TeamGroup Xtreem DDR4-4000 CL15 2x 16 GB Engineering Sample

[ir_cow]

TeamGroup has sent over a special media-only pre-binned T-Force Xtreem ARGB DDR4-4000 kit that is ready to be put through the paces! This 32 GB memory kit is dual-rank (2x16 GB) with amazing XMP timings of 15-15-15-35. We dive into the question of whether top-tier memory really grants more frames or it is just placebo.

Show full review

 

[Selaya]

It would be interesting to see how low the timings can go using a waterblock and raising the voltage to 1.8 V+. If this is something of interest to you, please leave your thoughts in the comments.

yes plsplspls

FOR SCIENCE!

 

[Athlonite]

Yeaqh tried for a month to get my GSkil Ripjaws 4000MT/s to run at those speeds and had to end up running them at 1:1:1 @ 1933MHz 1.35V

 

[ThrashZone]

Hi,
Yeah 1.5v for c15
Think I perfer my royals 4000c16 at 1.4v.

Did they add temp sensors on these or are you using a temp gun ?

 

[zlobby]

Now these are timings and MT/s I can work with!

 

[dgianstefani]

There's a Gskill 2x16GB 4000/14 kit that theoretically should be better.

 

[redhat_ownage]

what is this review? the only speeds you tested on this "binned god overclocking ram" was xmp and slower than xmp, how did this even get approved?
you used probably the second worst ddr4 platform possible, second only to x99, and on the AMD side you didn't even make it to 4000c15 because you you didn't turn off gear down mode, if team never sends you another kit of memory to test the world would not be sad about that.
worst review of a b-die kit I've ever seen.

 

[buildzoid]

you don't need watercooling to shove 1.8V into B-die. Just put a fan on it. The bigger issue you'll run into is the fact that allowing the OS to access and entire B-die stick when over 1.8V causes instability. I have a few videos running B-die at 4000+ CL12 and you need to set windows to only access 2GB per 8GB of B-die. So with a 32GB kit you get 8GB usable at 2V. Most kits start having these stability issues above 1.75V.

 

[ir_cow]

what is this review? the only speeds you tested on this "binned god overclocking ram" was xmp and slower than xmp, how did this even get approved?
you used probably the second worst ddr4 platform possible, second only to x99, and on the AMD side you didn't even make it to 4000c15 because you you didn't turn off gear down mode, if team never sends you another kit of memory to test the world would not be sad about that.
worst review of a b-die kit I've ever seen.

Not a review. Did you read it at all?

 

[Toothless]

what is this review? the only speeds you tested on this "binned god overclocking ram" was xmp and slower than xmp, how did this even get approved?
you used probably the second worst ddr4 platform possible, second only to x99, and on the AMD side you didn't even make it to 4000c15 because you you didn't turn off gear down mode, if team never sends you another kit of memory to test the world would not be sad about that.
worst review of a b-die kit I've ever seen.

Did you seriously made an account to complain about a showing of engineer samples? Read the article, constructive criticism or don't be an arse and not comment.

 

[The King]

From what I seen some Ryzen systems see a lose in overall performance @ 4000/2000 compared to 3800/1900.

Would have loved to have seen 3800 CL15/16 performance on the AMD system. Still thanks for this and posting the Thaiphoon screenshot.

Zentimings on the Ryzen system will be great as well.

 

[Icon Charlie]

Thank you for the write up.

A suggestion if you can find the Ram now that is.

I've been Rocking with the OLOY Blade Series of memory PC 4000 CL 15-16-16-36, which I have or their newer CL 14-15-15-35.

I down volt mine as stated previously on this forum to PC 3600 14-14-14-34

They used the Samsung B die which is why I went after them when they first came online.
No issues and its been 9 months or so of my usual usage.

 

[Mussels]

Okay, who do i have to sleep with to get RAM like this

Oh god.... so all the graphs on page 3 have Trident spelled as Tirdent
heh. Turd.


Edit:
I need to work on my memory timings. 64GB is great and all, but that best to worst difference is HUGE

 

[PapaTaipei]

Receives supposedly good RAM for enthusiasts, proceed to use it at low fps and in games/resolutions where it doesn't make a difference. GG!
Next time try 720p in CSGO and F1 and Quake Champions.

 

[ir_cow]

Okay, who do i have to sleep with to get RAM like this

Oh god.... so all the graphs on page 3 have Trident spelled as Tirdent
heh. Turd.


Edit:
I need to work on my memory timings. 64GB is great and all, but that best to worst difference is HUGE

View attachment 252182

Thanks for the catch. I was changing stuff and well...you see what happened. Team Group is apparently TeamGroup. If you go to the website, it's spelled three different ways... I picked the wrong one

*edit: updated

From what I seen some Ryzen systems see a lose in overall performance @ 4000/2000 compared to 3800/1900.

Would have loved to have seen 3800 CL15/16 performance on the AMD system. Still thanks for this and posting the Thaiphoon screenshot.

You got it!

 

[Mussels]

The summary for Zen 3 still seems to be:
Four ranks of ram, 3600+ (1:1 with IF) and the lowest CL you can afford

It's just seeing ~20% swings from simple things like two ranks vs four is enough to swing so many reviews one way or another

 

[The King]

You got it!

I don't see 3800 CL15/16 in the AMD chart, only the Intel chart has 3800 CL15?

With my current ZEN 3 setup am I running 3800 CL15 rather 4000 CL16 because performance was higher at 3800.
At 4000/2000 WHEA errors can have a negative impact on overall system peformance.

The summary for Zen 3 still seems to be:
Four ranks of ram, 3600+ (1:1 with IF) and the lowest CL you can afford

It's just seeing ~20% swings from simple things like two ranks vs four is enough to swing so many reviews one way or another

This is why I dont read/like any review on TH. They test all the Ryzen systems 8GBx2 and say running them at 3200 CL14 is good enough.

I would argue for the price of a 3200 CL14 8GBx2 kit you could get a 2X16GB kit with looser CAS like C16/18 and still see the benefits of running DR rank over SR.

 

[sector-z]

you don't need watercooling to shove 1.8V into B-die. Just put a fan on it. The bigger issue you'll run into is the fact that allowing the OS to access and entire B-die stick when over 1.8V causes instability. I have a few videos running B-die at 4000+ CL12 and you need to set windows to only access 2GB per 8GB of B-die. So with a 32GB kit you get 8GB usable at 2V. Most kits start having these stability issues above 1.75V.

Oh nice and 64GB equal 16GB possible ? I run 4X16GB 3800mhz 1:1 14-14-14-32 already at 1.56V

 

[DrCR]

you don't need watercooling to shove 1.8V into B-die. Just put a fan on it. The bigger issue you'll run into is the fact that allowing the OS to access and entire B-die stick when over 1.8V causes instability. I have a few videos running B-die at 4000+ CL12 and you need to set windows to only access 2GB per 8GB of B-die. So with a 32GB kit you get 8GB usable at 2V. Most kits start having these stability issues above 1.75V.

This is both awesome while also self revealing for me. If still a teen I’d be all over this but now it sounds like work rather than fun geekery. Like a geek’s version of mountaineering or some such.

 

[Mussels]

you don't need watercooling to shove 1.8V into B-die. Just put a fan on it. The bigger issue you'll run into is the fact that allowing the OS to access and entire B-die stick when over 1.8V causes instability. I have a few videos running B-die at 4000+ CL12 and you need to set windows to only access 2GB per 8GB of B-die. So with a 32GB kit you get 8GB usable at 2V. Most kits start having these stability issues above 1.75V.

My RAM has run off crying. I want to join it.

You do scary things with ram.

 

[Chrispy_]

The cost of these is likely to be eye-watering. Fancy box, ARGB, B-die.

The thing is, I'm running dirt-cheap 2x16GB Kingston 3600 CL16 kit that is within 20% the cost of the cheapest, low-end DDR4-3000 RAM on the market. I tweaked it lightly for fun and managed to run at 3733 CL14 but honestly saw improvements that could best be described as "run to run variance" so I daily it at the rated 3600 CL16.

Yes, this DDR4-4000 is faster, but by such an insignificantly pointless margin that spending more for anything other than chasing world records is utter madness. If you just want pretty ARGB lights, you can do that for practically no additional cost at the budget/entry-level end of the market.

 

[Blaeza]

When you lot have finished messing about with my ram, send it over please. Thanks

My Corsair scrap ram is currently running at 3800mhz, but CL22. Is that alright or should I start a new thread on ram ocing?

 

[Chrispy_]

From what I seen some Ryzen systems see a lose in overall performance @ 4000/2000 compared to 3800/1900.

Would have loved to have seen 3800 CL15/16 performance on the AMD system. Still thanks for this and posting the Thaiphoon screenshot.

Zentimings on the Ryzen system will be great as well.

The key is the 1:1 FCLK:MCLK divider which on some boards kicks in automatically beyond 1866MHz, and even if you have the option to force it to 1:1 at 1933 or 2000 it isn't guaranteed that your CPU can handle this. Zen2/Zen3 both use the exact same IO die (with the IMC) and AMD (Robert Hallock) confirmed that 1800MHz was reachable on most samples for Zen2's launch and they added 100MHz to that estimate for Zen3's launch due to yield improvements.

Even if you buy one that was manufactured this month, there's a non-trivial chance that your shiny new Zen3 won't be stable with a 2000MHz FCLK for DDR4-4000.

you don't need watercooling to shove 1.8V into B-die. Just put a fan on it. The bigger issue you'll run into is the fact that allowing the OS to access and entire B-die stick when over 1.8V causes instability. I have a few videos running B-die at 4000+ CL12 and you need to set windows to only access 2GB per 8GB of B-die. So with a 32GB kit you get 8GB usable at 2V. Most kits start having these stability issues above 1.75V.

ROFL, throwing away 75% of the RAM capacity you purchased just to get timings lower?
That's genius - and for record chasing that's a cool fact I wasn't aware of.

 

[Mussels]

The key is the 1:1 FCLK:MCLK divider which on some boards kicks in automatically beyond 1866MHz, and even if you have the option to force it to 1:1 at 1933 or 2000 it isn't guaranteed that your CPU can handle this. Zen2/Zen3 both use the exact same IO die (with the IMC) and AMD (Robert Hallock) confirmed that 1800MHz was reachable on most samples for Zen2's launch and they added 100MHz to that estimate for Zen3's launch due to yield improvements.

Even if you buy one that was manufactured this month, there's a non-trivial chance that your shiny new Zen3 won't be stable with a 2000MHz FCLK for DDR4-4000.


ROFL, throwing away 75% of the RAM capacity you purchased just to get timings lower?
That's genius - and for record chasing that's a cool fact I wasn't aware of.

3800Mhz is when they stop being in sync, the auto setting wont go above 1800 for infinity fabric - anything beyond that, it's all manually set
I havent tested every board, but seen it on 5 or so (mostly asus) with 3600/1800 as the limit for auto

 

[Chrispy_]

3800Mhz is when they stop being in sync, the auto setting wont go above 1800 for infinity fabric - anything beyond that, it's all manually set

I don't tend to buy much ASUS. Their high-end stuff is great but overpriced and their entry-level stuff is typically worse than Gigabyte/MSI/Asrock.

For MSI and Aorus I usually see the auto divider doing what AMD claimed was standard behaviour at Zen2's launch: