RAM problems? Basic notions and tests with TeamGroup DDR4 RAM
This blog entry is aimed at PC parts stores and their workers (in Perú) as well as end users, who have ever had problems with RAM in some platform (AMD or Intel) and basic information of which very few speak… Although you might not believe, having problems with RAM is more common than one thinks and this was quite true during 2017, when Ryzen processors were launched for the first time.
The problems were serious, incompatibility of memories with motherboards everywhere; you could not reach high frequencies and even I was not free of that … since I was the first in my country (Perú) to obtain a Ryzen 5 1600, almost a month before its launch by accident (this is how the XanxoGaming initiative began) BTW, I did respect NDA even though I did not sign any paper.
Fixing all the problems around that launch (to make my memories work at 3200 MHz, because that was what I «had paid for» ) made me research and, above all, experiment with RAM memories a lot.
I learned many things and despite the fact that I might have had some criticism around Ryzen, the platform and its initial problems, helped a lot with my training as a PC Tech media, as well as being a «hardware tester» and enthusiast overclocker. Nevertheless, some people (stores here at Perú) told us that the information around RAM is not enough and they even get confused and think that the product (RAM memories) have come DOA (Dead on Arrival).
Only one disclaimer; This blog post is not a criticism of brand representatives (in Perú), because here at XanxoGaming, we respect the work of all.
In the world of brands, there are two types of representatives; those who are very good at sales , but lack a «technical expertise» and, on the other hand, those who have a certain degree of technical knowledge (this may vary) or a mix of both, but the feeling is that there is not enough explanation given for certain problems, which curiously is basic information that many may ignore…
As I mentioned before, this will help stores that lack this information, as well as the final consumer and brand representative (in this case, RAM).
In this opportunity, we have a new brand entering the country, TeamGroup and although many in Perú and other countries may not know the brand, they are quite recognized worldwide, for being one of the «top brands» for overclocking ( beware, we do not want to ignore the quality / prestige of other companies ).
Thanks to them, this special opportunity is presented as a blog post and we have several RAM memory kits to test them with the AMD Ryzen platform and educate of MANY of the VARIABLES that occur when connecting different PC parts ( CPU, Motherboard, RAM ) especially when things go wrong and the reason behind the possible failures.
In addition, we will see if these kits, work well or not with the motherboard of choice and try some simple overclocking. We will cover and talk about some basic concepts such as:
-The behavior of RAM memories in Intel/AMD
-The IMC (Integrated memory module)
-Motherboards… Not all Motherboards are equal … and why
-QVL List and my opinion / personal experience with this «norm».
-Tests on TeamGroup RAM in Ryzen
This blog entry will also serve as documentation. The products in test are:
-TeamGroup T-Force Vulcan DDR4 2400 MHz CL14 (2x4GB)
-TeamGroup T-Force Vulcan DDR4 2666 MHz CL15 (2x4GB)
-TeamGroup T-Force DELTA RGB DDR4 3000 MHz CL16 (2x8GB)
The motherboard of choice for testing, is still considered a budget mobo, the GIGABYTE B450M DS3H , for the AM4 AMD platform (Ryzen, etc) with a Ryzen 5 2400G processor (APU).
Let’s talk about the first topic of today …
RAM behavior in Intel/AMD
Have you ever had blue screens of deaths? Well, most of these events are memory related and the support these kits can have in motherboards. This is not the only explanation, since there can be a wide range of reasons behind, but there is a big correlation that links RAM memories with the famous blue screen of death.
Intel has enjoyed several years on DDR4 programming and its caveats and it has advantage against AMD in terms of maturity of RAM support (on software and hardware level). There are two important factors and these are:
-The memory controller or the IMC (this is 100% hardware)
-The programming / support of the platform
This is the reason why in Intel one can see, the existence of several memory kits with extremely high frequencies, like 4000 MHz and above, work in Intel… while AMD, most of the cases works up to 3200 MHz (except for some exceptions).
Frequencies above 3200 MHz in AMD Ryzen, may have compatibility problems or need further tweaking. If the support is correct on the RAM memory kit on a specific board, the main cause one frequency works or not lies upon the processor’s memory controller .
The IMC: Integrated memory controller
You will have heard or read, » all the processors are not born equal «. The same applies with the IMC (integrated memory controller); » Not all IMCs are born equal » and some may run high frequencies, as others may not. In Ryzen this means 3333 MHz-3600 MHz and in Intel, 4000+ MHz .
That’s why you see in motherboards, although they list high frequencies of memory support, they do not assure it and they have an asterisk (*). The reason behind this is the IMC.
AMD has greatly improved RAM memory and its support in both first-generation Ryzen , its Ryzen APU and Ryzen second-generation products , through AGESA microcode updates.
Apart from these updates, the support that a manufacturer, such as ASUS , ASRock, EVGA, GIGABYTE, MSI, can influence if a memory kit works well or presents BSODs. There is no miraculous formula that can ensure that all kits work without problems.
Mind you, this is something that not only happens in AMD Ryzen , but there are also strange cases in Intel, especially when using several DIMMs or high frequencies.
Motherboards: not all are equal
Each manufacturer has a different memory support, for each platform and this varies even by models within the same brand. I will give you a concrete example:
-Manufacturer A model X is compatible with Z memories.
-Manufacturer A model Y is not compatible with Z memories and presents blue screens.
Several entry-level boards ($ 40-60) will not have the same RAM support as a mid-range or high-end boards. These differences can be attributed to memory controller of the motherboard (not to be confused with the IMC), the design of the parts in charge of the memory, as well as its support (programming) of the manufacturer.
Do you remember that I told you that there is no «magic formula or silver bullet» that simplifies things? Well, with all this information, you might have better understanding why.
Despite this, there is a way to check the compatibility of a RAM model with the motherboard, and this is called the QVL List.
What is the QVL List? Explanation, opinion and my experience with this measure
QVL is the Qualified Vendor List . What does this mean? It is a list of components pre-certified by the brand, for a certain model, with different additional components (such as DDR4 RAM) that, according to the manufacturer, should work without problems .
We have the GIGABYTE B450M DS3H board, so let’s see a screenshot of how your QVL List looks with DDR4 memories . To get to this option, simply use Google, look for the motherboard model (in this case, GIGABYTE B450M DS3H ) and go to the support option (Support) as you can see in this screenshot.
Within this option, there is a sub-option called Support List where we will find several PDF files of different types such as:
-List of memory support, for different types of CPU (processor).
-M.2 storage support list
-List of support for disk in solid state (2.5 «)
In the case of memories, there are different lists for AMD Summit Ridge processor (Ryzen 1000 series) , AMD Raven Ridge (Ryzen 2000 APUs) and Pinnacle Ridge (Ryzen 2000 Zen+ series) .
Simply download the one you want.
This shows the QVL of RAM memories. In the case of the list for GIGABYTE, it is classified by frequencies (3200 MHz, 3000 MHz, etc.).
The QVL List, although it is a good reference to know your RAM kit works without problems with a specific motherboard, it is far from being the magic formula under our experience and as you will see today.
«It is the same as the synthetic benchmark for games, which, although they serve as a reference, are far from being an indicative of behavior in real life games.»
The TeamGroup T-Force DELTA RGB DDR4 3000 MHz CL16 memory (2x8GB) is not listed in the QVL, but despite this, it passed our stability tests.
In the QVL, RAM KITs are listed by a single unit (a single DIMM) despite that you maybe bought a kit of two (or 2x8GB in this example). To verify information on the list, consider your kits as a single DIMM (1x8GB).
This is the 3000 MHz list of the TeamGroup memories and the only one that is 1x8GB is the code TF1D48G3000HC16CBK . The TF1D48G3000HC16CBK corresponds to the TeamGroup T-Force Night Hawk RGB 3000 MHz CL16.
If we strictly follow the list, we would not consider the 3000 MHz CL16 Delta RGB memories with this motherboard…
Under my experience, several memories that are not in the QVL have worked without problems, as well as in some very strange cases, memories that supposedly are in the QVL have presented errors, that is blue screens (even with BIOS updates).
Some of the blue screenshots of memories that are in the QVL can be attributed to the processor’s IMC. In the case of Intel (using an extreme example), if you have a 4266 MHz RAM inside the QVL with an i9 9900K processor and have blue screenshots, the reason is probably related with the IMC.
As we mentioned before, not all IMC are born equal and some simply behave better than others.
Professional overclockers, have to their collection and disposition, «pre-selected» (binned) CPUs that they know they overclock quite high (CPU frequency), but also work at a high RAM frequency and is due to the pre-binning of several CPUs with good IMCs.
Meanwhile in AMD, 3200 MHz in most cases is now possible without problems, but frequencies of 3466 MHz, will depend on the support of the board with the RAM kit and mainly the IMC (silicon lottery).
This behavior does not only occur at high frequencies , since a motherboard in a certain brand, at frequencies like 2666/2933, there are very rare cases of problems despite the QVL. As I said, there is no magic formula…
There is also something additional to mention and is that, depending on the motherboard manufacturer, they might have outdated QVL in their database.
To summarize this section in one paragraph, the QVL is a reference, but it is far from being a perfect tool .
One last thing to mention; the only additional reason beyond what was mentioned of kits not working properly, is that the unit has come badly manufactured (rare, but may happen or even DOA).
TeamGroup in Ryzen Memory Tests
Why Ryzen platform? We already explained it before, but we will repeat it again. Intel has been working with DDR4 memories for more years, so it presents fewer problems in general.
Things have improved a lot with AMD , but there is still that doubt, especially in our country (Perú), so it is our platform of choice. Let’s see our test system:
Processor: AMD Ryzen 5 2400G
Plate: GIGABYTE B450M DS3H
OS: Windows 10 Home October’s Update
Storage: GIGABYTE M.2 2280 PCIe SSD 256GB
As additional information, we have configured to 512KB, as memory dedicated to the integrated graphics that comes with the 2400G. This amount is more than enough to have good performance with this type of processor with integrated graphics and E-Sports games (to which it is aimed).
The three memories that we will test are the following.
-TeamGroup T-Force Vulcan DDR4 2400 MHz CL14 (2x4GB)
-TeamGroup T-Force Vulcan DDR4 2666 MHz CL15 (2x4GB)
-TeamGroup T-Force DELTA RGB DDR4 3000 MHz CL16 (2x8GB)
Sincerely, one of the best RGB RAM (visually speaking and of course under subjectivity). In real life they are better appreciated.
What do we test?
We stress RAM memories to the maximum, to see if they are stable and no errors occur. In the event it happens, even if there is ONLY ONE ERROR , it means that they are not stable for daily use.
That one error in the tests, means that, in the middle of gaming, the program could close or generate a blue screen (BSOD).
In addition, we will include a test of AIDA64 (RAM Latency), for each memory kit. The results that matter the most to us (apart from memory bandwidth and that the memories are STABLE ) is the latency measured in NS (nano-seconds).
An improvement of 2-3ns is noticeable, and a greater decrease than this can greatly improve the experience (for example, latency reduction from 83ns to 74ns) especially for this type of processor ( Ryzen APU ).
To finish and this is VERY IMPORTANT, the validation, only serves for the model of specific motherboard and brand that is being tested.
The memory support varies on budget motherboard, to a high-end mobo. On budget mobos, the RAM support may be limited, and on high end mobo, it is better. Beware, this also varies depending on the brand of the motherboard manufacturer, but this works as a general rule.
Some manufacturers have better memory support than others, but even the best brand, a kit might end up with issues (blue screen, it does not start, etc) so take our test bench into consideration.
Let’s start with our first kit.
Note and additional advice: Please, whatever the manufacturer’s model (on AMD especially), update to the latest BIOS.
TeamGroup T-Force Vulcan DDR4 2400 MHz CL14 (2x4GB)
To our surprise, if we do not apply the XMP profile (we assume you know what it is and how it is applied in the BIOS) the kit presents blue screen problems ( surprise!). So we recommend applying XMP immediately, since, in doing so, the stability test passed without problems, lasting at least 6 to 8 hours. The more hours, the better.
To have a second reference, in our stability tests, the golden rule is to reach the validation of 1000% (some are more extreme and aim to have a validation of 2000%). We went a little over 1000% for all these tests, especially in overclock.
The validation of the memories with zero errors, as well as the primary, secondary and tertiary «Timings» of the RAM, once the XMP profile has been applied .
AIDA64 MEMORY BENCHMARK – TeamGroup T-Force Vulcan DDR4 2400 MHz CL14 (2x4GB)
The result in the AIDA 64 test was 83.4 nano-seconds . Personally, we suggest using at least 2666 MHz memories for Ryzen APUs , but this time, it is not the focus. Let’s see the second test kit.
TeamGroup T-Force Vulcan DDR4 2666 MHz CL15 (2x4GB)
This might be the most popular kit for budget builders, for both Intel and AMD platforms. Once installed, we applied the XMP profile immediately and passed the stability tests. Here is the screenshot:
The memories are rated at CL15 , but AMD Ryzen does not accept odd numbers in CAS Latency (tCL) . If we have a memory with this latency (CL15) we see that the RAM works as CL16; this is NORMAL . Another example would be, if we have a hypothetical 3000 MHz CL15 RAM kit, they will work at CL16.
Let’s see the improvement in AIDA64 (latency test).
AIDA64 MEMORY BENCHMARK – TeamGroup T-Force Vulcan DDR4 2666 MHz «CL15» (2x4GB)
As you can see, there was a reduction of almost 3 nano-seconds (less is better), and also an increase in read/write bandwidth. Let’s see the last kit under test.
TeamGroup T-Force DELTA RGB DDR4 3000 MHz CL16 (2x8GB)
As always, activate the XMP on the Delta RGB 3000 MHz CL16 RAM. The kit also passed the stability test under our conditions.
In AIDA64, the increment on RAM frequency, from 2666 MHz to 3000 MHz, increased the data transfer performance quite a lot, and also decreases the latency to 74.4 ns , almost 6 nano-seconds less, which in this «Test», is substantial.
AIDA64 MEMORY BENCHMARK – TeamGroup T-Force DELTA RGB DDR4 3000 MHz CL16 (2x8GB)
Extra: Overclock to RAM
Well, we could have used the DELTA RGB kit to overclock, but we chose one that is budget oriented. The memories of choice are the TeamGroup T-Force Vulcan DDR4 2666 MHz CL15 (2x4GB) . We will make a very simple overclock, which most of you can do by simply increasing the frequency of the memories, as well as the voltage .
First, apply the XMP profile so the primary timings can be applied on the motherboard settings and then modify the RAM frequency. The frequency on this board, appears as Auto (by applying the XMP profile) but change it to a higher frequency above rated XMP frequency.
A disclaimer before we continue. Depending on the CPU (remember the memory controller, IMC, as a factor) RAM and motherboard, what you can get as the maximum overclock frequency may vary.
We tested the following frequencies:
2666 MHz >>>> 2933 MHz using 1.35v
2666 MHz >>>> 3200 MHz using 1.40v
2933 MHz Overclock @ 1.35v
The memory was stable at 2933 MHz, using 1.35v . As you go up frequently, you have to turn up the voltage. 1.35-1.40v is safe for daily use. Here the respective capture of overclock stability.
The improvement of latencies in AIDA64 by «overclocking» from 2666 MHz to 2933 MHz are very good, with a decrease of almost 6 nano-seconds (less latency is better). Also the read/write/copy bandwidth is increased.
AIDA64 MEMORY BENCHMARK – TeamGroup T-Force Vulcan DDR4 2666 MHz «CL15» (2x4GB) >> Overclocking at 2933 MHz @ 1.35v DRAM Voltage
It is quite likely that, for most users, using this configuration and with the increased voltage, get stability with this kit and respective plate. We wanted to go with everything and check if increasing to 3200 MHz, it would be possible to achieve stability.
3200 MHz Overclock @ 1.40v
We did not expect the memories to be stable, since obtaining a frequency (once overclocked) is not 100% assured . 3200 MHz for Ryzen can be problematic at times, for the memory controller. Considering that the base frequency for this kit is 2666 MHz , obtaining 3200 MHz is no small feat.
The memories managed to pass our stability test. Here is the capture.
The improvement increasing to 3200 MHz is even more impressive, reaching 71.2 nano-seconds in AIDA64 Latency test. This is a huge increase of the 80.8 nano-seconds offered at 2666 MHz (only with XMP).
AIDA64 MEMORY BENCHMARK – TeamGroup T-Force Vulcan DDR4 2666 MHz «CL15» (2x4GB) >> Overclock at 3200 MHz @ 1.40v DRAM Voltage
For those interested if the kit worked more than 3200 MHz, I have to say that unfortunately not … even with more voltage. I could had tried lowering some auto timings, but that would be to go beyond manipulating only frequencies and voltage.
Annex: Information about the manufacturer of the RAM memories TeamGroup
This section will be the most summarized of all, since for the three models of memory have the same RAM chip manufacturer. These three TeamGroup models use Micron chips and are Single Rank (SR). Be careful and do not confuse Single Rank with Single Channel, they are not the same.
Screenshot and information from the memory kits.
Final words …
For those who have the habit of going to the final section of each article (wink wink) , I have a bad news… There is no holy magic solution or simplified rule («silver bullet)» that can summarize everything in very few words of what we wrote today.
I hope that, those who took the trouble to read until the end , have a better idea of the problems that sometimes can occur, when building, choosing pieces in this «gray area» between the interaction of these PC components (CPU , board and RAM memories).
I hope this post has been helpful and if you liked (or disliked) leave a comment, to see if the reception of this type of material is welcomed and see if I can write something similar of this nature in the future.