We’re continuing our DDR5 memory kit tests, as we’ve received several over the past few days, including a copy of Teamgroup’s T-Force Xtreem, clocked at 8200 MT/s in CL38. Suffice to say, this is one crazy piece of kit! Our example of the day consists of two 24 GB memory sticks and requires a voltage of 1.40 volts. As a reminder, our test configuration has just been updated and we’re now using the Maximus Z790 APEX combined with an Intel i9-13900KS processor. In recent months, with the arrival of the Hynix A-die and the new Hynix M-die chips, frequencies have risen sharply, requiring a test configuration capable of supporting these XMPs and, of course, overclocking.
Don’t hesitate to ask us any questions you may have, either in comments or on our social networks. You can also join our community on the discord. We’d also like to take this opportunity to thank Teamgroup and especially Kevin Wu for giving us the chance to review this 2 x 24 Teamgroup T-Force Xtreem 8200 MT/s CL38 kit. We’re one of the only sites in the world to have received one!
We take a look at the kits available to us:
Since this is an article in which we’ll be doing comparative tests, we thought it would be a good idea to review the various kits we have. You’ll be able to find out more about the features of each one. You can also take a look at the section where our latest memory kit tests are available, and find the one that interests you. Don’t forget that as of today, our test platform is evolving, so we’ll have new performance tables.
- 32 GB Corsair Dominator Titanium 7600 MT/s CL36 (Hynix A-die)
- 32 GB Corsair Dominator Titanium 7200 MT/s CL34 (Hynix A-die)
- 48 GB Corsair Vengeance RGB 7000 MT/s CL40 (Micron)
- 32 GB GSKill Trident Z5 RGB 6400 MT/s CL32 (Hynix M-die)
- 32 GB Corsair Dominator Platinum RGB 7200 MT/s CL34 (Hynix A-die)
- 32 GB GSKill Trident Z5 RGB 6800 MT/s CL34 (Hynix A-die)
- 32 GB Teamgroup Deltra RGB 7200 MT/s CL34 (Hynix A-die)
- 32 GB Kingston Fury Renegade 6400 MT/s CL32 (Hynix M-die)
- 32 GB Corsair Vengeance RGB 6400 MHz CL38 (Hynix M-die)
- 32 GB Teamgroup Delta RGB 6600 MT/s CL34 (Hynix M-die)
- 32 GB Corsair Dominator Platinum RGB 6600 MHz CL32 (Hynix M-die)
- 16 GB Kingston Fury Beast RGB 6000 MHz CL40 (Hynix M-die)
- 32 GB Corsair Vengeance 5600 MHz CL36 (Samsung)
- 32 GB Teamgroup Delta RGB 6400 MHz CL40 (Hynix M-die)
- 32 GB Dominator Platinum RGB 5600 MHz CL38 (Samsung)
- 32 GB Kingston Fury Beast 4800 – 5200 MHz CL40 (Micron)
- 32 GB GSKill Trident Z5 5600 MHz CL36 (Samsung)
[nextpage title=”Packaging description:”]
Teamgroup’s kit is delivered in a cardboard box with a rather sober color code, different from the kits previously tested. Indeed, our sample of the day is part of the « Xtreem » series and the previous ones of the « Delta » series. The brand emphasizes the design of its kit, with a reproduction of its black module on the front panel.
On the rear, a label informs the future purchaser of the characteristics of the memory kit. There’s no cut-out to reveal the memory modules, as was the case with the « Delta » series.
There are other references in the “Xtreem” series, with a 7600 MT/s and 8000 MT/s kit available in 2 x 16 GB or 2 x 24 GB and our kit of the day, clocked at 8200 MT/s but only available in 2 x 24 GB.
Features and cooling :
Our kit is clocked at 8200 MT/s, has timings of 38-49-49-84 and requires a voltage of 1.40 volts to operate. It consists of two 24 GB modules for a total of 48 GB.
Heatspreaders are made of black aluminum. They are responsible for dissipating the heat generated by the PCB and memory chips. The heatspreader is aesthetically pleasing, with a bevelled top and fins to optimize cooling. It is 2 mm thick to increase its thermal capacity . The first side features the superb round “T-Force” logo and the inscriptions “Xtreem” and “Narvik Black – DDR5 – From T-Force Lab”. In the introduction, we mentioned Kevin Wu, who regularly posts tests of the kits on Facebook, enabling us to assess their excellent potential.
On the second side, a label lists some of our kit’s features. The dimensions of our T-Force Xtreem are 48.8 x 134.5 x 8.2 mm. As for the price, we haven’t yet managed to find it listed.
To conclude this presentation, please note that the kit is XMP 3.0(Extreme Memory Profile) compatible and comes with a lifetime warranty. So don’t forget to activate the profile directly via our motherboards’ bios.
[nextpage title=”How to identify your memory kit:”]
Basic features :
You’re going to tell me that this is pretty easy, since all you have to do is look at the references advertised by the manufacturer. Yes, but that’s not enough, especially if you’re planning to overclock it! Some information is given by the manufacturer, but you’ll have to dig around or even remove the heat sinks to find out the rest.
One side of the heatsink bears the words “Xtreem” and “Narvik Black – DDR5 – From T-Force Lab”, while the other has a label with the technical specifications of our model. This is what we’re going to look at in detail. For example, in the case of today’s model, here’s the information we know.
- Brand: Teamgroup
- Model: T-Force Xtreem
- Capacity: 2 x 24 GB
- Frequency: 8200 MT/s
- Timings: 38-49-49-84
- Voltage: 1.40 volts
- Review: –
In order to get an idea of the overclocking potential of your kit, this information is insufficient and you’ll need to go further. It’s also very important to know the revision of the PCB and the model of the chips used in your memory kit. Depending on this information, the kit’s overclocking performance will be very different.
To find this information, there is currently only one solution: disassemble the heat sinks. The Taiphoon Burner software does not currently support DDR5 kits. Nor is there any indication that this will be the case in the coming weeks.
CPU-Z always present :
Fortunately, our trusty CPU-Z software is always on hand, allowing us to verify this information, particularly with regard to the type of chips used in our kit. At present, there are three types of chip: Samsung, Hynix and Micron.
Our kit is perfectly recognized under Windows 10, and the various timings read perfectly. The memory chips used in our kit are Hynix. These are the chips with the best overclocking potential. However, one element catches our attention: the CPU-Z SPD tab shows an XMP profile at 7400 MT/s. Let’s reboot our PC to see if this profile is present in the BIOS.
The PCB of our memory kit :
We’re going to dismantle the heatsinks so that we can observe the PCB and identify the chips. Be careful not to accidentally detach any chips from the PCB! The best technique is to heat the heat sink with a hair-dryer, then gently pry it out, and be patient. As a reminder, the design of DDR5 modules is different from that of DDR4, with the presence of an on-board PMIC (power management integrated circuit) chip. The idea is to enable better power management and therefore system stability. Illustration photos courtesy of Kevin Wu, thanks to him.
Our kit therefore benefits from the Hynix H5CGD8MGBD-X021 3 GB chips, of which there are 8. As the “M” in the part number stipulates, we’re using Hynix M-die and not A-die chips, but this is a new generation that allows us to further improve the frequency rise. Our kit is an “SR” model for single rank, meaning that all the chips are on a single side. On the other hand, there are “DR” kits for dual rank , with chips on both sides of the PCB.
[nextpage title=”The importance of the XMP profile:”]
BIOS settings :
We can’t stress this enough, but your memory kit ‘s parameters are not automatically set once you’ve installed it. If this is the case for your processor, it’s not so for your memory kit, and you’ll need to go through the BIOS! This is what we had to do with our Maximus Z790 APEX to get the BIOS to load the XMP profile.
You can then configure it manually, choosing the right frequency, timings and voltage for your kit. For ease of use, we advise you to activate XMP (Intel) or EXPO (AMD) mode in « AI Overclock Tuner » to set it up automatically.
Without this, your performance will be impacted in all areas, and you’ll end up with a kit that doesn’t exceed the CL40 frequency of 4800 MHz. Performance will then be extremely limited. So we’re going to activate our XMP profile directly in the BIOS. And boom, good news: in addition to the XMP profile at 8200 MT/s.
ROG BIOS, whether on APEX or Crosshair models, benefit from memory profiles. They are a good basis for overclocking your memory kit. We’ll start by using them in our overclocking tests. Our kit is now set up in the BIOS and will work with the manufacturer’s specifications.
As you can see from the screenshots above, once the XMP profile has been activated, the settings are automatically detected. The frequency is set to 8200 MT/s, timings 36-49-49-84 are activated and the voltage is increased to 1.35 volts.
[nextpage title=”Intel configuration test:”]
Of course, this kit will only be tested on an Intel platform, as our comparisons have been based on this configuration from the outset. We’ve naturally opted for the ROG Maximus Z790 APEX motherboard, but we’re also testing the latest refresh model in parallel , the ROG Maximus Z790 APEX Encore version.
- Motherboard: ROG Maximus Z790 APEX (BIOS 0084) APEX ENCORE (BIOS 0070)
- Processor: Intel i9-13900KS
- Memory: Teamgroup T-Force Xtreem 8200 MT/s CL38
- Graphics card: MSI GT 710
- SSD: Western Digital SN850 NVMe SSD M.2
- Power supply: be quiet! Dark Power PRO 12 1500w
- Operating system: Windows 10 64-bit (22H2)
We also decided to stick with Windows 10 (22h2), as it performs better, especially when overclocking. We’ve noticed a significant drop in processor and memory performance since the latest Windows 11 updates, which sometimes distorts the results. Our benchmarks are run a minimum of three times, and the result is the average of these scores. Remember that we’ve also added the performance of our kit of the day with the XMP profile at 8400 MT/s CL38, since it’s shown to be perfectly stable H24. It would be a shame to do without it.
Aida 64:
This is a program that can provide you with a wealth of information about your configuration, but above all it has a memory benchmark. This calculates the read, write and copy speeds, as well as the latency, of your memory kit. It is often used to compare memory kits. It’s an excellent tool if you want to compare the performance of your memory kit with those we’ve already tested.
The version used for testing is 6.92.6603 beta, and we’ve used the performance of previously tested kits.
The performance of this first test is exactly what we expected. It starts off very strongly, with read speeds of 125,727 MB/s for the XMP 8200, rising to 128,230 MB/s for the XMP 8400. This is quite logical, since this Teamgroup Xtreem kit combines high frequency with fairly tight timings . So, on an Intel platform, you’ll need to prioritize frequency and tight timings.
The same applies to write speeds, which are very good, reaching 123,410 MB/s at 8200 MT/s CL38 and even 126,430 MT/s at 8400 MT/s CL38. The graphics logic remains identical to that of the previous model. The Corsair Dominator Platinum 4800 MT/s CL19 in DDR4 achieved 67,251 MB/s. Once again, we note a logical progression between the different kits according to their frequency and timing.
The findings are identical in copy, with once again the same duo at the top of our graph. The Corsair Dominator Platinum 4800 MT/s CL19 in DDR4 achieved a copy speed of 57330 MB/s. As you can see, it’s in reading and copying that DDR5 takes a clear lead over DDR4. Note also the difference between a DDR5 kit at 4800 MT/s (72,415 MB/s) and our kit at 8400 MT/s (123,193 MT/s). If we hadn’t activated our XMP profile in the BIOS, our kit, although sold as a 8200 MT/s model, would really have been clocked at 4800 MT/s. A reminder of the importance of enabling the XMP profile in the BIOS.
Finally, latency. This is the value on which all eyes are focused, as DDR5 lags a little behind. Our best DDR4 kit in this respect was the G.SKill Trident Z Royal Elite 4000 MT/S CL14, which achieved 41.4 ns. You’ll notice that the higher the DDR5 frequency, the narrower the gap. Note that a 3200 MT/S CL16 DDR4 kit has a latency of 50.3 ns. Here, it’s 56 ns for our Teamgroup T-Force Xtreem 8200 MT/s CL38.
Geekbench 3.4.4:
This benchmark is available in several revisions, version 5 having just been released a few weeks ago. It provides two scores: one single and one multi. Using this benchmark, we’re going to focus on the performance obtained on the ” Memory Multi-Core ” score, which differentiates memory performance.
The version used for testing is 3.4.4.
For the Geekbench 3 benchmark, despite the single-core frequency of our i9-13900KS being 6 GHz, we note a difference in score between the different memory kits. This is the first time that our kit of the day has fallen slightly behind, even if the difference is almost non-existent.
In multicore, you can’t change a winning duo, as our Teamgroup kit retakes the lead.
As soon as we isolate the memory score, real differences are visible between the kits. Our kit of the day scores 14,233 points! We’ll be keeping a close eye on this score when we start overclocking. It’s a good indicator of the benefits of overclocking.
In DDR4, our best kit scored just 7,748 points. Once again, DDR5 combined with the Raptor Lake processor offers a huge advantage.
Geekbench 5.5.1:
We’ve decided to add this extra benchmark, but why? The CPU benchmark uses new tests that more closely simulate the tasks facing processors in the face of recent applications. Geekbench 5 also increases the memory used in the benchmark to better reflect the impact of this parameter on CPU results. For the time being, we’ll be keeping both, as version 3 is still widely used. We have also migrated to the latest version, 5.5.1.
Whereas in Geekbench 3, we had a difference in single core, here in version 5, the gaps are smaller but, as in the previous version, we find a slight delay in our Teamgroup kit.
Finally, the score for our Teamgroup T-Force Xtreem kit at 8400 MT/s CL38 is 28,588 points. Still to give you a reference score in DDR4, our best score in Geekbench 5 was 11,179 points.
Geekbench 6:
We recently received a key for Geekbench version 6. We therefore took advantage of the change of configuration to add this new benchmark to our test suite. This is the latest benchmark and is intended to be more representative of current configurations. Primate Labs also assures us that the scores displayed by Geekbench 6 are more accurate in relation to the various existing devices and platforms. In fact, the test is much longer than the other two versions we use. Our tests will be carried out on the latest version, i.e. 6.2.2.
The scores in the single test are fairly close, which explains the slight but ultimately insignificant change in the leading order. Our sample of the day once again stands out from the crowd.
Let’s conclude this series of benchmarks with the multi part of Geekbench 6, where we end with the performance configuration encountered during all our tests. There’s no doubt about it, this is the best-performing kit we’ve had our hands on.
[nextpage title=”Overclocking on our APEX Z790:”]
Choosing your DDR5 kit according to memory chips :
As was already the case for DDR4, if you want to clock your memory, certain brands will benefit from better potential. At present, the kits announced by the various brands benefit from either Micron, Hynix or Samsung chips. The latter two brands will be the most interesting. As you can see, our Teamgroup T-Force Xtreem 8200 MT/s CL38 kit is equipped with Hynix chips. Unlike our last kits, these are not Hynix A-die chips, but the new Hynix M-die chips. These have the particularity of rising more easily in frequency, but being more complicated to tighten in terms of timing.
Key DDR5 voltages:
For memory overclocking, there are 5 key voltages. These are the voltages that will have an impact on overclocking and therefore on frequency rise or timing tightness. These are the CPU System Agent Voltage (SA), the DRAM VVD Voltage, the DRAM VVDQ Voltage, the Memory Controller Voltage and the IVR Transmitter VVDQ Voltage. Here’s where you’ll find these voltages in the BIOS of our ROG Maximus Z690/Z790 APEX.
Please note that, depending on the type of memory chip (Micron, Hynix and Samsung), the balance between these different voltages is different. If your PMIC module is not locked, by activating ” High DRAM Voltage Mode “, you’ll have access to voltages up to 2.00 volts. Now it’s time to overclock this Corsair DDR5.
Step 1: use available BIOS profiles
It’s a strange thing to say, but a few months ago, we were happy to have the BIOS profiles available as a good basis for our work. However, since the arrival of the Hynix A-die and now the new Hynix M-die chips, the frequency of kits has exploded, and in the end, the profiles in the BIOS are lower than the current frequency of our kit. There’s a 7600 MT/s CL36 profile and an 8800 MT/s profile for Hynix A-die chips, so we’ll have to build our own profiles.
Step 2: 8800 MT/s CL38 target
This is our first objective, to reach the 8800 MT/s CL38 frequency, as this type of chip should be able to do this quite easily. We set up the BIOS, modify voltages, timings and subtimings, and off we go.
The good news is that it starts without a hitch, and our two benchmarks go straight through. We’ve run several of them in a row, and there’s nothing to worry about in terms of stability. Finally, I thought the IMC of my i9-13900KS was limited to 8700 MT/s, but that’s not the case. In fact, this is the first time I’ve reached 8800 MT/s. As you’d expect, throughput in AIDA64 is through the roof, with read speeds reaching 137,650 MB/s and a Geekbench3 memory score of 16366 points.
Step 3: 8888 MT/s CL38 with 101 MHz BCLK
As my tests progressed, I found it impossible to boot up to 9000 MT/s. The only way to continue increasing the frequency is to increase the BCLK to 101 MHz. This will have the effect of increasing the memory frequency at the same time.
So here we are, with the same voltages/timings at 8888 MT/s. Performance continues to improve, with AIDA64 reading at 139,170 MB/s and Geekbench3 memory score at 16425 points.
Stage 4: 8950 MT/s CL38
It was from 8888 MT/s onwards that things got complicated, as I was no longer stable at all. In reality, and after many hours of testing, it was the SA voltage that was the key. We also lowered the voltage, since it’s now stable at 1.59v VDD and 1.47v VDDQ. We also deactivated the E-core and therefore carried out our benchmark at 8 P cores and 16 threads.
We never thought we’d get to 8950 MT/s in CL38 with this Teamgroup T-Force Xtreem 8200 MT/s CL38 kit! It’s simply enormous, especially to see our i9-13900KS IMC take it all in stride. Performance continues to improve, with read speeds of 139,670 MB/s under AIDA64 and a memory score of 16620 points under Geekbench3.
Step 5: yes, the 9000 MT/s CL38
Needless to say, when we’re at 8950 MT/s, the goal is of course to reach 9000 MT/s.
We achieved this under AIDA64 with a read score of 139,810 MB/s. As far as Geekbench3 is concerned, if you’re into memory overclocking, well, it’s the dijkstra passage that cuts the benchmark. We’re convinced that cooling the processor should solve the problem. We’ll be installing our DOD very soon.
[nextpage title=”The last word:”]
Incredible, that’s what best sums up our test of the Teamgroup T-Force Xtreem 8200 MT/s CL38kit. Before starting our article, we’d never imagined we’d be able to reach a frequency of 9000 MT/s. While the new Hynix M-die chips in the Teamgroup T-Force Xtreem 8200 MT/s CL38 kit perform very well, we mustn’t forget that our ROG Maximus Z790 APEX (Encore) motherboard and the IMC of our i9-13900KS also have a lot to do with it. It’s the combination of these three components that makes it possible to reach such high frequencies.
Teamgroup has also done a great job on the heatsink design to optimize cooling. As for the price, we don’t yet know, for the simple reason that this model has not yet been listed. We found the 7600 MT/s and 8000 MT/s models on the Octo24 website, priced at 247.98 euros and 297.99 euros respectively. We imagine that the 8200 MT/s version should be priced at around 320 euros, which would be an excellent price.
Teamgroup T-Force Xtreem 8200 MT/s CL38 :
We liked :
- A kit for overclockers!
- The original 8200 MT/s XMP, ready for high-end confis
- The sober design of the memory modules and their design to optimize cooling
- High-quality finish of memory modules
- The best-performing Hynix M-die DDR5 kit we’ve ever had!
- Excellent overclocking potential combined with the Z790 APEX and a CPU with a good IMC
- Lifetime warranty
We would have liked :
- greater availability: the kits are just arriving, but it remains to be seen how many will be available in Europe. We’ll keep our fingers crossed.
This is our first test of a memory kit featuring the new Hynix M-die chips, and the least we can say is that it packs a mighty punch. The XMP profile at 8200 MT/s is already ultra high-performance, so when you realize that 8400 MT/s is ultra-stable, what more could you ask for? The overclocking potential is also incredible, but only with a top-of-the-range motherboard and a processor with a very good IMC. Once again, thanks to Teamgroup, and especially to Kevin Wu, for giving us the opportunity to be among the first to enjoy the performance of this T-Force Xtreem 8200 MT/s CL38 kit.