Ryzen "2" ?

Hotwired · 24 Apr 2018 at 11:18

DragonQ said:
That's pretty cool. Because the chips essentially overclock automatically and have barely any manual overclocking headroom, undervolting gives them more TDP breathing space for clocking themselves.

That's a devious solution and completely the opposite to what people expect to be doing: raising voltage to increase clocks. But the automatic system has different rules to follow.

Undervolting enough to allow higher frequency within the power limit but not so low that clocks become unstable and the cpu backs off again.

Cooling is of course self explanatory, lower the better.

4K8KW10 · 24 Apr 2018 at 11:20

Which one is better:
1. Ryzen 7 2700 3.2GHz (max 4.1GHz) with 65W TDP for €289 or
2. Ryzen 7 2700X 3.7GHz (max 4.35GHz) with 105W TDP for €315

?

orbitalwalsh · 24 Apr 2018 at 11:23

4K8KW10 said:
Which one is better:
1. Ryzen 7 2700 3.2GHz (max 4.1GHz) with 65W TDP for €289 or
2. Ryzen 7 2700X 3.7GHz (max 4.35GHz) with 105W TDP for €315

?

think of it this way, can 26 Euros be put to something else in the system or a game

SiDeards73 · 24 Apr 2018 at 11:32

4K8KW10 said:
Which one is better:
1. Ryzen 7 2700 3.2GHz (max 4.1GHz) with 65W TDP for €289 or
2. Ryzen 7 2700X 3.7GHz (max 4.35GHz) with 105W TDP for €315

?

I think with Ryzen Gen1, the non X chips were better as you could OC them to the X chips pretty much, i think with Ryzen Gen2 this is reversed, your better off it seems with the X Version, leaving them at stock and fiddling with other settings to allow PB and XFR etc to work their magic, so your still kinda overclocking, but it seems your balancing your system to allow the chip to boost as far as it can, see the Undervolting stuff above for instance.

4K8KW10 · 24 Apr 2018 at 12:00

SiDeards73 said:
I think with Ryzen Gen1, the non X chips were better as you could OC them to the X chips pretty much, i think with Ryzen Gen2 this is reversed, your better off it seems with the X Version, leaving them at stock and fiddling with other settings to allow PB and XFR etc to work their magic, so your still kinda overclocking, but it seems your balancing your system to allow the chip to boost as far as it can, see the Undervolting stuff above for instance.

Yes, absolutely will do the undervolting.
The thing is that this 40W power consumption difference is too tempting to me and I would like to take the greener CPU with significant energy savings.
I will not pursue maximum performance but maximum power savings with optimal performance.

orbitalwalsh said:
think of it this way, can 26 Euros be put to something else in the system or a game

Yes, definitely something can be bought - a mouse, a keyboard, headphones or a better motherboard or memory.

Thanks for the replies, guys

chroniclard · 24 Apr 2018 at 12:51

Up and running.

Stock settings, 4340 single core boost, seems to be about 4000 all core.

Will tweak later.

los · 24 Apr 2018 at 13:00

chroniclard said:
Up and running.

Stock settings, 4340 single core boost, seems to be about 4000 all core.

Will tweak later.

What motherboard are you using? What is your memory running at?

chroniclard · 24 Apr 2018 at 13:22

los said:
What motherboard are you using? What is your memory running at?

Only had the CPU an hour so not much testing yet.

Its a CH6 motherboard on the 6004 official bios, memory is corsair LPX running at 3466 DOCP settings. I have not tested memory fully yet for stability but I will be trying to tweak the timings down. I had it running at 3333 with my 1700X and much tighter timings.

Will have a further play tonight.

chroniclard · 24 Apr 2018 at 15:50

Also for those interested, 2700X does about 700h/s on cryptonight mining algo with stock settings.

chrcoluk · 24 Apr 2018 at 17:40

so the undervolting that I mentioned has caught on then

chroniclard · 24 Apr 2018 at 17:55

Chewie said:
Just tried bclk 101, to see what would happen, and it seems to work just fine. Slight performances increase overall and max boost frequency achieved is now 4392.9Mhz.

Note that this is still with the -0.05 Vcore offset applied.

I got pretty much exactly the same result, -0.05 and 101 bclk. Max 4392 on some cores.

chroniclard · 24 Apr 2018 at 18:23

Some interesting results here, 4497 single core boost. Unsure why voltage is locked.....

chroniclard · 24 Apr 2018 at 18:38

Tighter mem timings increased multi core but not single. :confused:

chrcoluk · 24 Apr 2018 at 18:49

My 8600k gets 207 single at 4.8ghz, so that 186 is pretty good going.

chroniclard · 24 Apr 2018 at 19:14

Dont trust the voltage reading in hwinfo as cpu-z and ryzen master are showing much higher. Going to normal clocks.

4K8KW10 · 24 Apr 2018 at 19:25

chrcoluk said:
My 8600k gets 207 single at 4.8ghz, so that 186 is pretty good going.

Can you imagine what this means?

It means that Pinnacle Ridge and Coffee Lake have almost equal IPC.
Actually, according to http://www.guru3d.com/articles_pages/amd_ryzen_7_2700x_review,9.html, the difference is 2.7%.

JJimmyJ · 24 Apr 2018 at 19:49

chroniclard said:
I got pretty much exactly the same result, -0.05 and 101 bclk. Max 4392 on some cores.

Not to be 'that guy' but I'd be very surprised if both of you had the exact same offset with the exact same results, unless offset works differently on amd boards? Surely you have to find a set voltage that's stable for desired overlook then work out offset based off auto voltage for that clock? This is just my experience of offset over locks. I even had to do a positive offset lower my voltage on my i7 2600k?

chroniclard · 24 Apr 2018 at 19:52

JJimmyJ said:
Not to be 'that guy' but I'd be very surprised if both of you had the exact same offset with the exact same results, unless offset works differently on amd boards? Surely you have to find a set voltage that's stable for desired overlook then work out offset based off auto voltage for that clock? This is just my experience of offset over locks. I even had to do a positive offset lower my voltage on my i7 2600k?

What I mean is I got very similar max boost/clocks with the same settings. Our resulting voltages wont be identical but the performance was near enough the same.

JJimmyJ · 24 Apr 2018 at 20:18

chroniclard said:
What I mean is I got very similar max boost/clocks with the same settings. Our resulting voltages wont be identical but the performance was near enough the same.

Ah ok.
Sorry, probably going to swamp you with questions now

Did you have to change any other settings? (vdroop, llc etc - if they're even relevant on AMD). What was the lowest voltage it dropped to?
I'm moving to ryzen (2600 or 2600x), and the offset clocking is the way I've always clocked. Never really see it in the wild as much nowadays

haszek · 24 Apr 2018 at 20:29

chroniclard said:
Dont trust the voltage reading in hwinfo as cpu-z and ryzen master are showing much higher. Going to normal clocks.

Hwinfo is apparently very accurate but you're looking at the wrong place, VDDCR_CPU SVI2 TFN is what you should check.