Ryzen 7 3800X Show Different Boost Clocks with different motherboards

I would love to see similar tests that show what the all-core boost is on several motherboards, and it would be especially nice to see B350 and X370 motherboards included in such tests. How about it, HH?

 

I have the 450 variant and I can't complain either though I have a 3700X it gets up to 4375.

 

The original video is about the advertised boost clock, it kinda proves that CPU can boost to it. Go watch the video, it will clarify doubts.

 

And that rises pretty interesting question. How does those 3 boards at top get clock above max rated "w/(o)" enabling PBO.
If anything, it looks like inconsistent settings in across different BIOSes/boards.

And apparently, clock is not as important as performance. Not that I see peek ST performance +- few percent as important as total MT performacne of CPU.
Losing 5% from total performance is big deal (May sum as total loss of entire core). But losing it from 1 core boost does not really affect anyone!!!

Why? Because practically any workload runs in 2+ loaded cores. Especially gaming.

 

Why so many PCI-e x16 slots on a motherboard that has a CPU that only supports like 24 PCI-e lanes!?!

Man it does look nice though.....

Come on ThreadRipper 3 from Zen2 please!.!.!.

 

SLI/multi-gpu is barely used yet you will struggle to find a single motherboard without at least 2 x8 slots...
(Btw, I think slots sizes are usually called 2x, 4x and 8x)

 

There is still variability in multicore boost clocks - even on the same motherboards during the same workloads. I've experienced this in handbrake renders where encoding the same file will force all the cores to go anywhere from 3.8Ghz to 4.15Ghz on my machine for the entirety of the workload. I don't really mind it because it's still significantly faster than my 7820x but it can effect the time of the render by a few minutes over longer workloads.

 

I have 1 BIOS that can reach MAX boost and 4 BIOS on the same motherboard that can not reach max boost.

 

i just built a friend a 3700X - MSI 570X build and made him do tests with CPUz and Aida64 plus other tests,

While he did the tests Task Manager was observed and his chip was only running @Max 4.2.

I was worried something was wrong with it because it never reached the advertised 4.4

He was using the default cooler while he waited for Noctua to send him the proper bracket for AM4 since he never got one.

Once he gets the NH D-15 installed i'll make him check again.

I told him to check regularly for Bios updates but he's not comfortable doing bios updates and i live in Wales and he Scotland.

 

same boat I had 3900x and prestige creation in basket but since 3900x still aren't available here (pre-orders are going berzerk delivery dates go down then up, now 4-6 weeks ><) never finalized the purchase, I made the mistake of pre-ordering a high end cpu once, an old intel, by the time they had stock 6month+ had passed and the price was much lower and the cpu was soon obsolete (new release) now I'm waiting for 3950x a that point
I hope creation gets better tough or is in multi core as someone wrote

BTW DerBauer is gathering data :

 

Silicon lottery is one thing, but that should be about overclocking and undervolting the cpu. not false advertised stock/boost speeds, that should be hit without question, specially for say 3800x, when its then getting worse results than a 3700x. if it's a motherboard issue, then same thing, either they should say we don't support 3000 gen, or if it's from the 5xx series then... i kinda hope it's a faulty MB cause if not they should not be on the market.

Be like going to the store to buy 10 banana's and paying for 10 banana's but coming out with 8... you'd go back and either ask for a refund or to get the extra banana's. you shouldn't expect for these chips to hit 5ghz, but they should hit the advertised speeds at the very least.

 

Handbrake is a bastard. I know it very well as I was recently reencoding few hundreds of hours of content into x265. And while in OBS I use my own and thoroughly tested options for all alternative, in Handbrake, I just use multiple settings that are available from UI. (I should add parameter to enforce more threads or just force referencing more frames which easily puts more work on CPU... and usually results in smaller files, but even without optimizing compression, results are 2.5~3 times smaller files than original w/o IQ loss as they were mostly XviD 4.)

And then comes the "problem". With same settings I can get from 60 to 80% CPU utilization on same files. And from practical standpoint sometimes doing manual affinity to only real cores (no SMT) results in better fps encoding and sometimes leaving it on all does better job.
But since Handbrake changes CPU affinity to all cores with each new file in batch...

So I can imagine that in situation when CPU is not fully loaded, there are spare watts for higher boost. But I have not seen your particular state and behavior, therefore I can't really judge all circumstances.
But I know that clock is driven by temperature and 3 power draw limits. And that this 350MHz difference on all cores being loaded by exactly same workload seems like a bug.
And then I would wonder which one is buggy. Is 4.15GHz drawing more power than it should? Or is it that 3.8GHz does not use its power allowance?
I kind of doubt that 350MHz would be result of different temperature. Especially since I expect that you have good cooling.

 

Long distance relationships can be hard , how do you keep the flame burning ?

Lol

 

Make sure your budget motherboard lasts longer than a month.I'm more interested in who is going to overheat more than .100 difference on my clock.

 

Help der8auer get some data on what is really going on and how many people do or dont hit the boost clock speeds.

https://docs.google.com/forms/d/e/1FAIpQLScrtHk-YUlploicyuVpkd9tzwRuE3F1QNfIQHNgL6CNXwpy5w/viewform

 

A "full length" PCI-Express slot is PCIe x16.

PCI Express slots physically come in x1, x4, x8 and x16.

Electrically, a PCIe x1 slot only has 1 "lane". A PCIe x4 slot can have 1, 2 or 4 "lanes". A PCIe x8 slot can have 1, 2, 4 or 8 "lanes". A PCIe x16 slot can have 1, 2, 4, 8 or 16 "lanes."

 

In my experience it's a combination of AGESA version, biosversion, power profile, temperature and how much you have running in the background.

I bought a 3800x to seat in a GiBy x470 gaming 7. I asked for an updated bios so it came with the f41 bios out of the box. To my surprise and delight the boost clocks were above the expected 4500. It boosted right out of the box to 4550mhz. XMP 3600C17 B-die did it's job too, except oc-ing either ram or timings @1.47vRam (max). It's XMP or 'bust' at any voltage.

The tweaker I am, I started testing every new bios to date and bios f42a maxed my boost clocks to 4400Mhz, not a mhz more. The official F42 from last week gave me a max boost clock of ~4500, but only on 'Ryzen high performance plan'. The Ryzen balanced plan option gave a 4400mhz result too.

I still have the best results clockwise with the first AGESA and Zen 2 bios (f41) and just the windows power plans. 4550 all day long. (I lied, it's only microseconds long but hey, you got the point)

Neither the new AGESA nor the new bios fixed the gaming 7's RAM OC capability, so it's an ez choice to go back to f41.

On the other hand, that single core 4550 boost is so rare that you're wondering what program is light enough to profit from it? Probably some lucky idling processes that benefit right before the core/ccx goes in park...

Oh AMD.. how did you screw up your marketing this much? If you advertised it as a 4400mhz chip no-one would cry about it, because the performance is there.. the higher boost clocks would be the cherry on top. But now.. AMD messed up somewhere (or was it AdoredTV)

 

Lol, I gamble it's the first released AGESA /Bios that hits the mark.. the rest sucks? (As is the case with my x470 gaming 7)

 

open air, no direct airflow.
they are relying on just the CPU cooler airflow to cool the VRM's ?
They'd do better to put them in a closed case, like 99% of the people who buy them would do....and vary the airflow thru the case (Hi/Med/Lo)

 

Forget it, I need to read better

I have a x16 slot that runs at x2....but you already stated that....