Trident Z 3600 2x8 Cl 17-18-18

[Johan45]

This is a new revision of an older kit. It's Samsung single sided inside but die rev. is unclear at the moment.
Just wanted to post some preliminary results.
Stock 3600

3733 17-18 stock voltage 1.35

3866 same timings 1.37v


Here's a couple I've been working on 3733 Cl 13-13-13 1.63v



3943 16-14-14 1.63v

 

[Woomack]

I just installed 3600 16-16-16 kit. It's even working on XMP Weird was that on the heatsink of one module are 2 scratches and someone clearly had problems to put memory correctly into the box. I wouldn't expect that in Trident Z series.
I found only 1 kit in Polish and central EU stores. Price was the same as for most 3400 kits.

I see that tRFC is really high. My kit has over 631 on XMP. Thaiphoon Burner is reading IC as K4A8G085W?-BCPB
This IC is also in 16GB modules so I only wonder how will look 32GB modules ( if will be released as non-ECC ).

Maybe will post something later as these kits are on about the same IC.

Edit:

When I set XMP and manually 4000 then MSI Z170I set 18-18-18 1.35V:

 

[Johan45]

Nice, my Hero was having issues at 4000. Just below was OK.
Does your sticker say 5000 ?

 

[Woomack]

I'm testing on MSI or maybe I just put this kit into my daily rig and is working @4000 18-18-18 1.35V for last 2h. On the sticker there is 5004. IC seems the same as yours ( at least Thaiphoon says that ).
Maybe I will test this memory on ASUS but as I said, MSI is overclocking better. The only issue can be that MSI on the latest BIOS can't work at SA above 1.25V and that was limiting max clock on double sided kits. On the other hand early BIOS releases were overclocking slightly worse but had SA/IO up to 2.0V.

 

[Johan45]

That's weird that they would limit the SA like that. I know on auto mine shotts up to 1.3 but I dialed it back to 1.25 which I still thin is a lot. maybe I'm wrong. I'll have to give that a try at CL18-18-18 trfc700 and see if it behaves better. I still think some auto setting sare messin with me.

 

[Woomack]

I made it work @4133 18-18-18 1.35V and even booted @4200+ but for some reason when I pushed too much then after freeze and restart, bclk adjustment stopped to work. I'm not sure why but I can't make it work now even after clearing BIOS. My board has no higher ratio than x41.33.

 

[Johan45]

Uh Oh not another broken board ha ha

 

[Woomack]

After some more pushing at lower voltages I got info that something isn't right with my BIOS and best will be to reflash it. After recovery board still can't run at any other bclk than ~100. I'm not sure what could happen. I only set too high clock, board didn't want to boot so I entered BIOS, lowered bclk ... and since then it was only booting at 100MHz. Weird
I think I should make validation at 4270 when I could ... now I'm not sure if I make it work so high again. I can try to reflash BIOS as standard recovery is not writing all sectors.

Anyway, at 4133 it works pretty stable at 16-16-16 1.40V. I made only couple of quick tests as it was already 2 am and I had to go sleep. Today overslept to work but that's other story.

I think that on better BIOS or board this 3600 CL17 kit will run about the same. My kit has no problems to boot @ 4133 when I only set XMP and everything else at auto. If your memory can make ~4000 at lower timings then it's more than possible that it will work at ~4200+.
Btw. I was trying to set CL20-22 manually but my kit was less stable than at CL16-18.

 

[Johan45]

I tried that too, and that didn't help me either.

 

[Mandrake4565]

Interesting stuff boys, thanks.

 

[Woomack]

At the end it looks like without higher SA voltage I won't stabilize my memory @4133+. 4133 is working in benchmarks at 16-16-16 ~1.4V but is randomly crashing. The same when I set CL18, CL19 or 1.45V+ so it's not the memory. 4000 seems 100% stable in everything. I will try to downgrade BIOS to older version with higher voltages. Well, I don't have big choice, there is one version which is not causing additional issues and I can't find any betas.

 

[QuickFast]

Johan looks like all single sided 8gb sticks are samsung-B's can't wait to see what you two can do whit these sticks -there go's my XTU score I was going to buy the 3600cl16 the other day but I bought a i3 6320 instead

 

[Woomack]

Probably better idea point-wise for rankings. Memory above some point is not helping much in most tests. I guess that the best results will be at 3600-3800 but tighter timings. Between 3800 and 4133 there is like 1GB/s difference in bandwidth but also lower clock runs at much tighter timings so latency is lower and that helps in Spi32M and XTU.

Yesterday I was playing with MSI. On older, 1.3 BIOS I can set 1.33V VCCSA max ( 1.25V on 1.4 BIOS ). It let me set slightly higher max clock but Win10 is randomly crashing. BIOS is also randomly crashing and then bclk adjustment isn't working. The only way to fix it is to make full BIOS recovery.
At the end max memory clock is ~4260 19-19-19 1.45V. Higher voltage or more relaxed timings are causing additional instability.

 

[QuickFast]

Looking good Woomack -the 4000 retail sticks with the same ic's are rated at cl19-25-25 1.4volts

@Johan looks like you might have to wait for a new Bios for anything over 4000 with these sticks your the third person I saw hitting a instability wall at 3950 on a Asus board

 

[Johan45]

I wondered about that and that's exactly where it pukes too. I can take it up in windows one FSB at a time and after 3950 AIDA starts giving score about half what they should be.

 

[Woomack]

Looking good Woomack -the 4000 retail sticks with the same ic's are rated at cl19-25-25 1.4volts

So far memory works great. I'm only starting to think about switching motherboard but I'm not sure if it's the best idea since I'm only testing memory and writing reviews lately. I mean unnecessary cost especially that my budget is pretty low. I already sold 2 other memory kits to get that one. Not much more left for sale as I sold most of my stock to get laptop over a month ago

There are only 8GB and 16GB modules based on this new IC. There are no 2x8GB kits rated at more than DDR4-3600. 2x4GB kits are on other IC. It doesn't mean it's worse, just different and as we see, higher frequency is helping in some tests more than timings. I have almost the same results in quick tests like AIDA64 at 19-19-19 and 19-22-22 or 19-25-25.

Probably sub timings on higher density IC are causing some performance drops but it shouldn't be more than 2-3GB/s in tests like AIDA64. If results are about half than it should be then memory controller or motherboard have some issues with stability or memory is running in single channel mode ( even though CPUZ can show dual channel ).

On the Hero I had no problems to set high clock on 2x4GB Samsungs but on 2x8GB all results are better on MSI board. I guess that ASUS still hasn't updated BIOS for these higher density chips. MSI after last updates runs slightly better but they locked SA voltage so final results are about the same on 1.3 and 1.4 BIOS. It doesn't change fact it's a great result for ITX gaming board. There is G.Skill 4266 kit on the tested memory list too but I'm not sure how they made that as on my board bclk above 103 causes instability and for 4266 is required ~103.5MHz.

 

[QuickFast]

https://www.alternate.de/GeIL/DIMM-16GB-DDR4-4000-Kit-Arbeitsspeicher/html/product/1242732

 

[Woomack]

There are big problems with this store. They have a lot of stuff on the website ... but barely anything on stock and they make problems with shipping. At least they have office in Poland and are shipping all from Germany but most customers are complaining at everything.

Knowing Geil there is a 50% chance it will work at declared settings ... The only kits which I had and were working as declared were 2x4GB DDR4-3333 Dragon and 2133 Potenza ( it would be bad if it couldn't run at 2133 ). All who were buying their other DDR4 series, were complaining they can't even run it near declared speed ( kits above 2800 ). I mean check super luce reviews. 3200 kits could barely run at 3000, 3400 kits could run at about 3000-3200. We had some forum users who were excited that they ordered higher frequency Geil and later were not even showing results. I think I saw someone selling memory not much later.

I can say that Potenza 2133 kit which I had runs great and friend has it for last half year at 3200 stable. Dragon 3333 as you could see in my tests is also good but is not overclocking high because of used Hynix IC.

This kit in link looks good. I only wonder what will be real price. If price in link is actual then it costs about 30% more than TridentZ 3600 16-16-16. Anyway nice find. I bet we will see more kits on this IC soon.

 

[Johan45]

Yeah that's a crazy price just for a ram kit

 

[QuickFast]

It was Rule that pointed them out look at his timings looks like his OS might be holding him back from scoring higher
http://hwbot.org/submission/3080476_rule_xtu_core_i3_6320_738_marks