it vs the 4 Gb of my Pi Blacks so . . . there is that.

 

[WhipperSnapper]

Waaah!

So I installed my Corsair Vengeance Pro RAM (stock 2400 / CL 11 at 1.65 V) and the computer wouldn't boot at 1866 CL9 1.5 V and I couldn't even use my keyboard or mouse to work the BIOS menu. I had to use the MemOK! button to reset it. I ended up having to set it to 1600 CL9 1.5V (seems to run fine at this setting).

Kind of disappointing. I was hoping I might be able to get say 1600 CL8 at 1.5 or 1866 CL9 at 1.5. Maybe I'll try again but with 1.55. (I don't want to run at 1.65V to prevent any potential damage to my CPU's memory controller.)

Did I get a batch of crappy RAM, or is it possible that those particular sticks are just Voltage hungry since their default at 2400 is supposed to be 1.65?

 

[DrMrLordX]

They are probably voltage hungry. I need something like 1.6v to get my DDR3-2400 CL11 to do DDR3-1600 CL7.

 

[Dasa2]

your extremely unlikely to do any harm with 1.65v to ram provided temps are ok

 

[DrMrLordX]

It's not the RAM that worries him. It's the IMC. Intel IMCs do not like heavy VDIMM.

 

[lehtv]

I've never seen or heard of any IMC of LGA1150 CPU damaged due to 1.65V RAM. It's safe to say the risk is very, very low.