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

[spoffle]

Marketing has never been AMDs strong suit. The 5GHz rumour had a lot of weight behind it, maybe AMD were pushing the boat a little too far when trying to reach that milestone.

I can see 5Ghz on the super super highly binned parts after a few steppings.

 

[G550]

If I had responded to the survey and answered only what it asked, My 3800x would be listed as one of the CPU's that can't hit rated boost clock.
The problem is, and this is important to this discussion, **my CPU hits rated boost clock,** (even a little higher sometimes. )

So when people use this survey to say that "most chips don't hit rated boost", I think they are making that statement with insufficient data to back it up.

My 3600X was tested as per Roman's instructions and the results submitted so they're part of the figures he reported.

2 cores boost to 4350 (the uppermost of the two large peaks in Roman's 3600x frequency distribution)
2 cores boost to 4300
2 cores boost to 4275

The cores peaked at exactly the same clock frequencies in the single core CB R15 test as they do when the system is left idle / used lightly loaded for a few minutes (at default/Auto voltages max VID is 1.45volts, which the Asus BIOS AUTO voltage implementation respects).

With the limits the current SMU firmware component from AMD imposes, nothing I've tried so far causes the processor to boost any further than 4350.

Being 50MHz shy of the advertised 4400MHz, matters less than not having the ability to trade some of the stability margin AMD builds into its VID table of each processor to reduce voltages, and hence temperatures, whilst maintaining clocks/performance.
At the moment, any meaningful undervolt I apply drops the multiplier by 0.25-0.5 and worse, initiates "clock stretching" seemingly to maintain the stability margin, so I'm losing a bit of performance even with a mild 43mV undervolt.

 

[Smak]

Im glad its just not me, and that there is something going on since I was going paranoid. I managed to get 62.9ns before but didnt want to put more voltage on the RAM so I left it at 3733Mhz.

I managed to get back down to around 63.7ns last night by setting AMD Cool'n'Quiet to AUTO in bios (MSI MEG ACE X570) not sure what other boards call it, maybe try that? My 3700X was still able to idle at 0.9v with C'n'Q set to auto (which I assume actually means off) whereas before if I turned off CnQ I would get a constant 1.5v ish so something is different.

 

[Nghtmare]

I managed to get back down to around 63.7ns last night by setting AMD Cool'n'Quiet to AUTO in bios (MSI MEG ACE X570) not sure what other boards call it, maybe try that? My 3700X was still able to idle at 0.9v with C'n'Q set to auto (which I assume actually means off) whereas before if I turned off CnQ I would get a constant 1.5v ish so something is different.

That setting does not exist on my Taichi though, I had almost every setting on Default except for the RAM, and SOC which have not changed.

 

[DragonQ]

That setting does not exist on my Taichi though, I had almost every setting except for the RAM, and SOC which have not changed.

Same.

 

[Nghtmare]

Same.

Your RAM took a hit on performance too?

 

[opethdisciple]

I'd say it would be the program doing the measuring rather than the memory latency changing.

 

[Nghtmare]

I'd say it would be the program doing the measuring rather than the memory latency changing.

It could be, it is driving me a little insane because it does not seem to affect everyone by the looks of it.

 

[Smak]

That setting does not exist on my Taichi though, I had almost every setting on Default except for the RAM, and SOC which have not changed.

I believe it's under C-STATE on Asrock boards

 

[Nghtmare]

I believe it's under C-STATE on Asrock boards

Is that when the cores go to "sleep" on Ryzen Master? Because I had that active before this happened.

I have a question, when I went to swap the thermal pad on the chipset a few days ago, there was a little drop of Kryonaut on the 3rd RAM slot, I tried to remove it with my finger and it went inside the connection points. Since its non conductive and I use 2/4 for my 2 DIMMs I left it there. Could that be causing this?

 

[Deleted member 209350]

https://www.techradar.com/uk/news/a...00-close-to-equalling-intels-entire-cpu-sales

Absolutely incredible

 

[Nghtmare]

https://www.techradar.com/uk/news/a...00-close-to-equalling-intels-entire-cpu-sales

Absolutely incredible

Yeah it is, hopefully now devs start taking advantage of Zen 2 architecture since at the moment everything seems to take more advantage of Intel's architecture.

 

[Deleted member 209350]

Yeah it is, hopefully now devs start taking advantage of Zen 2 architecture since at the moment everything seems to take more advantage of Intel's architecture.

Exactly what im hoping for! All this power is useless if nothing is going to be optimized for it

 

[jigger]

I must have slept through that night. Unless you mean 'up to 30%' in which case you might want to apply to join AMD's marketing department.

I mean 1/3rd of performance lost. I need 30% more servers for the same amount of work.

 

[jigger]

https://www.techradar.com/uk/news/a...00-close-to-equalling-intels-entire-cpu-sales

Absolutely incredible

The 3500X sounds interesting.

 

[Smak]

Is that when the cores go to "sleep" on Ryzen Master? Because I had that active before this happened.

I have a question, when I went to swap the thermal pad on the chipset a few days ago, there was a little drop of Kryonaut on the 3rd RAM slot, I tried to remove it with my finger and it went inside the connection points. Since its non conductive and I use 2/4 for my 2 DIMMs I left it there. Could that be causing this?

Not sure what slots your board recommends to use but surely you could clean up with alcohol cleaner and put them back in them slots?

Regards latency from my own findings the only way I can get really low readings would be if my 3700X had the issue of not idling at low voltages, if it was stuck at 1.45-1.5v my latency would be improved.

 

[Nghtmare]

Not sure what slots your board recommends to use but surely you could clean up with alcohol cleaner and put them back in them slots?

Regards latency from my own findings the only way I can get really low readings would be if my 3700X had the issue of not idling at low voltages, if it was stuck at 1.45-1.5v my latency would be improved.

I am meant to use 2 and 4 for dual channel, the thermal paste fell on slot 3. I will try cleaning it up when I get home since I have 90% alcohol.

 

[Deleted member 209350]

The 3500X sounds interesting.

Certainly does. Is it possible that it could be a replacement for the Ryzen 3 series? Because I havent even heard any rumours about the Ryzen 3, and feel they might be trying to get rid of that line and just have the 5,7,9

 

[TrM]

Certainly does. Is it possible that it could be a replacement for the Ryzen 3 series? Because I havent even heard any rumours about the Ryzen 3, and feel they might be trying to get rid of that line and just have the 5,7,9

I think the ryzen g series replaces the ryzen 3 myself dropping the ryzen 3 at x price point would allways be completion vs there 4core 8 thread 3400g and that cpu gives good value at that price point really

 

[G J]

Who had thought that bringing a 6 core with hypertheading with great performance at under £200 would sell like hotcakes, if Intel refuse to drop prices and compete then all the better for AMD.

A 6 core no SMT 3500 for even cheeper, but what about the handful of games that cause it to turn into an unplayable stutterfest.