PC Processors

[ionstorm]

Hi AMD...

I have a 3700X only 1 core boosting just beyond 4.3ghz, nothing anywhere near 4.4..

My major gripe was I RMA'd this chip back to Amazon thinking it's faulty, but the new CPU is even worse (hell, I can't even have a stable 4225 OC on it like my current).

Anyway, neither CPU will hit the 4.4ghz advertised boost on any core at any time doing anything.  I've tried a Cinebench Single Core test and the max one core got was 4342mhz, with an average of 4250mhz for the entire test... let me add, that core boost wasn't the highest during testing.

Infact it seems to only boost it's highest on very light load, which is entirely pointless.  During an OCCT test I'm lucky to see 4.2ghz on hit, and it usually hovers around 4100-4150.

AMD, I've tried everything, I've even tinkered with the PBO settings and scaling, Ryzen Master states my max boost is now 4550mhz, I've got everything tweaked with PBO (not that i need to), and whilst YES it is faster than just normal PBO settings, it's nowhere near advertised.

Now i've had a 1700 and a 2700X, and honestly, my 2700X ran at 4.3ghz, 4.2ghz stable on low Volts for ace temps (fiddled in the hot weather we had, was epic).  So basically we're getting an IPC gain and clock decrease, which is fair enough, but you've sold me a 4.4ghz boosting chip (and don't get me started on the binning huge huge BAD move of it being on *some cores only*) cos that's not a rant you want to hear. I've tried lowering my DDR4 3600 down to 3200 to see if that helps, and it doesn't change a thing.

I also noticed a funny thing on Ryzen Master whilst running Cinebench, My Fasest CCX 0 and my Fastest Core in the System is not being used by the Cinebench test, infact during cinebench the C07 (Fastest) is Sleeping... Which seems like utter stupidity in an intense Single Core workload, like, what the actual hell?!

TLDR.

AMD, I bought a 4400mhz boosting chip, it doesn't boost to that, nor does the replacement chip (which is actually worse).

What you going to do for a long term Beta Tester of Ryzen?

[shinkojiro]

Whole Setup?

[brucer]

The issue lies with tsmc which manufactures the cpu, Amd, your motherboard manufacturer, the motherboard and its bios, amd dragging their feet..  What motherboard and bios version are you running?

[xlox]

I think 4.3 is pretty near 4.4

[brucer]

What would you do if your paycheck was almost what you earned?

 4.3 is not 4.4 as advertised..

[xlox]

this is no paycheck this is a bonus , but people are naive with advertisment it seems, still believing in santa , and not inform themselves before buying ... welcome in 2019 ...did you notice it is overlclocking , and what is advertised is MAX 4.4 on 1 core ? did you read reviews before buying ? did you noticed there is hardware websites that sell cpu more expensive when they are tested at some frequencies ...

if that doesn't convince you, maybe you should rma and read reviews to know what you buy ... in fact , i didn't buy my 3700x for using 1 core at 4.4, but to use 8 core at 4.2, and i get what i wanted ... in fact not getting 4.4 on 1 core seems only be a tweaking "problem" ,with no doubt it can do it... just have to find the right way to configure ... i think tools and drivers are not really as they should be , but performances are here ...

i did some tests and achieve to get 4.4 on several cores with ryzen master, I wanted only on 1 core but the fact is it's not because some parameters don't permit it that it can't , so it's only a matter of patience and to know what you want

[brucer]

Are you referring to youtube reviews after the cpus went onto the market and actual people started testing them or are you referring to the presale hype youtube videos by paid shills?   I mean I can go find any numerous youtube videos posted about low advertised boost frequencies after the release date. .. Der8auer and Hardware Unboxed youtube videos seems to come to mind, and Hardware Unboxed is even an amd shill, several others also made videos of it.. 

To add, This is exactly what I said was happening right from the beginning... Its a culmination of manufacturers at fault, tsmc, amd, and board manufacturers. It starts at the top and trickles all the way down the line..

[ddmeltzer8]

i cant believe that u actually said that!what a clown!

[ajlueke]

Hello ionstorm‌,

What are the other specifics of your setup?  A couple of tips I can share through my own testing.

The Windows scheduler will not appropriately load cores on any Windows version other than Windows 10 1903. 

Even on Windows 10 1903, boosting can break "especially in lightly threaded scenarios" in you enable XMP (D.O.C.P.).  This appears to be releated to the Bclk value.  Quite a few RAM profiles change this value to 100.0manually.  It is best to leave Bclk at default and enter timings manually.  Any Bclk value other than the default appears to affect boosting.  Somehow, it appears that manually setting the value, shuts off spread spectrum and forces a constant Bclk.  I don't think this behavior is intended, but is does negatively impact boost performance.

Any UEFI based off AGESA 1.0.0.3XX will not boost as high as those based off 1.0.0.2.  1.0.0.2 was the best performer as far as boosting went, but had more RAM stability issues.  If you don't have RAM issues, with the older AGESA I would recommend that for maximum boost performance.

[Earnhardt]

I'm sure its the same for all 3000 series CPU's:

Ryzen 7 3800X Show Different Boost Clocks with different motherboards 

[xlox]

ok brucer so , it's working as expected , case closed , you can stop spamming, we already all saw this article and it doesn't change anything. Just rma and don't buy any computer anymore , you don't understand how it works and don't want to

[shinkojiro]

more specifically don't buy into new architectures anymore. Intel has a long tried and true 10+ year old architecture that just works one can revert to. Wait some years for the willing pioneers to help iron out all the kinks before you mess with it.

[brucer]

amd should quit releasing, at best, beta products.

[brucer]

looks like i was correct..