vacancies advertise contact news tip The Vault
facebook rss twitter

AMD Ryzen CPU pushed beyond 5.2GHz on all 8 cores (video)

by Mark Tyson on 24 February 2017, 11:01

Tags: AMD (NYSE:AMD)

Quick Link: HEXUS.net/qadenk

Add to My Vault: x

AMD's new Ryzen CPUs are breaking records even before they reach general availability. Pro overclockers came to the Ryzen announcement party to see how far they could push the new Ryzen 7 CPUs and the results aren't disappointing - as long as you are happy with world record benchmark scores.

YouTube videos of the feat have been whack-a-moled but luckily I found a Vimeo of the Ryzen OC segment, which I've embedded above.

Lashings of LN2 were ladled on the 1800X

From an earlier screen grab taken by TweakTown, below, you can see the OC team had their Ryzen 7 1800X system running at 5.2GHz with all 8 cores active. That same screenshot shows a score of 2363 in Cinebench, which must be from earlier in the event. If you watch through the video you are briefly shown the moment when the AMD drafted overclocking team break world records in Cinebench, with a score of 2449cb. The record broken was a score of 2410cb.

As usual, to push clocks to the limit, the OC team used LN2 cooling and rather high voltages. The CPU-Z screenshot shows a core voltage of 1.875V was applied to achieve a 5.2GHz clock. I wonder how fast the Ryzen 7 1800X could have gone if cooled with LHe… This 95W processor's stock speeds are 3.6GHz to 4GHz but sometimes faster thanks to AMD's XFR technology.

Ryzen die shot from Dr Lisa Su's slide deck, click to zoom.



HEXUS Forums :: 11 Comments

Login with Forum Account

Don't have an account? Register today!
LN2 and 1.875V to get to 5.2GHz?

Hmm… having a quick look at HWBot the record frequency for the (also 8C/16T) 6900k is actually 5.22GHz, so they're there or thereabouts for the higher core counts … just wonder how that'll translate to overclocks for the lower core counts if they're based on the same die….
Impressive! my 8350 and 1070 are doing what I need my PC to do right now, but I really am getting more and more tempted to upgrade early just to show some support for these chips, especially given the price, maybe I will wait for a nice motherboard cpu and ram bundle to go online somewhere (considering I will need it all anyway) and use that as an excuse to wait a bit longer haha.
Seems pretty good. Hope there is some good mileage to be had in overclocking on water and with the Bus Speed as most of the recent CPUs (from Intel anyway) pretty much limited you to multiplier overclocks with only minimal increases to bus speed possible
scaryjim
LN2 and 1.875V to get to 5.2GHz?

Hmm… having a quick look at HWBot the record frequency for the (also 8C/16T) 6900k is actually 5.22GHz, so they're there or thereabouts for the higher core counts … just wonder how that'll translate to overclocks for the lower core counts if they're based on the same die….

That is the golden question, and possibly why the lower core counts are coming out later so the process is a little more mature (and inventory of binned parts increases).

What's interesting to me is how AMD release at closer to max limits in terms of frequency - Intel might just be playing safer, but it could well be AMD are getting a little more out of the chips with the frequency/power tuning magic.
kalniel
… it could well be AMD are getting a little more out of the chips with the frequency/power tuning magic.

If this is the same 14nm process that Polaris is fabbed on, that may have a bearing too:


from: https://forums.anandtech.com/threads/new-zen-microarchitecture-details.2465645/page-93#post-38366723

If the voltage curve for Zen is even vaguely similar that would explain them being able to release close to the ragged edge - a long flat voltage plateau, then a steep upcurve. I can even substitute known Zen base clocks on that graph - if the R7 1700's 3GHz base clock equates to the 900MHz Polaris clock at the end of the voltage plateau, the R1800X's 3.6GHz base clock would equate to roughly 1100MHz, requiring around 20% more voltage. Rather pleasingly, since power is proportional to the square of voltage, that would mean the R7 1800X should require around 45% more power than the 1700, and it's TDP is around 45% higher ;)

That's all a little bit too clean for my liking, but I think it probably points in the right direction. That may well mean that 5GHz is about as far as you're likely to get any Zen processor, although it may be that the thermal inertia of a large chip with only half the cores active will mean it's acheivable with air or water on a 4C chip, rather than LN2…