AMD Zen 2 (Ryzen 3000) - *** NO COMPETITOR HINTING ***

[Grafter67]

what kit do playstation and xbox use?

 

[willhub]

Well 4.1Ghz 1.25V (around 1.27v) and idling at 29 degrees (door open and tower is near door)

1.29V load and 75 degrees.

 

[stoke1863]

so are people manually changing voltages right now? or are the default auto voltages safe?

 

[willhub]

so are people manually changing voltages right now? or are the default auto voltages safe?

Default it's toasty, I'm trying 4.1GHz all core and playing with voltages, I can't get this thing to work stable over 4.1GHz all core. trying for 1.35V

 

[nicktay2605]

so are people manually changing voltages right now? or are the default auto voltages safe?

Not sure about all board manufacturers but Jayztwocents has just posted an overclocking discussion of the 3900x and the stock voltages the Asus Crosshair board tried to set were very high, e.g. 1.47 vcore, resulting in 40°c idle, and a memory voltage of 1.48

 

[chroniclard]

Not sure about all board manufacturers but Jayztwocents has just posted an overclocking discussion of the 3900x and the stock voltages the Asus Crosshair board tried to set were very high, e.g. 1.47 vcore, resulting in 40°c idle, and a memory voltage of 1.48

Ryzen+ regularly uses 1.5+ when boosting, dont think its an issue (assuming its working correctly)

 

[willhub]

I couldn't manage to link to it but read an article from 1 day ago AMD confirms no Zen 2 desktop parts you'll get much out of them, most likely the 65W 3700x is the best bet for overclock apparently and we're better off overclocking ram and infinity fabric

https://www.google.com/url?sa=t&sou...EEAg&usg=AOvVaw25eCosIa0WMv3NCehRguSQ&ampcf=1

 

[harney]

Things could be looking good for DAW performance if your interested

https://www.gearslutz.com/board/music-computers/1266481-ryzen-3000-series-3.html

page 3

 

[Phemo]

Well I wasn't going to bother, but after seeing the 3700X in stock and figuring I can sell my 2700X to recoup a whack of the cash I thought I might as well. Be interesting to see how it does in my ITX box though obviously I'm not expecting massive gains. Loving that I can upgrade CPU without having to change motherboard for once.

 

[willhub]

Is 470/4800 in Cinebench R20 looking ok for 4.1GHz?

 

[CAT-THE-FIFTH]

It is a massive improvement, I'd like to see more evidence and my 5960X scores 53.7 on that same benchmark so finally Zen 2 is faster in Arma 3
five years later by 6% or so. I'm pleased for that result as it means AMD have caught Haswell and are likely pushing past Skylake now/soon.
I'd be buying one if I wasn't on 8 cores already.

The test was also using a Vega64 too.

 

[Kashinoda]

The test was also using a Vega64 too.

Arma is almost entirely CPU bottlenecked in my experience, that and server performance.

 

[CAT-THE-FIFTH]

Arma is almost entirely CPU bottlenecked in my experience, that and server performance.

Any chance you can run the benchmark if you have the game??

It is this benchmark:
https://steamcommunity.com/sharedfiles/filedetails/?id=375092418

 

[awaybreaktoday]

DID YOU KNOW? The new version of AMD Ryzen Master can show you core behavior that other tools can't or don't! Examples include: cc6 sleeping cores, sub-2200MHz idle cores, sub-1V idle voltages, fastest CPU physical core, your motherboard's VRM capacity, and more.

 

[Panos]

DID YOU KNOW? The new version of AMD Ryzen Master can show you core behavior that other tools can't or don't! Examples include: cc6 sleeping cores, sub-2200MHz idle cores, sub-1V idle voltages, fastest CPU physical core, your motherboard's VRM capacity, and more.

Master and Wattman are exotic pieces of software that are not as simple as move one slider to the right.

 

[Kashinoda]

Any chance you can run the benchmark if you have the game??

It is this benchmark:
https://steamcommunity.com/sharedfiles/filedetails/?id=375092418

I get my 3700X tomorrow so this will make for a good comparison when I run it again.

4790K @ 4.4GHz (all core), 16GB 1800MHz RAM, GTX 1080 Armor OC

1920x1080 Ultra Preset - 33.8 FPS
3440x1440 Ultra Preset - 32.3 FPS

1920x1080 High Preset (same as review posted) - 40.7 FPS

Note sure how Jacky60 is getting 53.7 FPS on that, I guess Arma is making use of those 4 extra cores.

 

[chrcoluk]

For people concerned about idle voltages have a read here of AMD Robert's explanations.

https://www.reddit.com/r/Amd/comments/cbbfce/3900x_being_overvolted_on_amd_ryzen_power_plans/

In short use the AMD Windows Power Plan and don't worry about it.

yeah amd power plans are aggressive, and have a very high min clock speed. I would love to understand their reasoning for this, but I accept the explanation. However as an end user I would for sure personally verify the response by checked ryzen master to be sure the cores are actually been parked as I Wouldnt be happy with a 1.3+ idle voltage.

Also even without low idle voltages c states should lower the voltage, so curious if the OP on that reddit page had tinkered with c states, the AMD rep may be suggesting that C1 to C3 states are all been skipped and going straight to parking.

My 2600x at non turbo voltages (and under clocking disabled) will have voltages of about 1.05v, (99% in power plan). However AMD's power plans have extremely aggressive p-states where by cpu's will spend really small amount of time at non turbo clocks when under any kind of load. I expect this all contributes to what the poster was seeing.

I personally dont notice any difference in responsiveness between microsoft and amd power plans, but the mean average idle voltage is considerably lower on microsoft so thats what I go with on my ryzen rig when its bare metal windows.

Some info here on it.

https://forums.guru3d.com/threads/t...ng-cooling-part-2.423134/page-17#post-5679536

So they made p-states more aggressive similar to what speedshift does on intel, one of confirmed reasons was for benchmarking (not surprised) but a real world game is also given as an example, if its actually improving actual real world usage then fair enough its hard to argue with that.

--update after finishing read amd rep post---

It seems confirmed p states are too aggressive, apps like hwinfo when polling are triggering xfr clocks so polling apps always see voltage for boosted workloads. I expect AMD will likely retune p-states or work with app vendors so they poll in a way to not trigger the behaviour.

 

[antsm81]

3800x clocked to 5.9ghz? https://www.techpowerup.com/257234/...-the-market-with-ryzen-3000-series-processors

Edit: looks like its actually a single core overclock.

 

[chrcoluk]

Which is no where near as intensive as a Prime torture.

cpu based rendering is probably the most intensive real world usage most processors go through, so his point is valid in my opinion, if I was rendering on my cpu for hours I would definitely do intensive stress testing. not sure if it would be prime 95, but it would be at least some thing like real bench encoding running for couple of hours or something.

 

[Potatowithearsontheside]

I'm not denying the game engine is archaic, I'm not disputing that things could be different however that doesn't account for AMD CPUs being seemingly unable to perform in this game until now. All the AI is on pretty much a single thread and the core game engine dates from before multi-core CPUs nearly 20 years ago. I'd have thought that was plenty of time for AMD and if the optimisation could be done better by a primary school pupil don't you imagine it would have been done by now for the price of a couple of cans of pop or McDonald's?
If I was doing productivity work, video editing etc I'd buy Zen 2 3900X or 3800X but I'm not. Even Steve at gamers Nexus said the same, buy Zen if you use your PC for work (content creation etc) and gaming it will be OK but if it's ONLY for gaming then 9900K is still the way to go.
If 3800X/3900X were maintaining 4.8Ghz with 2 or 3 cores consistently I'd buy 3900X tomorrow but the clockspeed deficit and failure to maintain advertised clocks precludes that.

Maybe, just maybe, AMD have got better things to do with their time than optimise for one **** game that you play...? You know, its not like Intel performs that much better at it either, so have they had their fingers up their butts ever since because they've failed to further optimise for the same **** game...?
As an engineer you design based upon future needs, not what one bloke on the internet has decided is the be-all and end-all of CPU performance metrics. There's a reason why Pacman isn't used in gaming benchmarks. Get with the times.