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

[G550]

Just saw this, i'll do this later, i have to go out shortly.

Avoid installing KB4512941 for now if you have already disabled the Web search functionality within Windows Search (BingSearchEnabled registry key set to zero to disable Microsoft's intrusive search behaviour).

After installing the update I noticed that my 3600X was no longer downclocking and the processor core voltage remained stuck at its maximum, with taskmanager showing Cortana stuck at 13% processor usage and clearly the culprit.

The Windows Search box will still pop if you click on the icon, but is no longer functional.


Looks like a Microsoft bug (intentional or otherwise), and plenty of discussion about it already:

https://www.ghacks.net/2019/09/02/c...windows-10-1903-update-reported-to-microsoft/

 

[jigger]

Between 14 and 15% on the lows, 10% on the averages yes

Its similar to what HUB found, tweaking the timing manually with DRam Calc.

3800MT/s, 1900Mhz IF: 57 FPS minimums
3800MT/s, 1900Mhz IF: Manual (DRam Calc) 68 FPS minimums +19%

That is pretty big performance jump.

 

[humbug]

@G550 thanks for the heads up.

If they don't it seems strange for AMD to recommend that he should use that.

I don't understand why people are defending AMD for this, don't get me wrong Zen 2 are great CPUs and 100Mhz makes next to no difference in the real-world performance of them, for me this is simply a matter of AMD misleading people, whether they set out to do that is irrelevant, it would've been simple to lower the max advertised boosts by 100Mhz and probably no one would've batted an eye, it probably would've even done them some favors as it wouldn't have made PBO redundant.

I'm not directing this at you personally BTW humbug, it's just a general moan about how trying to excuse AMD for over promising and under delivering does them (AMD) no favors.

I agree, and i do think this is misleading, a black mark against AMD there but i think they will learn from it, watch AMD do what Intel do when it comes to Ryzen 4000, just quote the base clock and that's it.

This was an attempt at pushing the boundaries of honest marketing and it was completely unnecessary.

As i said to the Video.

I have a 3600, its around +/- 4.1Ghz in games and Cinebench, so 100Mhz (2%) short of the 4.2Ghz, if they advertised it as a 4.1Ghz CPU or even a 4Ghz CPU i still would have bought it, you are right, its a very good CPU, its perfectly stable even with very overclocked Crap Corsair 3000MT/s to 3333MT/s OC (Hynix AFR) RAM, it runs quite cool under a 120mm AIO and the performance is excellent, to me that's all that matters and i'm very happy with the CPU.

AMD are on the up, they have the mindshare back, people like and are impressed with their products once more, just chill out AMD and let your products speak for themselves, no need for this dodgy marketing.

 

[ben90]

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.

 

[jigger]

@G550 thanks for the heads up.



I agree, and i do think this is misleading, a black mark against AMD there but i think they will learn from it, watch AMD do what Intel do when it comes to Ryzen 4000, just quote the base clock and that's it.

This was an attempt at pushing the boundaries of honest marketing and it was completely unnecessary.

As i said to the Video.



AMD are on the up, they have the mindshare back, people like and are impressed with their products once more, just chill out AMD and let your products speak for themselves, no need for this dodgy marketing.

I hope AMD don’t go the way of Intel’s advertising. TBH the boost clock will probably be addressed.

 

[Illuminist]

One thing i found that many may not know about, if you want your FSB clock to stay at 100Mhz, not 99.67Mhz or whatever..... turn off Spread Spectrum.

Any idea where I will find spread spectrum, noticed yesterday my BLK clock was 99.033mhz making my pc 4.389 or something. I set it as 100mhz in the bios but it isn't

 

[humbug]

Any idea where I will find spread spectrum, noticed yesterday my BLK clock was 99.033mhz making my pc 4.389 or something. I set it as 100mhz in the bios but it isn't

Depends on the motherboard, everyone is different. Don't know...

 

[G550]

Anyone else noticed that running a Ryzen 3000 processor with a manual/"fixed" clock no longer allows Windows to alter the processor multiplier according to system load / power plan settings?

With a 2600x fitted to my Prime X470 pro board, the BIOS would expose two parameters (FID FrequencyID and DID DividerID) to set the multiplier.
With the Ryzen 3600x in the same board, the BIOS just presents you with a single box where you key-in a set multiplier.

With FID/DID (multiplied internally by 2) set to give a 42.25 ratio, the 2600x scaled between 2.2GHz and 4.225GHz depending on system load.
Ran it like that for over a year.

With a fixed multiplier entered in the one box the BIOS exposes for the 3600x however, it just sits at that multiplier and will never downclock.

Anyone able to have their manual clock scale with load, or is this just a new limitation imposed by the design of the 3000 series processors?

 

[G550]

Any idea where I will find spread spectrum, noticed yesterday my BLK clock was 99.033mhz making my pc 4.389 or something. I set it as 100mhz in the bios but it isn't

Somone asked about that on a different forum recently. Apparently for Aorus boards, its in the tweaker section of the BIOS, but you must have CPU Clock Control set to Auto for the setting to become visible.

 

[humbug]

Anyone else noticed that running a Ryzen 3000 processor with a manual/"fixed" clock no longer allows Windows to alter the processor multiplier according to system load / power plan settings?

With a 2600x fitted to my Prime X470 pro board, the BIOS would expose two parameters (FID FrequencyID and DID DividerID) to set the multiplier.
With the Ryzen 3600x in the same board, the BIOS just presents you with a single box where you key-in a set multiplier.

With FID/DID (multiplied internally by 2) set to give a 42.25 ratio, the 2600x scaled between 2.2GHz and 4.225GHz depending on system load.
Ran it like that for over a year.

With a fixed multiplier entered in the one box the BIOS exposes for the 3600x however, it just sits at that multiplier and will never downclock.

Anyone able to have their manual clock scale with load, or is this just a new limitation imposed by the design of the 3000 series processors?

When experimenting with overclocking i think it did this, yes.

I don't bother overclocking as i don't gain anything, with reasonable volts, 1.35v, its only stable upto about 4.1 to 4.15Ghz which is what it runs at in games anyway, anything higher than that i need silly volts, when i see some Ryzen 3000 owners, mostly 3900X owners getting 4.4 even 4.5Ghz at 1.30 or slightly higher volts i'm always like HOW THE FFFFFFFFFFFFFF??????????????????

 

[Illuminist]

When experimenting with overclocking i think it did this, yes.

I don't bother overclocking as i don't gain anything, with reasonable volts, 1.35v, its only stable upto about 4.1 to 4.15Ghz which is what it runs at in games anyway, anything higher than that i need silly volts, when i see some Ryzen 3000 owners, mostly 3900X owners getting 4.4 even 4.5Ghz at 1.30 or slightly higher volts i'm always like HOW THE FFFFFFFFFFFFFF??????????????????

4.5ghz at 1.3675v
4.4 at 1.35, but runs at 1.31v
Why doesn't it do that for you?

 

[humbug]

4.5ghz at 1.3675v
4.4 at 1.35, but runs at 1.31v
Why doesn't it do that for you?

Crap silicon? binning...

 

[G550]

I don't bother overclocking as i don't gain anything, with reasonable volts, 1.35v, its only stable upto about 4.1 to 4.15Ghz which is what it runs at in games anyway, anything higher than that i need silly volts, when i see some Ryzen 3000 owners, mostly 3900X owners getting 4.4 even 4.5Ghz at 1.30 or slightly higher volts i'm always like HOW THE FFFFFFFFFFFFFF??????????????????

Well, I never spent much time testing overclocks, but my 3600x does ~4.275G at 1.28ish volts which is where I've just set it back to given the much lower thermal output at that level, even if that does leave it sitting permanantly at its max clock.

4.2G wasn't quite stable at 1.2V and would probably need ~1.22V at a guess.

4.325G wasn't quite stable at 1.32V when I tried it, would probably need 1.33-1.34V, but at that voltage I find you're well into diminishing returns and rapidly becoming temperature limited anyway given the heat density due to small size of the 7nm CCD's (core dies).

Personally, I wouldn't go much above 1.3V for a fixed overclock unless you can manage the heat in situations where its heavily loaded (ie watercooling or a substantial air-cooler).

 

[andy_mk3]

My 3900x will do 4.5ghz at around 1.4v, but only just. I wouldn't call it massively stable. Its reasonably happy at 4.4ghz but 4.35ghz seems the sweet spot where I can run 1.3v and have it stable.

 

[Murphy]

How do you feel about Intel? People lost 30% performance over night and Intel knew it would happen.

IMO that's a false equivalence and i suspect you're attempting to bait me, but I'll bite.

It's a false equivalence because that 30% performance loss is dependent on workload and the performance in specific workloads or performance in general is not advertised or even listed in official specifications, that's not to say losing performance because of shoddy security practices are something that doesn't leave a bad taste in the mouth, however a loss in performance due to optional security fixes released after the product was launched is not the same as selling someone a product that you say does X when it only does what you say it does in a minority of cases.

Personally i couldn't care less about the clock speed of a CPU and i suspect AMD may have felt increased pressure to push clock speeds when those silly 'leaks' surfaced showing 5Ghz and plenty of people seemed to care more about the clock speeds than the performance, it's possible AMD didn't want people to be disappointed when they found out Zen 2 was anything from 500-1000Mhz lower than the 'leaks' suggested.

 

[humbug]

Some argue Intel are getting the performance they are by skipping a lot of the security checks in the code, Coffeelake came after Spectre and Meltdown.

Or are Intel just less competent CPU architects than AMD?

 

[jigger]

IMO that's a false equivalence and i suspect you're attempting to bait me, but I'll bite.

It's a false equivalence because that 30% performance loss is dependent on workload and the performance in specific workloads or performance in general is not advertised or even listed in official specifications, that's not to say losing performance because of shoddy security practices are something that doesn't leave a bad taste in the mouth, however a loss in performance due to optional security fixes released after the product was launched is not the same as selling someone a product that you say does X when it only does what you say it does in a minority of cases.

Personally i couldn't care less about the clock speed of a CPU and i suspect AMD may have felt increased pressure to push clock speeds when those silly 'leaks' surfaced showing 5Ghz and plenty of people seemed to care more about the clock speeds than the performance, it's possible AMD didn't want people to be disappointed when they found out Zen 2 was anything from 500-1000Mhz lower than the 'leaks' suggested.

TBH I think it’s you looking to bait but, if think 50mhz on single core is a bigger deal..

I don’t remember initial reviews mentioning low boost clocks, maybe the difference is a combination of, board, bus speed, chipset drivers and micro code.

 

[Murphy]

TBH I think it’s you looking to bait but, if think 50mhz on single core is a bigger deal..

I don’t remember initial reviews mentioning low boost clocks, maybe the difference is a combination of, board, bus speed, chipset drivers and micro code.

I wasn't the one to ask a leading question, that was you.

I don't think 50Mhz on a single core is significant, i assume that's the statement you intended to make as that first sentence is a little disjointed, I actually said in the post you quoted to pose your leading question that "Personally i couldn't care less about the clock speed of a CPU", what i care about is the reputational damage that listing a product with a specific boost clock that the majority of customers can't achieve has done to AMD when instead of over promising and under delivering they could've under promised and over delivered, they could've simply set max boost clocks 200Mhz lower and allowed anyone who switched on PBO to break the max boost ceiling they placed on each CPU.

The reason why most reviews didn't mention the low boost clocks is because they tested on AGESA 1.0.0.2, a pre-release AGESA version, a version that AMD told reviewers shouldn't be used something like 2-3 days before launch, something (afaik) only two reviewers had the time to retest with the release version 1.0.0.3 of AGESA.

 

[Twinz]

At this point, if we take the survey as accurate, it shows that most people don't hit advertised boost in a single application.

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.

AMD is hiding behind a technicality, and I'm glad they are getting called out for it. Max boost has had an expectation attached to it for a long time and they used that expectation to imply capabilities that don't exist.

However, I have yet to see enough evidence to conclude that a minority of the chips can hit their boost speeds. This survey certainly doesn't prove it.

Also, I get that AMD supposedly told him to use this specific application and that was stupid since it's not a good way to get the chips to hit max boost clock. However, the box doesn't claim "max boost speed in CB20 ST".

 

[Disco_P]

How do you feel about Intel? People lost 30% performance over night and Intel knew it would happen.

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.