FEATURED Team Group Dark Pro 8Pack Edition 2x16GB DDR4-3600 CL16 - TDPPD416G3600HC16GBK

[gameinn]

@Woomack

I was looking more at your 3600 CL14-14-14-28 screenshot. Your aida scores are:

Read - 53,267
Write - 55,632
Copy - 48,071

Now refer to this video/spreadsheet table:

The comparable dual rank XMP profile with no subtiming tweak of 3600 14-15-15-35 has the following aida scores:

Read - 54,323
Write - 54,125
Copy - 52,518

You beat it on write but read was approx 1,000 behind and copy was far worse losing by 4,000. Why is this? I would have expected atleast 54,000 read and 52,000 copy.

 

[Woomack]

I run these tests at mostly auto sub-timings if it's not described otherwise. Depends on the motherboard, there are different sub-timings when you leave everything at auto. Also used CPU and its speed affects memory bandwidth in these synthetic benchmarks. The guy in the link posted results which are saying nothing about sub-timings or pretty much anything else. I could manipulate results to show magical +5GB/s but it's not the point. In these reviews I show what users can expect in, let's say, the easy way. It's not a performance contest.
I could give a full table of timings but on 95% motherboards it wouldn't work anyway. In the end, more advanced users are adjusting everything manually but most still wish to know what to expect so they won't waste money on a pretty average RAM that can't OC. Less advanced users won't play with sub timings anyway so they count to see basic settings and expected frequency. If anyone needs some tips then I'm always helping.

So again, there are so many variables that I can't tell you straight why there are these bandwidth differences. There are too many settings which are affecting that and don't believe when someone in the web says there were no tweaks. Even if they change one timing then some others under auto may also change.

 

[gameinn]

Ok it's just when I saw tRFC at 280 I thought it was tightened sub timings and not just 14-14-14-28 primary.

 

[Woomack]

I don't remember details, maybe I did some more changes. It was already a while ago but as I said, there are other variables which are affecting AIDA64 bandwidth so without full specs, can't compare results from two different sources. I usually focus on main timings and depends if it's Intel or AMD some additional settings (as I already said in my last post), while tRFC often gives more than main timings (at least when at auto is really relaxed). It's also one of the timings which is in SPD/XMP so if you leave it at auto then for DDR4-3600, depends on the brand, will run at ~560-780 when for Samsung it can easily run below 400. If you have high series OC motherboard then for Samsung IC there are OC profiles with full tables of timings. The same for most ASUS ROG or MSI motherboards. I wasn't playing with Gigabyte but their higher series used to have that too.
One more thing. On higher series motherboards, there is an option to boost results in AIDA64 (and some other benchmarks). Depends on your hardware it may add 1-2GB/s too. I'm never using it as I try to provide comparable results for readers.

There used to be rankings for AIDA64 results but no one is doing it anymore and I don't have time to keep it updated. It would be nice if more users shared some results as most only come here to read some stuff and ask for tips. Some years ago the whole OC community was sharing much more. Now most hardware forums are dead. Everything moves into gaming and because of the current hardware, there is no point to overclock (or just can't) so it's waste of time out of competitive benchmarking.

 

[BugFreak]

...I often run a couple of applications, multiple website tabs, and a game at the same time so sometimes I'm passing 16GB of RAM.

Woomack, what games are you seeing memory use over 16g? I've been kicking around going from 16g to 32g but so far my monitoring shows me staying around 10-12g while playing. I am not generally running a bunch of other apps too but I'm just curious.

 

[Woomack]

I play Black Desert Online and since this game has many "afk activities" then I often minimize it and play something like Civilization 6 or something else. BDO used to use about 8GB but there were some updates with data compression etc and now it's more like 5GB. Civilization 6 on a huge map is like 5GB+, especially when you play for longer. Win10, clean, is like 2.5GB but have to add some services, especially when the PC runs 24/7 and some stuff is all the time in RAM so it can be 4GB. Any web browser nowadays with a couple of tabs can be like 2-3GB+( I have maybe 20 pinned tabs and I've seen worse ). Photo editing in my way of loading multiple photos at once is like 10GB+ by itself. So I can use 16GB+ easily.
There is one more thing. Windows is locking some RAM predicting possible tasks. The more RAM you have, the more will OS reserve. I'm not even starting with SQL database as it will use much more RAM than it actually needs but at home it's not often used.

 

[gameinn]

Woomack, what games are you seeing memory use over 16g? I've been kicking around going from 16g to 32g but so far my monitoring shows me staying around 10-12g while playing. I am not generally running a bunch of other apps too but I'm just curious.

Watch these. 32 GB is about the experience.

 

[mov3on]

Today I have a bit unique memory kit as it's marked as the 8Pack Edition while the famous overclocker 8Pack is out of the scene for some longer. Either way, there is no Dark Pro memory kit listed on the Team Group website above DDR4-3466 while the tested memory kit is rated at DDR4-3600 CL16-16-16 1.35V. What makes it special is that the kit contains two 16GB memory modules and is also available at a reasonable price but the availability is limited to the overclockers.co.uk online store.
With the higher demand on RAM, I decided to purchase this kit for my daily PC. I often run a couple of applications, multiple website tabs, and a game at the same time so sometimes I'm passing 16GB of RAM.

Back to the memory itself. As it was expected, it's using Samsung B-die IC so if we wish to overclock and keep optimal performance then it's still the best IC around, even though not manufactured anymore.
The memory has no RGB lighting what more often is an advantage. I guess that all that RGB stuff is way too popular and many users have enough of that. Manufacturers started to focus on the lighting instead of improving performance or compatibility.

Below are some photos of the memory kit. It's black so perfectly fits my next build with the MSI Z490I Unify.

Spoiler

View attachment 210998

View attachment 210999

View attachment 211000

View attachment 211001

View attachment 211002

View attachment 211003

View attachment 211004

The Dark Pro has one XMP profile at 3600 16-16-16-38 1.35V. Officially there is a voltage range up to 1.45V but on the label, we will see 1.35V. Of course, there are no problems at XMP settings.

Spoiler

View attachment 210997

I'm not sure if it's a matter of early BIOS or something else but the memory overclocks great up to DDR4-4133 while above that can't even boot. Also scaling with the voltage is a bit weird and looks like:
- 3600 1.35V
- 3800 1.40V
- 4000 1.45V
- 4133 1.50V
Below these voltages, memory is not even booting, regardless of timings and other settings. For example, the memory can boot at 4000 CL14 1.45V but can't at 4000 CL19 1.40V. Also, I'm not sure if it's a matter of BIOS or the memory itself as I had no time on more detailed tests on other motherboards.

What you can count on is:
3200 CL13-13-13 1.40V+
3200 CL14-14-14 1.35V
3600 CL14-14-14 1.45V
3600 CL16-16-16 1.35V
3800 CL16-16-16 1.40V
4000 CL16-16-16 1.45V
above is still in tests and the highest stable setting so far is 4133 CL18-18-18 1.50V what doesn't look any special.

The tightest at DDR4-3600 is CL12-12-12-28 1N 1.6V. It wasn't stable but I wasn't expecting that. Here is the AIDA64 screenshot.

View attachment 210992

All tests were performed on the i5 10500 CPU, MSI Z490I Unify motherboard and MSI GTX1660Ti graphics card. Nothing really special but enough for some good results.

I have bought exactly same kit (I think) for my Ryzen 5000 system. Was hoping I will be able to run it at 4000 C16, but sadly I can't.

For some reason 3800MHz is as far as I can go, and even 3800 is not 100% stable, so I'm running it at 3733MHz. When I set the clock to 3866 - PC doesnt even boot, BIOS is reseting. I have even tried 1.5V and super loose timings, didn't help. Maybe I was unlucky with the silicon, idk, or maybe I've got slightly different kit.

 

[Woomack]

My 5900X (and 5600X/5800X before) runs exactly the same so it's like:
up to 3800 1:1 100% stable with 2x8GB
up to 3733 1:1 100% stable with 2x16GB
up to 3733 1:1 100% stable with 4x8GB
up to 3733 1:1 100% stable with 2x32GB
up to 3600 1:1 100% stable with 4x16GB
up to 3600 1:1 100% stable with 4x32GB <-- this one requires fully manual settings or is crashing at XMP

So far I've seen only a couple 5950X that could reach DDR4-4000 ... and only 5950X. On the other hand, results at DDR4-4000 are not much better.

Personally, I like to play with Micron IC more nowadays. No special performance differences out of synthetic benchmarks but somehow more satisfaction with results.

 

[mov3on]

My 5900X (and 5600X/5800X before) runs exactly the same so it's like:
up to 3800 1:1 100% stable with 2x8GB
up to 3733 1:1 100% stable with 2x16GB
up to 3733 1:1 100% stable with 4x8GB
up to 3733 1:1 100% stable with 2x32GB
up to 3600 1:1 100% stable with 4x16GB
up to 3600 1:1 100% stable with 4x32GB <-- this one requires fully manual settings or is crashing at XMP

So far I've seen only a couple 5950X that could reach DDR4-4000 ... and only 5950X. On the other hand, results at DDR4-4000 are not much better.

Personally, I like to play with Micron IC more nowadays. No special performance differences out of synthetic benchmarks but somehow more satisfaction with results.

Good to know I'm not alone. I thought maybe I'm doing something wrong, or my kit is just not good.

 

[Woomack]

Sadly, it's far from promised better IMC and higher IF clock. My 3700X and 3900X had better IMC/IF. For pure memory tests, I'm using Ryzen 4650G as it goes up to 4533 1:1 and 5400 1:2 stable.