G.Skill Trident Z 3200 MHz C14 32 GB

[cadaveca]

G.Skill has long been a bastion of affordable high-performance memory, and their Trident Z line-up is the ultimate version of what DDR4 they have on offer. Featuring a large heatsink ready for some abuse, G.Skill's Trident Z sticks now can also be had in other colors!

Show full review

 

[ZeppMan217]

Curious - why isn't memory overclocked as part of the review? CPUs/GPUs always have an OC section but not memory, even though being OC friendly is mentioned.

 

[cadaveca]

Curious - why isn't memory overclocked as part of the review? CPUs/GPUs always have an OC section but not memory, even though being OC friendly is mentioned.

It is because OC is not guaranteed, and memory makers often change which IC is in a SKU, which affects OC results too. Also, the X99 platform used for testing in the review does affect OC potential in a big way. The CPU was the limit, not the memory (since it does OC farther on Z170/Z270). If I posted results on the platform used for testing, it'd be showing the CPU limits, not the memory limits.

 

[buggalugs]

I always found Gskill to have much slower timings and sub timings than other brands like Corsair....or sometimes main timings are similar but subtimings are much slower like double. DDR4 has disappointed me in general with worse latency than DDR3.

Heres a question, are there any DDR4 memory with 1T command rate? I dont think there is but it would be good to know.

 

[erixx]

"Sub-timings are a difficult area to navigate. If you are someone who overclocks memory in great detail, what is going on is pretty basic, but if you do not, you might not be aware of such things and the benchmark results might, as such, say something different from what is really going on to you. Sub-timings are one of those things that differentiate brands that use the exact same memory chips and the same PCBs, but end up with different results. They can also have a significant impact on performance, and in far greater a way than primary timings do. They also are sometimes dictated by the memory stick's SPD table, and sometimes by the board itself, leading to a complicated set of criteria I am sure many do not consider, and places like stores and memory manufacturers don't seem too eager to disclose such things readily. This is a big part of why it isn't smart to mix kits from different sets or brands even if they seem to be the same; sub-timings can change from kit revision to kit revision, never mind that the actual memory chips can change as well. Deciphering what is going on is NOT for the inexperienced."

This ending paragragh holds so much thruth!

The superb Aida memory results you show, I can only get same or close with 4400 Mhz Uncore speed! (5930K here) Soon I will have a 6850K to play with, hoping for a more stable 3200 RAM speed. I may ask you for help again

 

[swirl09]

This is the kit I use in my current main, very happy with it

Its often a part overlooked, for good reason, as the results can vary and its not a high priority for a gamer. But over the years I have found that when I have good performance in a new title and someone with an identical or near identical setup as far as cpu/gpu/clocks are concerned, the difference tends to come down to storage and RAM.

So while it still shouldnt be a priority when building a new machine for gaming, my advice would be if you have the budget, fit it in. There are some benchmarks around too showing the differences, sometimes its next to nothing (and even losing a frame here or there with pricier RAM), while at times there are some good gains which can make a real difference.

 

[erixx]

Please Dave, would you tell me where (or under what name) the RAM SPD "tRC" value can be found in the UEFI BIOS? I swear that after several Asus boards I can never find it!

 

[FR@NK]

Heres a question, are there any DDR4 memory with 1T command rate? I dont think there is but it would be good to know.

I have the 4x16GB version of this kit and it will run at 1T command rate. But like Dave said, it depends on your CPU.

I can only get same or close with 4400 Mhz Uncore speed!

Broadwell-e's uncore speeds dont overclock as much compared to haswell-e. My chip wont even do 4000 MHz and alot less when then the cores are overclocked.

 

[bug]

I always found Gskill to have much slower timings and sub timings than other brands like Corsair....or sometimes main timings are similar but subtimings are much slower like double.

Timings are always a trade off. It's easy to set defaults a bit higher, but that will probably get you in trouble with more motherboards.

DDR4 has disappointed me in general with worse latency than DDR3.

Do not compare numbers directly. In real life, latency is measured in nanoseconds. 1 clock@800Mhz and 2 clocks@1600MHz are actually the same exact latency.

 

[Pjokerxp_]

here is mine, I can run it at 1T but does it realy change things in games?

does these benchmarks looks right ? because in the review it seemed alot faster

 

[webdigo]

is a heatsink on a DDR4 really necessary when they run at like 1.35v?

 

[cadaveca]

is a heatsink on a DDR4 really necessary when they run at like 1.35v?

In some circumstances, yes, as sticks can get fairly warm when pushed hard in temperate environments. It is also not always about cooling; memory IC's needed refresh rate can drift with temperature, so by spreading heat across a DIMM, you can help ensure that all IC have roughly the same refresh needs.

Please Dave, would you tell me where (or under what name) the RAM SPD "tRC" value can be found in the UEFI BIOS? I swear that after several Asus boards I can never find it!

Nope, can't tell ya. Not a setting any more.

Heres a question, are there any DDR4 memory with 1T command rate? I dont think there is but it would be good to know.

No, because it doesn't matter. You can set 1T just fine though, although nearly no real difference is noticed.

 

[Hossein Almet]

I was able to tighten the timing of my Trident Z 3200, 16, 18, 18, 38, 561 to 16, 16, 16, 36, 465 from the 6800k, X99 gaming pro carbon, without any issue. When the tRFC reduced to 451, the motherboard has no problem booting it, but W10 refused to play ball, ie when I restarted it, BSOD appeared, after successful getting into the BIOS I loosened the timing a bit, and it's all fine.

 

[Eknex]

Are in the multiprogram tasks real differences between one of 2666mhz and 3200mhz?

 

[bug]

Are in the multiprogram tasks real differences between one of 2666mhz and 3200mhz?

Not unless you do something that's particularly memory intensive. For everyday usage, you can't tell the difference.

 

[basco]

Please Dave, would you tell me where (or under what name) the RAM SPD "tRC" value can be found in the UEFI BIOS? I swear that after several Asus boards I can never find it!

Row Cycle Time (tRC).
Determines the minimum number of clock cycles a memory row takes to
complete a full cycle, from row activation up to the precharging of
the active row. For optimal performance, use the lowest value you can,
according to the tRC = tRAS + tRP formula. For example:
if your memory module's tRAS is 7 clock cycles and its tRP is 4 clock cycles,
then the row cycle time or tRC should be 11 clock cycles.

like dave said not a setting anymore.
there are more hidden ram timings that different vendors hide or not(ex. rampage V has more timings to play with then my sabertooth)

and there is no tool i know with which you can read it like memtweakit or aida64

 

[erixx]

Row Cycle Time (tRC).
Determines the minimum number of clock cycles a memory row takes to
complete a full cycle, from row activation up to the precharging of
the active row. For optimal performance, use the lowest value you can,
according to the tRC = tRAS + tRP formula. For example:
if your memory module's tRAS is 7 clock cycles and its tRP is 4 clock cycles,
then the row cycle time or tRC should be 11 clock cycles.

Thanks a lot basco. The little problem is: AIDA, for example, reports a supposed tRC value (from the stick, we should guess), one for Intel spec timings, and one for XMP timings, but nowhere in the BIOS you can enter it. Just like Dave said.

Looking forward to learn much more. I can get my CPU at 4400 and RAM at 3200 and put benchmarks on fire! But of rock solid stability I stay at 4200 and 2800.

 

[bug]

Thanks a lot basco. The little problem is: AIDA, for example, reports a supposed tRC value (from the stick, we should guess), one for Intel spec timings, and one for XMP timings, but nowhere in the BIOS you can enter it. Just like Dave said.

Looking forward to learn much more. I can get my CPU at 4400 and RAM at 3200 and put benchmarks on fire! But of rock solid stability I stay at 4200 and 2800.

If you want to learn more, check this out: http://archive.techarp.com/freebog.html
It may miss some newer stuff, but there's enough in there to keep you busy for a while (look for the "memory subsystem" section, I didn't find a direct link).

 

[buggalugs]

I



No, because it doesn't matter. You can set 1T just fine though, although nearly no real difference is noticed.

I have to disagree, run a aida memory benchmark at 2T and then at 1T, Latency numbers are better and even throughput can be better. Like just say you get 30,000 MB/S write speed at 2T, you might get 30,700 at 1T. Same with read and copy....but if your memory cant handle it, it might perform worse at 1T, or the system will crash with memory errors. I have found you can run 1T at default settings but with overclocking or using XMP, you cant. Theres usually a tradeoff between frequency and latency.

These are real measurable figures, is the average person going to notice a difference? Probably not, but most people dont notice a difference if they're running stock CPU speeds vs overclocked CPU. Or running memory at 2133Mhz vs 3200Mhz. Some of us are tweakers thou. I notice a snappier system at lower latency.

 

[cadaveca]

These are real measurable figures, is the average person going to notice a difference? Probably not, but most people dont notice a difference if they're running stock CPU speeds vs overclocked CPU. Or running memory at 2133Mhz vs 3200Mhz. Some of us are tweakers thou. I notice a snappier system at lower latency.

For sure, and that's why I said "nearly no". See, the thing is, you need to know exactly what part of doing something with your PC that memory is actually used, and in a way that overall performance impacts the end result. Programmers do know this, and then tend to write stuff to deal with that effectively, so memory is very nearly never a bottleneck to system performance for average usages. At the same time, there can be no denial that low latency has an impact, and that is no matter what the measured throughput is. DDR4 overall introduced some measurable latency increases, but those are mostly dealt with within the structure that supports the memory, the memory controller and the interface to L3, so although measured latency is higher, performance is still faster Other than die shrink and minor changes to Intel's Core design, we've had the same CPU since SandyBridge launched, and it has been the GPU side of things that has seen the most improvements over the succeeding generations. However with all the changes to memory performance and the smaller transistors, there is a huge performance difference to be had moving from a 2600K to a 7700K, even if just by the raw clockspeed the CPU offers. Today we get CPUs whose natural clocks are those that we overclocked to with SB, and the GPU side of those chips is hugely faster too.

 

[X800]

Here is my result . I have kit of 2 of these G Skill F4-3200C14-8GTZ

 

[Vlada011]

This kit could fit perfect in my X99 RIG.
CORSAIR ask too much for 32GB of Vengeance LED and Dominator Platinum.
But somehow 16GB for now look me fine. Anyway Skylake X299 will need new kits.
I thought even to invest in Domy Special Edition... but no one could give me warranty for QVL on Skylake Extreme.

 

[iBruce]

Hey Dave,

I have the Trident Z F4-3200C12Q-32GTZSW kit. Having trouble with thermals in XMP 2.0.

Also have the Dual Channel kit for a Z170 rig, and when I set XMP with the Dual Channel kit, my total watts used sits around 135watts. Yet with the Quad Channel kit and a mild overclock of the 6800K at 4.4GHz, my total watts used reaches 200watts PLUS at idle. With the CPU overclock alone and memory stock, watts are around 145.

What am I doing wrong?

I'm using an Asus R5E10 motherboard.

 

[cadaveca]

What am I doing wrong?

Your board is likely changing other settings and voltages for the larger kit; VCCIO, VCCSA, and Input voltages are what I'd be looking at.

 

[iBruce]

Your board is likely changing other settings and voltages for the larger kit; VCCIO, VCCSA, and Input voltages are what I'd be looking at.

So I should drop the simple-click XMP 2.0 memory settings (on both kits) and set up these outrageous Trident Z 3200/14 sticks manually all the way.