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

[TNA]

When you guys get your chipset drivers, do you get it from the mobo manufacturers website or direct from AMD? Any advantages/disadvantage of using AMD's one?

 

[Illuminist]

When you guys get your chipset drivers, do you get it from the mobo manufacturers website or direct from AMD? Any advantages/disadvantage of using AMD's one?

The amd one always seems to be way more recent than the gigabyte website, so I go for that usually

 

[TNA]

The amd one always seems to be way more recent than the gigabyte website, so I go for that usually

Thanks man. Will go for that next time

 

[alec]

Almost certainly just the monitoring software not understanding the clock speeds. Remember Ryzen clock speeds can adjust extremely quickly and in 25 MHz increments. The performance difference is ridiculous if this is accurate but I am not convinced.

I don't think this sort of improvement can be attributed only to clocks. Rather some improvement in branch prediction specific to gaming loads.
My secret hope is that it fixes the anomaly. Zen 2 is roughly equal or sometimes better in all workloads, single and multithreaded, to Coffee Lake. Except most games where it is consistently worse.
Can't wait until we get more game benchmarks

 

[DragonQ]

Didn't you both notice that despite the lower clocks, the Agesa 1.0.0.4 keeps all the cores loaded. So, it isn't about the frequency but about how the game threads are distributed and executed by the CPU.

It's possible they've done something to change thread loading in Windows but I can't see how that'd be in the AGESA, surely it'd be in chipset drivers or Windows updates.

 

[Panos]

@TNA @humbug
Because we thought the 10% fps improvement was pretty weird, found more things to add into the context.
Someone posted CB20 with 3600C16 ram on 3800X stock with AGESA 1004.

https://www.reddit.com/r/Amd/comments/dpf037/3800x_1004_beta_bios/

Now compare the improved 3800X above (in addition to my previous post video), with this 9900KS review with Ryzen CPUs on AGESA 1003 ABBA and using 3600C16 for all CPUs. (both Intel & AMD).

https://www.youtube.com/watch?v=Kv0yCUYOPFQ

On CB20 we see 5% flat perf increase!

 

[4K8KW10]

It's possible they've done something to change thread loading in Windows but I can't see how that'd be in the AGESA, surely it'd be in chipset drivers or Windows updates.

There is no Windows update and there is no new chipset driver. The latest one is old from September 27.

 

[Zeed]

@TNA @humbug
Because we thought the 10% fps improvement was pretty weird, found more things to add into the context.
Someone posted CB20 with 3600C16 ram on 3800X stock with AGESA 1004.

https://www.reddit.com/r/Amd/comments/dpf037/3800x_1004_beta_bios/

Now compare the improved 3800X above (in addition to my previous post video), with this 9900KS review with Ryzen CPUs on AGESA 1003 ABBA and using 3600C16 for all CPUs. (both Intel & AMD).

https://www.youtube.com/watch?v=Kv0yCUYOPFQ

On CB20 we see 5% flat perf increase!

Where did you get magic 5% from ??

 

[Panos]

Where did you get magic 5% from ??

4860 to 5061 is over 4%

 

[Zeed]

https://twitter.com/1usmus/status/1188018589112123393

AGESA 1.0.0.4B NEW FEATURES SMU 46.54 (additional boost) CPU Ratio Apply Mode (CPU or CCX) VDDG CCD/IOD Voltage Control Mode (Async or Sync + manual adjustment) PBO Max Offset Voltage (-0.1v max) Indirect Branch Prediction Speculation Performance increase

 

[Zeed]

4860 to 5061 is over 4%

AAA ye its Mainly due to higher boost on 1.0.0.4 al core and 3800 finally hitting Boost that AMD said it was boosting to. Took em 6 months but finally fixed boost up.

Still Amd got nothing on 9900ks in games if someone is on 2080ti runnning 1440p with 144hz monitor

 

[Pannalicour]

AAA ye its Mainly due to higher boost on 1.0.0.4 al core and 3800 finally hitting Boost that AMD said it was boosting to. Took em 6 months but finally fixed boost up.

Still Amd got nothing on 9900ks in games if someone is on 2080ti runnning 1440p with 144hz monitor

Lol Intel still got nothing on 3900x, streaming and gaming simultaneously on slow preset division 2 x264 obs in that price point.

 

[jigger]

AAA ye its Mainly due to higher boost on 1.0.0.4 al core and 3800 finally hitting Boost that AMD said it was boosting to. Took em 6 months but finally fixed boost up.

Still Amd got nothing on 9900ks in games if someone is on 2080ti runnning 1440p with 144hz monitor

You mean apart from the 3700/3800

 

[Zeed]

Lol Intel still got nothing on 3900x, streaming and gaming simultaneously on slow preset division 2 x264 obs in that price point.

Yuop made epic move to grab em day 1. Now Your looking at 550 at lest for one....

 

[humbug]

@TNA @humbug
Because we thought the 10% fps improvement was pretty weird, found more things to add into the context.
Someone posted CB20 with 3600C16 ram on 3800X stock with AGESA 1004.

https://www.reddit.com/r/Amd/comments/dpf037/3800x_1004_beta_bios/

Now compare the improved 3800X above (in addition to my previous post video), with this 9900KS review with Ryzen CPUs on AGESA 1003 ABBA and using 3600C16 for all CPUs. (both Intel & AMD).

https://www.youtube.com/watch?v=Kv0yCUYOPFQ

On CB20 we see 5% flat perf increase!

I'll take the performance jump in whatever way it comes

Having said that we do know the Windows scheduler hampers Ryzen in some games, AC being one of them, if AMD can figure out how to fix Microsofts scheduler for them we would see gains like that, this was my thinking.

 

[Twinz]

My R20 MT went from 5144 best ever on ABBA to 5154 first try. (3800X stock clocks)

Run-to-run benchmark variance seems to have tightened up considerably. It's as if something in the new microcode is making the CPU "try harder"...for lack of a better way to put it.

 

[gpuerrilla]

Having said that we do know the Windows scheduler hampers Ryzen.

Yeah theres enough evidence out there that Microsoft could also offer potential gains if they optimised their end for it.

 

[Vulture]

Guys i got yesterday 3800x, i have latest bios and chipset driver but single core speed can't go higher than 4.43 ghz (tested on cinebench and watching speed on ryzen master). Minimum clock speed fluctuation can be as low as 3.81 ghz (saw on task manager) despite specification base of 3.90 ghz, is it normal?

 

[TNA]

Guys i got yesterday 3800x, i have latest bios and chipset driver but single core speed can't go higher than 4.43 ghz (tested on cinebench and watching speed on ryzen master). Minimum clock speed fluctuation can be as low as 3.81 ghz (saw on task manager) despite specification base of 3.90 ghz, is it normal?

New BIOS coming out soon. Might solve that issue.

 

[Vulture]

New BIOS coming out soon. Might solve that issue.

I thought 1.0.03 abba should solve that, and is it normal for task manager to report base clock 3.89 ghz instead of 3.90, and minimum clock speed fluctuation can go as low as 3.80 ghz instead of 3.90?