FEATURED GEIL 8GB ( 2x4GB ) DDR4-3333 CL16 1.35V - GWW48GB3333C16DC

[Woomack]

Today I have something ( I hope ) interesting - 2x4GB DDR4-3333 CL16 dual channel memory kit from Dragon series.
What is so interesting about it ? It's probably the only enthusiast grade memory without heatsinks but has also unique looks what is not hard to notice on the attached photos.

Tested memory kit has been purchased in a retail store for about $100 inc. my local 23% tax. I just couldn't resist

As you see on the photos below, memory has white PCB and in these highest clocked kits also white memory chips. Since I mentioned memory chips then GEIL used Hynix SK IC and that actually explains pretty tight timings of 16-16-16 at DDR4-3333.

Photos were made on a fresh MSI Z170I Gaming AC, ITX motherboard which just appeared in the stores. I will try to show how high it's possible to overclock GEIL memory on this motherboard. If I notice any issues then I will check it also on ASUS Maximus VIII Hero.













I wish to see this memory available with red, green and yellow LEDs. Right now there are only blue what doesn't really fit to most gaming/enthusiast motherboards. Most of these motherboards are in red/black and lately also gray colors.

Since there is more photos than usual then I move everything else to the next post.

 

[Woomack]

Tests at XMP settings

All tests were performed with 6600K CPU overclocked to 4.2GHz. Also cache frequency is 4.2GHz as is all my tests. It's easy to achieve frequency so all readers can compare results to their home computers.
I will try to add 1h+ AIDA64 memory and cache stability test to all results. If it won't pass then I will let you know.

XMP: 3333 16-16-16-36 2N 1.35V

 

[Woomack]

Overclocking Results

2666 14-14-14-28 1N 1.20V

Spoiler

2800 14-14-14-28 1N 1.20V

Spoiler

3000 15-15-15-28 1N 1.20V

Spoiler

3200 16-16-16-28 1N 1.20V

Spoiler

2666 12-12-12-28 1N 1.35V

Spoiler

2800 12-13-13-28 1N 1.35V

Spoiler

3000 13-14-14-28 1N 1.35V

Spoiler

3200 14-15-15-28 1N 1.35V

Spoiler

3333 15-15-15-35 1N 1.35V

Spoiler

3000 12-14-14-28 1N 1.45V

Spoiler

3200 13-15-15-28 1N 1.45V

Spoiler

3333 14-16-16-28 1N 1.45V
These settings are actually passing everything @14-15-15 but are crashing in AIDA64 after about ~30-40 mins.

Spoiler

 

[Mandrake4565]

Certainly nice looking sticks, I'm looking forward to see your results Woomack.

 

[Woomack]

I added 1h+ AIDA64 stability @3333 15-15-15 1.35V to the 3rd post.
I'm not sure if I make it run higher. Next ratio is 3466 and at this frequency memory is not booting at all. Also at 3333+ it's not booting when I set memory voltage above 1.45V. Maybe it's mobo fault but Samsung sticks went up to 4000 without bigger issues.

I think that next step will be tighter timings at 3333 or lower clock. Maybe I will check higher clock on ASUS board other day.

So far I made couple of quick runs @3333 13-15-15 up to 1.45V but it's crashing in more demanding tests.

 

[Mandrake4565]

Woomack, you were able to pass 1 hr of the AIDA stress test but you can't run demanding benchmarks at 3333?

 

[Woomack]

Not at 1.5V+. That's the point, too high voltage at higher clock = stability issues on Hynix. This memory is 100% stable @3333 15-15-15 1.35V but won't even boot at 1.5V on my board.
It's like:
3000 can work at ~1.75V
3200 can work at ~1.50V
3333 can work at ~1.45V
3400+ can probably work at ~1.35V but 3466 requires higher voltage.
I'm not sure what am I missing here but the same was on X99 with some other Hynix based kits. Some liked higher voltages at 3200+, some not but all could run @3000 1.7V+

 

[Woomack]

All results at 1.20, 1.35 and 1.45V are in the 3rd post.
I can say that the most interesting looks 1.2V as not many other kits can even boot at 3000 15-15-15 1.20 or 3200 16-16-16 1.20V while this memory works stable.

On MSI Z170I Gaming Pro AC the lowest tRAS available is 28 so I couldn't check lower.
Also on this board I'm not able to make this memory boot at 3466 and bclk change is not working so I couldn't test it at ~3400 ( bclk stucks at 99.5MHz if I try to raise its frequency ).
I will probably add something more when I move to other motherboard or when MSI release any better BIOS.

All results were tested in popular benchmarks and 1h+ AIDA64 memory and cache stability test.

 

[dudleycpa]

All results at 1.20, 1.35 and 1.45V are in the 3rd post.
I can say that the most interesting looks 1.2V as not many other kits can even boot at 3000 15-15-15 1.20 or 3200 16-16-16 1.20V while this memory works stable.

On MSI Z170I Gaming Pro AC the lowest tRAS available is 28 so I couldn't check lower.
Also on this board I'm not able to make this memory boot at 3466 and bclk change is not working so I couldn't test it at ~3400 ( bclk stucks at 99.5MHz if I try to raise its frequency ).
I will probably add something more when I move to other motherboard or when MSI release any better BIOS.

All results were tested in popular benchmarks and 1h+ AIDA64 memory and cache stability test.

Thank you for your hard work!

I'm starting to think/realize that the "sweetspot" for Skylake seems to be 3000/3200. The tighter the timings the better (but it is for the benchers only). It is strange but on Newegg the voltage is listed at 1.35V Model GWW48GB3333C16DC. I'm starting to wonder if the memory running good timings at 1.2V is a good sign of good memory?

If I have this wrong please forgive me - did you say that in a different post that the motherboard was the issue with some testing? The MSI Z170I Gaming Pro AC may have been causing issues? If so have you tested on the ASUS Maximus VIII Hero?

 

[Woomack]

This memory has 3333 16-16-16 1.35V XMP profile so newegg is correct. I was testing it on various voltages.

MSI Z170I is a good board but there were only 2 BIOS versions. Even on initial version 1.0 Samsung memory can make 4000+ in dual channel what is great result considering it's "only" a gaming motherboard and what's more in ITX format. However Hynix IC ( on which is based tested in this thread Geil memory ) is not overclocking as good as on ASUS motherboards. I had no time to test Geil on ASUS M8 Hero as I have 2 reviews to finish but I will test it some time later. I can only tell you that this Geil is my best Hynix and "worse" Hynix kits could make 3400+ stable on M8H.

I see that Skylake likes higher frequency more than tight timings and 3466-3600 seems to be the best balance between timings and frequency. Most Samsung based 3000+ kits can make it without problems. Hynix usually can't make much above 3200 on lower voltages ( below 1.5V ).

 

[dudleycpa]

This memory has 3333 16-16-16 1.35V XMP profile so newegg is correct. I was testing it on various voltages.

MSI Z170I is a good board but there were only 2 BIOS versions. Even on initial version 1.0 Samsung memory can make 4000+ in dual channel what is great result considering it's "only" a gaming motherboard and what's more in ITX format. However Hynix IC ( on which is based tested in this thread Geil memory ) is not overclocking as good as on ASUS motherboards. I had no time to test Geil on ASUS M8 Hero as I have 2 reviews to finish but I will test it some time later. I can only tell you that this Geil is my best Hynix and "worse" Hynix kits could make 3400+ stable on M8H.

I see that Skylake likes higher frequency more than tight timings and 3466-3600 seems to be the best balance between timings and frequency. Most Samsung based 3000+ kits can make it without problems. Hynix usually can't make much above 3200 on lower voltages ( below 1.5V ).

Thank you for the update. It is an interesting time on O/C forums!

Being "only" an ITX gaming board it does very well. If I was a gambling biped (and you know I am) I would have thought the Asus M8 Hero would done at "least" as well as the MSI board.

Do you see any patterns where 3400+ and 1.5v will be the "new standard" for Skylake? This assumes the cost isn't too high - the Geil reviewed is $90 on Newegg. Much too expensive unless one is benching!

 

[Woomack]

New 3200 Samsung based kits can make 3600-3866 CL17/18 1.35-1.40V and cost about the same. Most of them won't make so tight timings at ~3200 but Skylake generally likes higher frequency more. In time we will see improved BIOSes for most popular motherboards so I expect that memory will overclock better.

Now about motherboards. Most gaming or standard series won't make more than ~3600 as BIOS is programmed for stability, not better overclocking. 2x8GB or 2x16GB kits can't make more than ~3400 and 16GB or more RAM is getting more popular lately so most users won't care if motherboard can make 4000+ or not.

ASUS ROG motherboards ( maybe except Ranger ) can all make 4000+ in dual channel. I couldn't pass 3600 on Ranger but maybe new BIOS is helping in higher clocks.
ASRock Z170 Gaming ITX could make 4000 only on 1 memory stick and probably won't make more. It also had limited settings in BIOS and max voltages of 1.5V for CPU and memory.
ASRock Z170 OCF is officially the best memory overclocking board but hard to say if other boards are not as good on new BIOS versions.
ASUS M8 Hero or Gene can make 4000+ and I actually passed that clock on Hero.
MSI Z170I Gaming can make 4000+ in dual channel and I also saw that frequency. I don't know what about Maximus 8 Impact but this MSI is better than any other ITX Z170 board on the market. On MSI site is info that they tested it with 4200 modules so it clearly has potential. I still have to check the latest BIOS ( released 2 days ago , date on MSI site is wrong ).

I'm writing this post on 3600 17-18-18 1.35V memory on the mentioned MSI board. I will probably check it soon at higher clock.

Back to the Geil memory, it's still one of the best options on the market and also great looking. Barely any memory can make 3333 15-15-15 1.35V. Even CL16 is good.

 

[Johan45]

Thanks Woomack, I still like the look of that stuff.

 

[Mandrake4565]

Thanks Woomack, I still like the look of that stuff.

Same here