AMD Ryzen Threadripper 2990WX Overclocked to 5.955 GHz On All Cores

[Raevenlord]

Hardwarebot has graced us with the confirmed feat of an AMD Ryzen Threadripper 2990WX reaching almost 6 GHz across all of its cores. The overclocking effort, supported by an extreme overclocking setup with liquid nitrogen - and as such, worth more for the sheer PR and "fun tech factor" associated with it. Overall, the overclock up to 5.955 GHz on all cores amounts to a practically 100% increased clockspeed over its base, all-core 3 GHz base clocks.

The overclocking effort was achieved bu user IvanCupa on MSI's X399 MEG Creation motherboard paired with 8 GB of single-channel DDR4 memory (easier on the CPU to support than if more channels were populated). We'll see whether this chip can be taken even closer to the stratosphere than this - remember it's only been a few days since the chip's release to the wild. "Tinkerers be tinkering still."



View at TechPowerUp Main Site

 

[[XC] Oj101]

Wow, it's been about a decade since I've heard "Hardwarebot"

 

[jsfitz54]

Windows 10 Pro 16299? Do we all need to roll back to this version?

 

[dj-electric]

Important note - this is CPU validation only, not a performance or stability test, short or long, in any way.

 

[bug]

Somebody let @Liviu Cojocaru know we need his "this doesn't mean anything to me" comments in here.

 

[phill]

Impressive result Wonder if they could get a 6Ghz run in CB 15 or wPrime or something??

 

[GreiverBlade]

[sarcasm]oh ... not a "ultra custom made for it mobo"? like the one from the 5ghz Intel? ... [/sarcasm](i know it's a MSI X399 MEG Creation no worries ... )

albeit being well ... not useful for a daily driver ... it's still impressive to achieve 5.955ghz on 32 Cores (and 64 Threads )

 

[BluesFanUK]

What's the point?

 

[leppie]

This is just 1 core...

 

[Ruru]

This is just 1 core...

Still all cores and SMT enabled. Usually on Intel platform these high OC's are with few cores and HT turned off.

 

[Jism]

What's the point?

Gives some indication to what the TR is capable of on LN2. But nothing more. To get it actually stable and bench solid you are looking at 5Ghz or less. It proberly requires a shitload of power on almost 6GHz.

 

[Frick]

Still all cores and SMT enabled. Usually on Intel platform these high OC's are with few cores and HT turned off.

Enabled, but running 600Mhz.

 

[R-T-B]

Enabled, but running 600Mhz.

Yeah, the article kind of glances over that detail...

 

[HTC]

Impressive result Wonder if they could get a 6Ghz run in CB 15 or wPrime or something??

Not 6GHz but 5.387GHz: for 32 cores, that's very impressive, IMO.

 

[Patriot]

Not 6GHz but 5.387GHz: for 32 cores, that's very impressive, IMO.

It also stomps the 5ghz 28c xeon score... as you might expect. Frankly I was expecting the memory starved dies to hold it back more than it seems to.
UMA seems to be the right approach.
Intel @5GHz 28c 7,334
AMD @5.387GHz 32c 8,532

 

[RealNeil]

"Tinkerers be tinkering still."

Whoo-Ya! I love it!

 

[HTC]

It also stomps the 5ghz 28c xeon score... as you might expect. Frankly I was expecting the memory starved dies to hold it back more than it seems to.
UMA seems to be the right approach.
Intel @5GHz 28c 7,334
AMD @5.387GHz 32c 8,532

That depends on the application: some are affected more then others, and the penalty is quite severe, in more extreme cases.

Start watching from 17:04 minute mark:

 

[Hood]

This is just 1 core...

Also only single channel RAM...

 

[HTC]

Also only single channel RAM...

For CPU-Z record screenshots, that's actually normal.

 

[Lucky_n00b]

Hi there, just wanted to make a small clarification here just in case this info is overlooked

I'm an overclocker working for small Indonesian-only OC site (oc.jagatreview.com), also sometimes I cooperate with vendors and other partners doing some overclocking tests for an upcoming pre-released products.
Both me and my friend IvanCupa do a joint LN2 session for 2990WX Extreme OC testing on the MSI X399 MEG to show some extreme OC result on the NDA Release day, I'm doing benchmarks (Cinebench R15/GPUPI for CPU 1B) and IvanCupa is doing CPU-Z clock validation.

For the CPU-Z Validation, please do note that:
- Yes, ALL 32-Core is including SMT is still enabled when booting from BIOS, but
- We searched or Best Cores using Ryzen Master, and downclocked all the other cores while doing so. So effectively it's only one core that clocked that high.
- We show this on the screenshot submitted on hwbot (core number 26 only overclocked, others downclocked)

It turned out that there's a difference of at least 150Mhz between 'best' cores and 'worst' cores for validation, and this is the best cores out of our 2990WX.
Due to LN2 and time constraint, we haven't tested doing LN2 validation for ALL-core turned to EXACT same speed. Would love to do that later though

Anyway, correct voltage was the one on MSI Command Center Lite. 1.68v. Temperature was Full pot (-184C or so)

Hopefully that clears things up

*)Sorry if bad English

Best Regards,


Alva

 

[eidairaman1]

It also stomps the 5ghz 28c xeon score... as you might expect. Frankly I was expecting the memory starved dies to hold it back more than it seems to.
UMA seems to be the right approach.
Intel @5GHz 28c 7,334
AMD @5.387GHz 32c 8,532

Pretty good considering bulk of programs are using an intel written compiler anyway.

 

[HammerON]

Hi there, just wanted to make a small clarification here just in case this info is overlooked

I'm an overclocker working for small Indonesian-only OC site (oc.jagatreview.com), also sometimes I cooperate with vendors and other partners doing some overclocking tests for an upcoming pre-released products.
Both me and my friend IvanCupa do a joint LN2 session for 2990WX Extreme OC testing on the MSI X399 MEG to show some extreme OC result on the NDA Release day, I'm doing benchmarks (Cinebench R15/GPUPI for CPU 1B) and IvanCupa is doing CPU-Z clock validation.

For the CPU-Z Validation, please do note that:
- Yes, ALL 32-Core is including SMT is still enabled when booting from BIOS, but
- We searched or Best Cores using Ryzen Master, and downclocked all the other cores while doing so. So effectively it's only one core that clocked that high.
- We show this on the screenshot submitted on hwbot (core number 26 only overclocked, others downclocked)

It turned out that there's a difference of at least 150Mhz between 'best' cores and 'worst' cores for validation, and this is the best cores out of our 2990WX.
Due to LN2 and time constraint, we haven't tested doing LN2 validation for ALL-core turned to EXACT same speed. Would love to do that later though

Anyway, correct voltage was the one on MSI Command Center Lite. 1.68v. Temperature was Full pot (-184C or so)

Hopefully that clears things up

*)Sorry if bad English

Best Regards,


Alva

Thanks for the post and welcome to TPU!

 

[Gasaraki]

What a CLICKBAIT headline. It's 5.9 on ONE CORE, not all cores.

 

[Patriot]

That depends on the application: some are affected more then others, and the penalty is quite severe, in more extreme cases.

Start watching from 17:04 minute mark:

I am aware, that said, Cinebench is a memory sensitive render or at least it was back when I set the 11.5 WR's
I held out against much higher clocked boxes due to super low ram latency. Used Cas 5 ddr3 1600 on 48c Magny Cours for 43.5 @ 3.8ghz.

Also noting... Almost all of these cinebench runs violate the rules by using win10 ...

 

[bug]

What a CLICKBAIT headline. It's 5.9 on ONE CORE, not all cores.

Neah, it's just a slight omission. It's 5.9GHz on all overclocked cores