• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.
  • The forums have been upgraded with support for dark mode. By default it will follow the setting on your system/browser. You may override it by scrolling to the end of the page and clicking the gears icon.

Team Group T-Force CKD XTREEM DDR5-8800 48 GB CL42

ir_cow

Staff member
Joined
Sep 4, 2008
Messages
5,106 (0.83/day)
Location
USA
Team Group is pushing boundaries with its XTREEM series, leveraging CUDIMM technology to deliver exceptional overclocking performance. This particular kit boasts a 2x 24 GB configuration, carefully binned to achieve impressive speeds of up to 8800 MT/s, with an additional 9466 MT/s profile. Join us as we put this memory through its paces and see how it compares to the competition in our comprehensive review.

Show full review
 
I've been kicking around this kit since it launched, but keep telling myself to wait for it to hit $300 or less (it's down to $320 this week!). It's funny how even minimal gains can be a draw when you want something new (and a better DRAM heatsink in my case).

Interesting to find out they include a much higher speed secondary profile (I wonder if the 8400 CUDIMM kit has something as well), but that does kind of explain the cap at 8800 when others have gone over 9000. It seems like a smart decision from a set the profile and go behavior as well due to Gear 2 compatibility issues over 9000.

One note on the review: while not super important a shot of mem tweakit with stock settings would be nice.

I'd be curious to see what single rank versus dual rank looks like on ARL at the same speed/timings. I don't think any has hit the market yet, but G.Skill has advertised high speed kits as well as low latency kits which are dual rank. I doubt the additional cost would be worth it unless one needed the capacity, but it'd be interesting to see how it plays out.
 
Good review. Kinda not worth it. DDR5 6400 seems to be the sweet spot and holding steady.
 
Last edited:
Does CKD DDR5 like this has some sense with Z790 board and 13900K or is it a total useless purchase?
 
Does CKD DDR5 like this has some sense with Z790 board and 13900K or is it a total useless purchase?
They might have better memory IC bins than lower clocked non CUDIMMs, but there is no advantage from the CKD. I doubt much of anything useful could be gained from using them with RPL and especially if looking to run as a daily driver.
 
ir_cow said:
On MSI motherboards, Intel 200S Boost—which changes the Die-to-Die (D2D) and Next Generation Uncore (NGU) ratios to 32X—is strictly limited to DDR5-8000. Any XMP profile using higher speeds than this will automatically be set to DDR5-8000, using the XMP Profile 1 primary timings. We also found that the maximum D2D Ratio without increasing the VNNAON voltage changes at increased memory speeds. For example, DDR5-8800 required 0.85 V (VNNAON) for 32X ratio, whereas DDR5-8000 did not need an increase from the stock value of 0.772 V.

Page 11. That sentence sounds incomplete, because of the word "that".
I'd be curious to see what single rank versus dual rank looks like on ARL at the same speed/timings. I don't think any has hit the market yet, but G.Skill has advertised high speed kits as well as low latency kits which are dual rank. I doubt the additional cost would be worth it unless one needed the capacity, but it'd be interesting to see how it plays out.
+1.

Also a versus between a "Manual profile" (using D2D, Ring Cache and also NGU if needed) with non-extreme, daily use settings/values and the official 200S Boost.

@thestryker6

There is a direct comparison here between 1R and 2R with 13900K using F5-6400J3239G16GX2-TZ5K and F5-6400J3239G32GX2-TZ5RW.
With Arrow to the knee Lake the only (indirect) comparison is between this and this.
But there are differences regarding the test systems so the results can't be compared directly. Still we can kind of deduce what the deal is.
Does CKD DDR5 like this has some sense with Z790 board and 13900K or is it a total useless purchase?
If I'm not mistaken extreme speeds like this require a Z790 with 2 DIMM slots.
 
Does CKD DDR5 like this has some sense with Z790 board and 13900K or is it a total useless purchase?
The Asus Z790 Apex Encore has some quirks (on a specific BIOS) that make some CKD kits work with bypass mode..not sure about this specific kit, the VColor Manta does work though (Bypass mode, so it will just work like an ordinary DDR5)

Good review. Kinda no worth it. DDR5 6400 seems to be the sweet spot and holding steady.
Depends on what you're doing and how fine tuned your system is..like I mentioned on the Ryzen Zen Garden thread, between my fully tuned i9 and my Ryzen 9 9950X3D there was no significant difference that would justify the expense..(even on UCLK=MCLK scenario on 4k I was losing more FPS vs the 2:1)
 
@Sol_Badguy I fixed that sentence. Also the highest retail dual-rank kit I have is GSKILL F5-6400J3239G32G . Since it's A-Die, I assume I could get it to 8000 after a bit of work on Arrow Lake, but the highest retail 64GB kit I could find was DDR5-7000. I was planning on adding the 6400 dual-rank in the charts. Now i'm thinking with 6400 JEDEC, 1R and 2R, it will confuse readers.

The Asus Z790 Apex Encore has some quirks (on a specific BIOS) that make some CKD kits work with bypass mode..not sure about this specific kit, the VColor Manta does work though (Bypass mode, so it will just work like an ordinary DDR5)
The point of bypass mode is avoiding these problems. Since CKD requires a command to activate, in theory all DDR5 MBs are compatible to boot. However the CKD function will not work. My personal experience with this bypass mode working is iffy. The MB at least needs to be updated enough to recognize 24GB DIMMs. My selection of motherboards is small now that I no longer review them... I just tell people to avoid using on legacy platforms unless they can do returns easily.

Does CKD DDR5 like this has some sense with Z790 board and 13900K or is it a total useless purchase?
Technically it should have a better binned IC, but it will be up to you to manually set the speed / timings you want. Also "write to SPD function" bricks the CKD currently, so no saving custom profiles if you want to resell. I would not spend this much when there is plenty of 7200, 7600 and 8000 kits that will do a equally good job for a lot less.
 
in theory
not all and not at every bios revision, for the Apex Encore it has to be on a specific bios modded from HWBOT to boot up..
 
not all and not at every bios revision, for the Apex Encore it has to be on a specific bios modded from HWBOT to boot up..
Not true. I'm using retail BIOS for that MB. Lots of conflicting reports. Unless you are saying CKD mode works on the modded BIOS, that could very well be true because I saw it at CES 2023.
 
Last edited:
@Sol_Badguy I fixed that sentence. Also the highest retail dual-rank kit I have is GSKILL F5-6400J3239G32G . Since it's A-Die, I assume I could get it to 8000 after a bit of work on Arrow Lake, but the highest retail 64GB kit I could find was DDR5-7000. I was planning on adding the 6400 dual-rank in the charts. Now i'm thinking with 6400 JEDEC, 1R and 2R, it will confuse readers.
Yeah unless you're explaining why it's included some people might not notice and some people might get confused thinking it's a chart error or something.
It probably warrants it's own short article, like an ARL memory case study / deep dive etc. where you investigate this aspect.
Once you draw the conclusion then you don't have to use both kits in further reviews/comparisons.

I realize this means more work for you but at the end of the day it's also added value, and you probably only need to do it once, well maybe twice if you also check CUDIMM hyper speeds (assuming it's stable).
You could probably do 3 speed steps: 6400, 7200, 8000 to cover the usual non-CUDIMM speed range, assuming the 2R kits cooperate at high speeds.

One other wish would be the inclusion of more non-gaming tests. In the original preview there are two extra tests compared to your current methodology: GPT2 inference and WinRAR.
More work I know but also more value especially since these are practical workloads.

Thank you again. :)
 
Yeah unless you're explaining why it's included some people might not notice and some people might get confused thinking it's a chart error or something.
It probably warrants it's own short article, like an ARL memory case study / deep dive etc. where you investigate this aspect.
Once you draw the conclusion then you don't have to use both kits in further reviews/comparisons.
I would be a good article. The work involved wouldn't bother me. I am time limited. This is my hobby side job after-all. I still do a normal 40hr+ day job. Only so many hours in day :/

One other wish would be the inclusion of more non-gaming tests. In the original preview there are two extra tests compared to your current methodology: GPT2 inference and WinRAR.
More work I know but also more value especially since these are practical workloads.
I have UL Proycon now installed. Unsure how useful it is for real-world comparisons yet. Just like 3DMark, half these benchmarks are just a number that is meaningless besides concluding score "lower is better" or "higher is better". Also unsure the logistics of getting W1z benchmark suite up and running on my test systems. We don't share a office.
 
Depends on what you're doing and how fine tuned your system is..like I mentioned on the Ryzen Zen Garden thread, between my fully tuned i9 and my Ryzen 9 9950X3D there was no significant difference that would justify the expense..(even on UCLK=MCLK scenario on 4k I was losing more FPS vs the 2:1)
While I don't want to seem like I'm calling you a liar, I am going to say that I greatly doubt game frame-rate issues are being caused by your system RAM running at 6400 VS 8800. Whatever is going on is somewhere else. The only real benefit to this kind of RAM is in things that are system RAM access intensive, such as scientific compute tasks. Gaming does not fall under that umbrella, regardless of the title and the results of this review prove that.
 
I have Z790 Aorus Master and currently sitting at 6200 36-36-36 (sweet spot for my Samsung ICs). I actually am very happy, never had a single crash in years but I would like to upgrade memory sooner or later. CKD seems quite a bit an evolution (even if actually it's just an added chip on the memory and ICs are the identical SK Hynix you can found anywhere, maybe just a better binning) but seeing there could be compatibility issues (my MB should definetely support 24GB modules but CKD won't be used, at max bypassed) I will wait for these:
- https://www.kedglobal.com/korean-chipmakers/newsView/ked202408290009
- https://www.mouser.it/new/micron-technology/micron-1-gamma-dram

There should bring better frequencies, less voltage and without CKDs. I think they will be compatible.
 
Last edited:
Whatever is going on is somewhere else. The only real benefit to this kind of RAM is in things that are system RAM access intensive, such as scientific compute tasks. Gaming does not fall under that umbrella, regardless of the title and the results of this review prove that.
Well buddy, call me whatever you like (I am very sorry for that if you feel/think I am a failure on that part), on a "Fully" tuned system (OS included) you can feel every bit of Mhz adding up, yes, I know you'll probably post some tables/graphs from the test from TPU with "their" test results, but my results are different since every single value on the RAM timings are tuned, at the end of the day, we as users are also different in ways being sensitive to such changes, but I am way past being just "normal or ordinary" I can almost feel the difference between using an ordinary NVME vs an Optane SSD for system drive use, even using a better binned chip vs turds can make the experience better if you know what you are doing.

for us Enthusiasts, speed matters..

Unless you are saying CKD mode works on the modded BIOS
CKD on bypass mode, not pure CKD, some won't even boot up, it will work only on bypass mode, benefits? you just get a good binned ram kit, normal users? pain in the ass to try and test..on retail bios its mostly a miss for some who already tested..as far as I know, only the Asrock Z790i Lightning boots up with CKD kits in bypass mode perfectly on its retail BIOS
 
Well buddy, call me whatever you like (I am very sorry for that if you feel/think I am a failure on that part)
Buddy? LOL! Um, ok.
on a "Fully" tuned system (OS included) you can feel every bit of Mhz adding up
As someone who works with and on memory intensive compute tasks frequently, I know exactly what I'm talking about.
I know you'll probably post some tables/graphs from the test from TPU with "their" test results
Nope. I don't need to, they speak clearly for themselves. Unless you can show examples of what you claim...
for us Enthusiasts, speed matters..
It can, but ONLY if the tasks you are engaging in can utilize the extra performance effectively. Games and desktop computing as a general rule do not.

It's called diminishing returns. Will this memory at 8800MT work better than 6400MT? Sure. Will that difference matter? That depends. Will it matter to gaming? Hell no.
 
I have Z790 Aorus Master and currently sitting at 6200 36-36-36 (sweet spot for my Samsung ICs). I actually am very happy, never had a single crash in years but I would like to upgrade memory sooner or later. CKD seems quite a bit an evolution (even if actually it's just an added chip on the memory and ICs are the identical SK Hynix you can found anywhere, maybe just a better binning) but seeing there could be compatibility issues (my MB should definetely support 24GB modules but CKD won't be used, at max bypassed) I will wait for these:
- https://www.kedglobal.com/korean-chipmakers/newsView/ked202408290009
- https://www.mouser.it/new/micron-technology/micron-1-gamma-dram

There should bring better frequencies, less voltage and without CKDs. I think they will be compatible.
Unless you're looking to spend time tweaking memory a kit of 7200 CL34 is probably the best choice (I don't really think there are going to be better kits put out at this speed). It should just work on most RPL parts and should be a solid improvement over what you've got in both bandwidth and latency.
 
Will it matter to gaming? Hell no
well hell yeah, for me..I wouldn't be using it if there wasn't any benefit to it..on AMD platform? probably but for Intel (specially the Raptor Lake) it does make sense. I don't really have that much of a liberty at time to prove somebody wrong/right, I will just keep things to myself. (at least I tried is what matters, not trying to please anyone as its really not my nature to fish for recognition of any sort)

Letz juzt agree to dizagree.. :roll:
 
well hell yeah, for me..I wouldn't be using it if there wasn't any benefit to it..on AMD platform? probably but for Intel (specially the Raptor Lake) it does make sense. I don't really have that much of a liberty at time to prove somebody wrong/right, I will just keep things to myself. (at least I tried is what matters, not trying to please anyone as its really not my nature to fish for recognition of any sort)
Hey, you do you. If you want to spend 3X the money for an extra 1% or 2%, feel free. You're never going to notice/see those extra 3 or 4 frames in gameplay, but at least you'll know they're there. Oh, wait, that's right, you won't see them, because the 8800 speeds are actually slower than the 6400 speeds..
TADA! Funny that, eh?
(Now let's all wait for the cherry-picked comeback pointing out the one game tested, CyberPunk2077, where the 8800 was 2 frames faster, even though all the rest showed the 6400 being faster)
Lets just agree to disagree.. :roll:
Let's also learn how to spell. Or were you just trying to feel all the cool? And you can agree to be all the incorrect you want to be..
 
Last edited:
Oh, wait, that's right, you won't see them, because the 8800 speeds are actually slower than the 6400 speeds..
You're comparing a 2x 16GB kit to a 2x 24GB kit where the subtimings are lower on the former. The only other 2x 24GB kit in the charts is the DDR5-8000 which runs slower than the 8800. If they were all 2x 24GB kits then they'd likely be in order of absolute latency and/or bandwidth.
 
You're comparing a 2x 16GB kit to a 2x 24GB kit where the subtimings are lower on the former. The only other 2x 24GB kit in the charts is the DDR5-8000 which runs slower than the 8800. If they were all 2x 24GB kits then they'd likely be in order of absolute latency and/or bandwidth.
Oh, so different cherry-picking then?
 
Because it makes no difference. Your argument is a red-herring.
What are you talking about? I know at this point I probably shouldn't be engaging with your utter nonsense but I genuinely don't understand your problem. It's a fact that sub timings on 24GB modules are higher than 16GB and this impacts performance.

In this review an entire page was dedicated to explaining it: https://www.techpowerup.com/review/team-group-t-force-xtreem-ddr5-7200-48gb-cl34/20.html
 
What are you talking about?
Go read the review. THAT'S what I'm talking about.
I know at this point I probably shouldn't be engaging with your utter nonsense
Oh?
but I genuinely don't understand your problem.
That is because you not reading the information contained in the above review and understanding it.
It's a fact that sub timings on 24GB modules are higher than 16GB and this impacts performance.
But by how much? 1%? Maybe 2%? Does 1 or 2 percent justify twice the cost?
And what conclusion does that offer? The CS2 result in that very page you cited show 9fps difference(out of 624fps) for the average and an 8FPS difference(out of 348fps) for the 1% lows.

So is that a big enough improvement for twice the money? I think not! That is my point.

This kit at $180(MSRP) is NOT worth the price over a 6400 kit for $100 or 6000 for $90. You can offer your elitist argument all you wish, but at the end of the day, this kit is NOT a good value and NOT worth the increased price. End of discussion.
 
Back
Top