Found this deal. -21% off. CAS 19! Still seems much cheaper than what is sold here. I am unsure about newegg warranty terms but will probably go with them as the price is right!
G.SKILL Ripjaws V Series 8GB 288-Pin DDR4 SDRAM DDR4 3600 (PC4 28800) Desktop Memory Model F4-3600C19S-8GVRB
I'm going with it's soo cheap because of how high the cas latency is
Won't notice the difference at all while 'Drunk' ;P
The effect that CL has is linked directly to MT rating; effective latency of high CL RAM will be quite reasonable if transfer rate is also high. It's more nuanced than "high CL is bad".
It's fine to compare CL at a given MT rating, but don't automatically dismiss higher CL if the RAM is rated for faster speeds.
DDR4 3600 at CL 19 is worse than DDR4 3000 at CL 15 for most people.
In terms of true/effective latency, yes - by a maximum of less than 6%.
In terms of bandwidth, no.
Horses for courses.
@Shobai: What applications would absolutely need the extra 600MHz or even 400MHz (over 3200 CL 15 which is common)?
edit: should clarify, always happy to learn more about how pc's work. legit question.
@neferseki: 300Mhz, or 200MHz. DDR4 is rated in MT/s.
Even so, you've left out the most important part of your question - you should be asking "of the applications that I use now (and intend to use in the future) which respond better to increased bandwidth and which respond better to improved effective latency".
For games, ARMA 3 is the most obvious example (even though it's getting long in the tooth). Depending on your source, there's ~10% more FPS via DDR4-3600 over DDR4-3000, all else being equal, and improved frame times.
@Shobai: Yes yes we all know DDR is double data rate, No need to be pedantic over semantics. The industry calls it MHz just leave it at that.
For your example of ARMA 3 though, theres no way that isn't affected by CAS Latency though right? What difference does CL19 over CL15 make in that scenario? Is frametimes affected only by bandwidth or is that affected by CAS too?
I don't like looking at things in a complete vacuum, still need to consider all factors.
@neferseki: No, the industry absolutely does not - all RAM manufacturers and resellers label it appropriately as MT/s. People speaking otherwise are talking from ignorance.
As for ARMA 3, that's why I say "depending on source". Some do a proper control, where effective latency is kept constant while bandwidth alone is changed, but most tech publications are run by people working from ignorance and just lock the timings then wind the frequency up (which ties the two together, as you rightly point out).
This is CL13 @2400mhz, could probably even run it at CL12@2400 if you put 1.5v (maybe less) at it.
CAS is how long (time) it takes for the information to travel from the stored location to the output pin in cycles per second. This time is a fixed number, in this case 10.5ns. If for example you have a first gen Ryzen and don't want to exceed the rated speed of the memory controller on this chip which is 2666mhz, you'd run CL14.
I bought this cause ram deals seemed to be much less frequent than Ryzen deals but now I'm getting worried.
RAM prices should overall decline a little more over the year but not sure where rock bottom will be.
Ultimately, it's likely not to be too much lower so don't stress about it.
fingers crossed. going on a little holiday in a fortnight. hopefully A) Ryzen 3 get firmed up release date and B) no insane sales on Ryzen 2nd gen and C) no massive ram price drops to make me feel like a sucker
Roughly mid-year is estimated launch for Ryzen 3000. Few months to go!
https://notkyon.moe/ram-latency.htm
base on calculation 10.5 nanoseconds
roughly the same as 3000mhz cl16 10.666 nanoseconds, ofc 3000mhz c16 unlikely to oc to this frequency without some voltage and twinking
so not too bad, would ryzen take this xmp, I wonder
The site linked, and the OP description as well, make me sad: it's incorrect to use "mhz" (sic) where you're all using it.
DDR is "double data rate" (although, amusingly enough, the word is that DDR5 will be quad data rate…). This means that data is clocked on the rising and falling edges - don't let that concern you, just remember that all ratings that you read are expressed in terms of "MT" (mega transfers) and for DDR that's at twice the clock speed. DDR4-3600, then, has a clock rate of 1800MHz ("shouldn't that be 1.8GHz?").
Why does that matter? Because latency is a time-based measurement and needs the be determined from a time-based measurement; MT rating is one step removed from being a time based rating, so any calculation of effective latency needs to translate a given MT rating into a true MHz value before it'll be accurate.
https://www.crucial.com/usa/en/memory-performance-speed-late…
as per crucial true latency calculation, say: ddr4 2400 c17
17 /2400 * 1000 *2=14.1666ns agrees to that web page calculator
Yes, that's correct.
You can clearly see what I'm talking about there, can't you? The equation is in the form:
True latency = (CL rating x some arbitrary number x a second arbitrary number) / MT rating
I can help here, though, because the arbitrary numbers aren't actually arbitrary. Rather:
Some arbitrary number: this is a scaling factor to convert the result from micro seconds to nanoseconds
A second arbitrary number: this scaling factor converts the MT rating to MHz. If/when DDR becomes QDR (with DDR5, perhaps?) the value to use will be '4'.
So, as per my contention, don't call (e.g.) DDR-3600 "3600mhz" ("milli-what?") or even "3600MHz". It's 3600 MT/s, double data rate, and the RAM is running at 1800 MHz.
If you are trying to run at lower speeds and tighten the timing,and we assume most people would buy 2 sticks, this is a better deal per stick at the same latency:
https://m.newegg.com/global/au/product/N82E16820231933
But I would spend a a couple dollars more and get this one:
second one! yes.
If your intended task is latency-bound, go for that one for sure. The OP kit would have to be able to tighten to CL18 to match that second kit.
Having said that, if the task is bandwidth-bound, would that second kit be able to wind up to 3600 MT/s? How much would you have to loosen CL to get there?
They're probably both running the same batch of chips under the heatspreaders. You're probably just as likely to be able to take one kit to match the other as vice versa.
good to learn something new, thank you! though I must admit I have no desktop system and no intention to have one. but I enjoy watching how processing power is becoming more and more affordable. I could use more processing power for what I do or want to do, but ability to do it from any cafe or lounge is more important to me.
It should depend on the intended workload - for instance some games prefer more bandwidth and aren't as sensitive to latency, so you're better off loosening timings and upping MT/s. A counter example is that some other games are less sensitive to bandwidth but prefer reduced 'effective latency'.
Think about what you want to do with it now, and what you expect to want to do with it into the future, then select something you think will get you there (and roll the dice on overclocking and timing tightening, if desired).
I would say that 3000 cl15 will run at 3600 cl19 without any issues.
I'd take the chance that such a set would, too, but you've linked 3000 CL15 - reckon they would? Probably, who knows?
These sticks are most likely Hynix CJR.
You can see at the bottom of that thread that people have overclocked them to 3200 C14 and 3600 C15.
At this price it's a steal compared to B-die, provided you want to get your hands dirty with RAM overclocking.
I am on phone at work and that site isn't playing nicely, are you referring to the single 8gb stick only? Also what voltage did they use to get those timings? I am pretty tempted by the timings you just gave me
The overclocks in that thread used 2 sticks in a motherboard with 2 RAM slots, so that'd be the best case scenario.
They used 1.35v.
Picture
I've just bought 2 of these sticks and will see how far they go on my 2600 and B450 Tomahawk.
Thank you. I have an x370 itx board so that should work well.
I can see they are not 14-14-14 though. Don't know how much difference it will make in the end. The cheapest b-die kit on newegg is around 240, I havent slept in 3 days to work out what ram to buy
@giovannip: I mean, you could get away with 3000 16-18-18 without noticing the difference.
https://imgur.com/a/pgiLe
Hi, I'm going to be doing this myself, if you could post your results, would help. Couldn't find other threads on the Tomahawk + 3600 timings.
@Cave Fire: I wouldn't expect 3600MHz to work stably, though 3466MHz should be doable.
RAM just came in today, so I'll update you with some overclocking results.
@Cave Fire: Typically flat timings are Samsung B-die. Hynix tends to have the 2nd and 3rd timings 2-3 higher than the first.
I can't seem to stabilise 3533/3600 16-20-20-40 even with 1.45v, probably limited by my mobo and/or IMC.
I've got 3466 16-19-19-38 stable with 1.40v. Currently testing 3466 15-19-19-38.
I need help, my old mobo that has DDR3 x4 slots. I have been using mixed RAM 2 + 1 for some 5 months now. Then last week, 2/4 slots suddenly not recognized any RAM stick. One manage to boost up once, but not any more. Any ideas ?
Reseat the ram. That just means unplugging all of them, removing the dust that might have fallen in and then reinstall the ram.
If still not working,
Update the BIOS, or reset the BIOS settings using the reset jumper.
Remove the sticks of ram and leave only one or two in, test and retest after moving sticks of ram to different slots till you have tested all Dimm slots
If some of the slots do not detect ram then you might have a partial motherboard failure
Thanks for the reply, I did all the above except reset the BIOS. The strange thing is that, a G.Skill stick was workinng on 1/2 non working slot, but I haven't manage to get it works again …
Someone posted in one of the other Newegg deals that the ram manufacturer's post a high speed while the latency is high. So it's not an "apples to apples" comparison. A cas 15 3000 stick may be able to O/c to 3600, and is objectively better?
People get very hung up on latency (see above) when it's not very important. Higher clock rate is more important.
I feel like ozb should stop posting all these loose timing high frequency 'deals'. They may cause people to make a bad purchase if they end up with loose timed ram that won't run at rated speeds on ryzen.
i just bought two sticks 40min ago, and I intend to use it with ryzen… If it ends up being slower than 3200 will have to wear that I guess.
Awesome find.