RyZen 3000 Boost Issue: What's your take?

[lexluthermiester]

It depends on what would be considered overvolting. Ryzen 3000 CPUs do run cores at 1.4-1.5V when boosting.

I haven't seen that. The highest voltage bump during boost I've seen is 1.375.

 

[londiste]

I haven't seen that. The highest voltage bump during boost I've seen is 1.375.

Single/dual core boost or all-core boost?
1.47-1.5V is what I see on one core. ~1.45V on two cores and less after that.
While high voltages are not sustainable on all cores anyway, the limiting factor is primarily power limit kicking in on 3+core load

 

[silkstone]

Single/dual core boost or all-core boost?
1.47-1.5V is what I see on one core. ~1.45V on two cores and less after that.
While high voltages are not sustainable on all cores anyway, the limiting factor is primarily power limit kicking in on 3+core load

On my 3600 I get 1.45v idle and 1.35v on load, though I still need to look into setting bios voltage to 'normal'.
My idle clocks are pretty much always at boost, but I do have some background tasks that keep my cpu at a constant 10-20%

 

[xtreemchaos]

ive not voted, all I can say is give it time, same happened with the 2700x at first my one wouldn't boost above 4.150 2 bios updates later I think f40 "Aorus"and it was hitting 4.300 like a good little processor. I agree where said its down to mobo and bios updates.

 

[Nordic]

My 3900X seems to boost just fine. Poll sure came across as trolling to me.

You have made the common mistake of assuming your experience is the same as everyone elses. Most threads about ryzen 3000 have people including myself complaining about boost not working as it should.

 

[Divide Overflow]

You have made the common mistake of assuming your experience is the same as everyone elses. Most threads about ryzen 3000 have people including myself complaining about boost not working as it should.

I made no assumptions on anyone else's experience. I simply reported mine and my opinion on the poll.

 

[Wavetrex]

Tbh I already forgot that my CPU was supposed to boost to "4.4"

After a month and a half of use, the only thing I care about is that my games play well and don't stutter, and my workloads complete fast enough, and especially... that it's faster (sometimes significantly) than what I had before.
if it does that at 4.3 or 4.4.... I stopped caring after first week of testing.

I didn't buy the CPU for the number written on the box, I bought it for the benchmark results which clearly showed how much faster it is than the previous one which I had. End of story (For me anyway).

 

[TheLostSwede]

Apparently, Ryzen Pro can do 5GHz...

 

[jesdals]

Well stil have to say, not worried with boost speeds when running 4500mhz on all cores

 

[phanbuey]

Well stil have to say, not worried with boost speeds when running 4500mhz on all cores

Golden chip.

 

[Octopuss]

Apparently, Ryzen Pro can do 5GHz...

That makes no sense.

 

[Wavetrex]

Apparently, Ryzen Pro can do 5GHz...

That makes no sense.

Well, it can, under LN2, so it's not necessarily a lie ;-)

But one should not forget that when Intel switched to 14nm with Broadwell, it was barely able to scratch 4.3 Ghz, with maybe 4.4 golden sample chips, any more and heat went completely through the roof.
It took them a bit of + ++ ++++ +++++++ to actually hit 5Ghz with that process.

I won't be surprised at all if AMD releases a new stepping sometime early next year (maybe in Threadrippers?), that can boost to 5.0 (if lucky).
It's only about getting those manufacturing issues resolved.

 

[TheoneandonlyMrK]

Im surprised by all this , my 2600X can be made to sit at 4.325 quite stable, stable enough to game If you can cool it which i can if the gpu is not also overclocked, which it never is, I actually game at 4.225 all core boost via a heavily abused Pbo algorithm so i cant help but wonder what gives , im so so tempted but dont think this a good value guy time to buy.

 

[Vario]

Well, it can, under LN2, so it's not necessarily a lie ;-)

But one should not forget that when Intel switched to 14nm with Broadwell, it was barely able to scratch 4.3 Ghz, with maybe 4.4 golden sample chips, any more and heat went completely through the roof.
It took them a bit of + ++ ++++ +++++++ to actually hit 5Ghz with that process.

I won't be surprised at all if AMD releases a new stepping sometime early next year (maybe in Threadrippers?), that can boost to 5.0 (if lucky).
It's only about getting those manufacturing issues resolved.

Just a guess here but I think it most likely clocked poorly because of the L4 cache EDRAM / Iris Pro.

 

[biffzinker]

Just a guess here but I think it most likely clocked poorly because of the L4 cache EDRAM / Iris Pro.

Shouldn't of mattered since the L4 eDRAM was on a separate die.

 

[Wavetrex]

Just a guess here but I think it most likely clocked poorly because of the L4 cache EDRAM / Iris Pro.

That doesn't explain why Broadwell-E was clocking much lower than Haswell-E on the HEDT platform, as none of them had an iGPU or EDRAM.

It's just the process... the "brand new" 14nm was worse than the well understood 22nm.
And now history repeats itself with 10nm (but in a much worse way).

Actually, that seems to be the case for AMD and TSMC... current TSMC "7nm" is not exactly better than GloFo's "12nm", as @theoneandonlymrk just said... 2000 series clocked very similarly with 3000.

BUT... we'll see what 7nm+ brings. 2020 is going to be even more interesting than 2019 !

 

[xrror]

I think AMD marketing got a little overenthusiastic with the numbers on the box. Which is a shame since they really didn't need to - and now there's this shadow over AMD with the perception that their clock speeds might be a lie.

Honestly they could have dialed back their box speeds by 200Mhz, and then everyone who did see the higher than 4.2Ghz would have been estactic, because it would be like AMD was giving you a bonus for nothing. That PR spin would have written itself.

"Hey everyone, AMD chips are actually faster than they advertised on the box!" ... says John and Jane Doe on reddit, etc.

I hate all forms of "boost" with a vengance. So, let me just say that I won't be saying this is an AMD exclusive issue. It's even present in the GPU scene. Heck, it practically started there.

Yea, the whole boost clock era we're in now is just a sad reminder that the days when a new process nodes = more performance are over. We're forced into this "mobile first" "power efficiency age" scraping for every last clock because physics is a harsh mistress. And there's no fun anymore with old school voltmods to crank clocks because this new stuff just burns immediately if you're not sub ambient.

 

[lexluthermiester]

You have made the common mistake of assuming your experience is the same as everyone elses. Most threads about ryzen 3000 have people including myself complaining about boost not working as it should.

Then try the advice I offered earlier, lower your voltage to no higher than 1.3v and your boost clocks will be fine.

 

[silkstone]

Then try the advice I offered earlier, lower your voltage to no higher than 1.3v and your boost clocks will be fine.

My 3600 crashes when I do this.

Current all core boost = 3.9
single core boost = 4.1
load voltage 1.35
idle voltage 1.45-1.5

Updates bios and chipset driver as well as changing the bios config. Still have the same issues.

 

[biffzinker]

My 3600 crashes when I do this.

Current all core boost = 3.9
single core boost = 4.1
load voltage 1.35
idle voltage 1.45-1.5

Updates bios and chipset driver as well as changing the bios config. Still have the same issues.

Yeah mine wants 1.43V for 4.175 GHz. I thought it was stable from stress testing but found out it was failing to run the Physics Test in Passmark's PerformanceTest.

 

[tabascosauz]

Then try the advice I offered earlier, lower your voltage to no higher than 1.3v and your boost clocks will be fine.

Bruh moment... that's not how it's working for a good number of us at all.

On Normal modes my board gives the chip an exact same dose of Vcore as it does on Auto, up to 1.4V full load. Why? Because it needs 1.3V for 4.0 and 1.35V for 4.1, that's why. Now scale that up to 4.35GHz and you'll see that not every Matisse is a golden egg like you assume them to be.

Now, if we're talking sub-4GHz, then you'd be wholly correct. 3.6GHz needs no more than 1V, which is probably where the hardware should be advertised, strictly speaking. But AMD wants to beat Intel and their own Pinnacle Ridge SKUs so we're demanding more than this 7nm experiment can currently deliver.

 

[Nordic]

I don't think it is fair to say that 3.6ghz is where this hardware should be, but it certainly does not like going over 4.2ghz.

 

[TheLostSwede]

Then try the advice I offered earlier, lower your voltage to no higher than 1.3v and your boost clocks will be fine.

If I touch my CPU Voltages at all, my system won't boot, so there's that...

 

[Vlada011]

I would like to know how much boost AMD 3900X on temperatures up to 55-60C on all cores 12 with HT Enabled?
Not frequency of first 1-2 cores.
I still think that whole picture is not here and because of that it's not time for investing in AMD.
Yes I agree, investing in Intel with inferior Interface is bad but I have feeling that within 2 years max, maybe less, 1 or 2 important features of AMD X570 will be obsolete compare to Intel and AMD or only Intel.
PCI-E Interface or DDR5, or maybe both and that's period I could use X99 without problems.

 

[mstenholm]

I would like to know how much boost AMD 3900X on temperatures up to 55-60C on all cores 12 with HT Enabled?
Not frequency of first 1-2 cores.
I still think that whole picture is not here and because of that it's not time for investing in AMD.
Yes I agree, investing in Intel with inferior Interface is bad but I have feeling that within 2 years max, maybe less, 1 or 2 important features of AMD X570 will be obsolete compare to Intel and AMD or only Intel.
PCI-E Interface or DDR5, or maybe both and that's period I could use X99 without problems.

I will tell once I get mine. 480 mm rad placed outside the case should give decent cooling. The rig have standing ready for at least five weeks now, well except the darn CPU. Ordered on the 7th of July.