6GHz across 32 cores, ThreadRipping mayhem

Subject: General Tech | August 16, 2018 - 02:28 PM |
Tagged: amd, threadripper 2, 2990wx, overclocking, LN2

The low cost workstation class 2990WX has been verified as running at 5.955GHz on an MSI MEG X399 Creation board, with the help of a lot of liquid nitrogen.  The Inquirer has links to the setup that Indonesian overclocker Ivan Cupa needed in order to manage this feat, which required fans to cool certain portions of the motherboard as well.  You are not likely to see this set up installed in a server room but the achievement is no less impressive as that is an incredible frequency to reach.  Check it out in all it's glory.

View Full Size

"So far, it would seem that AMD is on top when it comes to willy-waving, though it's worth noting that overclocked performance is a tad nebulous and real-world in-app performance is really where choosing an Intel or AMD chip comes to play."

Here is some more Tech News from around the web:

Tech Talk

 

Source: The Inquirer

August 16, 2018 | 03:24 PM - Posted by a-non (not verified)

It is more than misleading to call it 6GHz across 32 cores, only one core achieved the shown 5955.4 MHz.

August 16, 2018 | 03:46 PM - Posted by Anonymouskj (not verified)

you're absolutely right only one core reach 5.95 ghz

August 16, 2018 | 09:27 PM - Posted by Paul A. Mitchell (not verified)

https://d1ebmxcfh8bf9c.cloudfront.net/u43611/image_id_2052102.png

Core Speed (MHz): 5875 on Core #26

All other Cores: 600

See also: "Peak Speed 5.875 GHz"
in the row across the top:
"Temperature, Speed, Power & Current"

August 16, 2018 | 05:03 PM - Posted by MoReBenchesNeededOn10 (not verified)

Forget the LN2/OC and those Abject Hacks at The Inquirer, let's talk about all the Windows benchmarking on TR2/32 core that's a bit behind the same sorts of Benchmarks Done on some Linux Distros.

Now that's some interesting results and TR2 needs some re-benchmarking going forward on Windows once all the problems are sorted out. Windows 10's scheduler looks to be having issues with that Compute only cores paired with the Full SOC cores that have their own local Memory Channels available TR2 design. So that's more optimization for TR2 simply because of that NUMA/NLMA(No Local Memeory Access) DIEs parings that is unique to TR2(24 and 32 core) SKUs.

Also does anyone think that AMD may just have a 16 core TR2 with 16 cores that are spread across 4 dies on a TR2/SKU or will AMD just stick with that standard 2 dummy DIEs and 2 FULL SOC DIEs/4 memory channels for 16 all TR2 SKUs.

August 16, 2018 | 05:14 PM - Posted by MoReBenchesNeededOn10 (not verified)

Edit: for 16 all
to: for all 16 core

August 17, 2018 | 08:53 AM - Posted by a-non (not verified)

So, you guys want to correct and admit your mistake or be petty and give us an example of how much you value just churning out small stories over a simple source check that should take 5 minutes at worst in this example?

August 19, 2018 | 08:05 AM - Posted by Aymd (not verified)

So Intel needed a chiller to reach 5ghz on all 28 cores, AMD needed LN2 to reach 5,9ghz on one core, AYMD.

Post new comment

The content of this field is kept private and will not be shown publicly.
  • Lines and paragraphs break automatically.
  • Allowed HTML tags: <a> <em> <strong> <cite> <code> <ul> <ol> <li> <dl> <dt> <dd> <blockquote><p><br>
  • Web page addresses and e-mail addresses turn into links automatically.

More information about formatting options

CAPTCHA
This question is for testing whether you are a human visitor and to prevent automated spam submissions.