fry178:

Funny how ppl always assume you have to be brain dead to spend that kind of money on a 16gb kit. How about someone that makes +100k a year, no kids or other things to worry about, and is fine with wasting their earned cash on stuff like this instead of drugs and the like?

If you are very young and with 100k+ a year maybe. I was not making much money till i grown up in my late 30, and when i did, i stopped being compatible with those kinds of money trashing. When i was 20+ i would have done it. So those are for a very restricted group of people. Young, unexperienced, excited and rich people.

Oh yeah, not saying its a good value or that it makes sense, but so are many other things, and not everyone has the same priorities, or we would all drive the same small Chinese cookie box as a car ;-)

Would rather get a 2080Ti than ram this fast. Same price...which is the better upgrade? Nah going for the ram instead....

Support 2500 FCLK or hit wall 1900 ? or GTFO !

Robbo9999:

Ok, so this is 5000Mhz memory and initially the timings look ok with it being CL18, but then those other timings are 26-26-46. I feel like I'm seeing an increasing trend with RAM kits now where the first latency number is a lot lower than the second & third numbers - like this kit in this article which is 18-26-26-46. Now I'm assuming they've loosened the other timings so they can get CL18, but what kind of negative impact does it have for those second & third timings to be so loose in comparison to the first? When I bought my DRAM kit the latencies were more like 16-16-16-46, so it seems like a new trend that they have such big differences between the timings - it feels like marketing more than anything else, just so they can call it CL18 even if the other timings suck. I wonder how much compromise in performance is going on here?

I think its just the nature of the newer memory ics, b-die is kind of an outlier, it can do very tight secondary timings with little issue, but the hynix, micron(this kit uses micron) and newer samsung memories, do not share this quality, this is why you can now find alot of new cl16 3600 kits with 16-19-19-39. I dont believe they are loose enough that it negates the advantage of the high memory speed. we can estimate the latency of each timing roughly, 5000mhz cl18-26-26 kit 1000/2500mhz(real frequency is 1/2 of the rated speed) = 0.4ns per cycle x 26 trcd cycles =10.4ns , 3200 cl14-14-14 1000/1600mhz =0.625ns per cycle x 14 trcd cycles = 8.75 3600 cl 16-19-19 1000/1800 =0.555ns per cycle x 19 trcd cycles = 10.55 ns The latency of each secondary timing isn't that different from a normal kit, though you must also keep in mind that the cas latency is far more important, so even though it has avg secondary timing latencies, the cas latency is only 7.2ns , which is better than 3200 cl12 , and you also have the benefit of much higher bandwidth. while the looser latencies will have an effect on performance, the kit is cycling so fast, that it probably doesn't matter that much, only extremely good b-die would probably out perform it latency wise, but not in bandwidth (in theory) whether ryzen 3000(or skylake derivatives) can actually make use of the extra bandwidth is another story, it is unlikely to make much of a difference. it is a waste of money.

The problem with Ryzen 3000 (Zen 2) is that once the IF (Infinity Fabric) clock is no longer in sync with DDR clock, the memory controller enters a "buffered" mode in which data is temporarily stored inside it, before being passed onto the CCX. That temporary storage takes about 8 ns, so it basically makes a 3600 CL16 memory into a CL 28 memory. Ouch. Obviously once the bytes start flowing in, the latency is no longer relevant... but for latency sensitive applications (like games), going over maximum possible IF clock is only going to reduce performance. I have performed hundreds of tests on my 3700X with my Ballistix Elite, trying all the possible clocks and latencies (which were not crashing), from 3200 CL12 !! to 4400 CL18 (but at that point the voltage required for even finishing benchmarks was close to 1.5v and I wasn't comfortable about it anymore) The result? DDR4-3733 CL16-19-16-42 T1 (Gear Down Off) (which works on a relatively low voltage of 1.42 for such an overclocker's kit) was the highest performing in the majority of tests. Not even running at 4400 with the tightest timings possible on that clock was not enough to even get close to 3733 "Linked" ( fclock = 1866 ) What AMD said that 3733 is the ideal frequency was not just marketing talk. They probably performed not thousands but tens of thousands of tests with all kinds of memory and were really telling us the best they achieved. (Note: Some chips do 1900 fclock stable (3800), but mine doesn't unfortunately. Still, I'm satisfied with 1866, I have an awesome CPU that easily rivals the much more expensive 9900K while it only costed me around 340 Euro, and reused the motherboard which I already had from Ryzen 1700. Best deal ever !

So damn stupid....... Maybe if it were a 128gb set across two sticks, but this is just plain stuuuuuuuupid.

fry178:

Funny how ppl always assume you have to be brain dead to spend that kind of money on a 16gb kit. How about someone that makes +100k a year, no kids or other things to worry about, and is fine with wasting their earned cash on stuff like this instead of drugs and the like?

And sometimes it's also about time investment. If you've got the cash to burn, just buy the best and most expensive stuff available and be done with it.

What's even more funny is that these are using the exact D9WFL "E-die" chips from Micron, which exist in many of the higher-end memory out there. https://www.amazon.com/Crucial-Ballistix-Desktop-Gaming-BLE2K8G4D36BEEAK/dp/B07NPKTRS1/ This is literally the SAME memory. For $149. Volt it to 1.5-1.52 instead of 1.35 XMP default , put a fan to blow air over them and there's a good chance it will work with the same timings as these Corsair modules, at 2500Mhz 18-26-26-46 (or very close). That's $1076 extra for a number printed on a tiny piece of paper glued to the heat-spreader. Great scam Corsair, great scam !

Makes you think if nvidia sold dram...

Yakk:

Makes you think if nvidia sold dram...

They do. With every single one of their videocards...

@wavetrex except the one sold by corsair as 5000, is "certified" to run that speed, any trouble its with corsair, not "you". same reason not everyone is tuning/tweaking their own car (or other things).

That is some fast RAM. Well done Corsair!

Also to give perspective as to how much this ram is limited by current cpu offerings, de8auer was able to achieve ~60gbs r/w using a similar 5000mt/s kit The theoretical bandwidth of this kit is 80gb/s! There is virtually no memory scaling beyond 3800mhz on ryzen With similar kits at 4400mt/s and higher on intel , its the same story, exceeding 60gb/s is difficult no matter how fast the kit is.

that moment when your RAM is faster then your CPU... o_O

I'll offer 100$ for them max.

Well when you look at the pricing for this DDR3 ram on Amazon, the DDR4 5000mhz seems a steal! See how everything is relevant in life? https://www.amazon.com/Corsair-Vengeance-2400MHz-Memory-CMY32GX3M4A2400C11R/dp/B00GM83C86/ref=sr_1_18?keywords=Corsair+Vengeance+Pro+ddr3+2400mhz&qid=1570973388&s=electronics&sr=1-18 https://media0.giphy.com/media/ilqOZXJ8pxfQnnLEjK/giphy.gif

Hence their name ("corsair"), a synonym for "pirate"... 🙄

Just in time to milk the last few months of DDR4 before the arrival of DDR5, which will have speeds of 4800-6400mhz as the norm.