Discussion in 'Frontpage news' started by Hilbert Hagedoorn, Aug 22, 2019.
clearly yes... at 68 Euro it's even 1 Euro less than my B450
It was most certainly not sustained, but that isn't really the issue nor the point of this investigation. AMD only claim "up to" for their single thread "boost" rating and only for "lightly threaded loads". Most of the new Ryzen CPUs simply cannot maintain their highest clock/voltage state under any kind of meaningful load (yes single thread). I say most as the 3600s seem to fair okay with it's lower rating.
No, this is about some people not even hitting the rated clockspeed at all.
In my testing of a 3700X stock CPU "boost" of "up to" 4400 scores less in single thread than fixed 4300 which should not be happening. If I watch what's going on I can see the core boost high initially but then drop off to lower than where it can run and with much more voltage than necessary. Sometimes scheduling fails to get the load onto one of the "good" cores so it'll never hit the clock in that case.
Another user with first hand experience
I've yet to hit my rated speed either on both my ASUS x570E or my Gigabyte Aorus but I do get way higher clocks out of the Gigabyte. That being said I never had both on the same bios variant so idk.
We can read. It's practically no airflow scenario. Just IR radiation. Put it inside case and have front fans or top fans which ever given PC case have and temperatures will be different.
And there is no argument for Fanless PC case when we talk about $500 CPU. Of someone buys that, they can surely spend few $ on airflow.
I cannot agree more. The only software that benefits from high boost clocks is 'idle processes' because it only hits the boost clocks right before the CCX goes in park. There is no such thing as a sustainable Zen2 boost clock.. my dreams of an overclockable Zen proc went up in smokes (thnx Adored) before launch.. but I thought I got to have a 4500mhz chip all day... compu'ah sais no.
OK, so I did a Cinebench 20 run on my 3800X, and it boosted to 449? MHz, cant remember last digit exactly.
I do remember that it was 44.8x multiplier, which I think would be a better judge of boost, as we may not all have the same base clock.
I have a 99.8MHz base clock, not 100MHz, as far as I can tell.
(edit : quick calc comes up at 4471MHz, but it was 448 or 449 something I saw)
Its not connected to the internet at the moment, with a clean install of Windows and GoG and Steam clients,and a few bench .exes on the desktop.
I left it for a while, watching a film on another PC, came back, and saw in HWInfo 610, that the CPU multiplier had gone up to 45.3x, and the CPU hit 4516.7MHz boost clock, on one core.
It seems it took 1.498V, as that was the max voltage seen.
stock cooler, the wraith LED thing...79.8°C Max.
So, it does boost on mine at least, just weirdly while idling, or whatever chose to run while I wasn't looking .
1804 188.8.131.52 AB Bios, Asus Tuf B450M-PRO Gaming.
Just got the memory at 3800C16, and IF to 1900MHz, thats all i've changed in the Bios, nothing relating to PBO or whatnot (getting the Rad and stuff from the post office tomorrow morning).
edit2: This was done with the previous chipset drivers (AMD_chipset_drivers_nonVGA_1.7.29.0115).
Updating now, and testing CineBench R15 as well)
edit3 : just hit 4391.9 MHz on at least 4-6 cores, unzipping Cinebench to the desktop, using windows unzip thing.
edit4 : One IMPORTANT thing about this Motherboard, the CPU Voltage frequency thing is 300 kHz default, and seems to only go up to 350 kHz (you have to enter a value in a box, and 375/400 failed).
Updated the chipset drivers to 1.8.19.0915, and ran Cinebench R15 single thread (3x) and only hit 44.8x, 4466.6 MHz. 1.4V average, with the odd peak to 1.425V and one very quick peak near the end to 1.456V on the 3rd run.
At 99.8 MHz Base, that still comes out at 4471.04 MHz (44.8 x 99.8).
If it was a 99.7 base clock, that would be 4466.6 MHz.
I'm now wondering if the base clock can be changed.....
edit : I reset the HWInfo clock and data, and left it idling doing nothing, and at 0.6% cpu, it doesn't drop below 3.8GHz on all cores apparently, 38.3x multiplier.
All cores just sitting there.
I would bet all of this is down to the Mobo manufacturers, and their Bios skills with Zen2.
They don't have any. Any of them.
This is their learning curve.
I'm very happy with my CPU, and don't believe AMD are to blame here.
They don't furnish the final Motherboard Bios's to the manufacturers, its up to them to get it right.
Intel CPU arch has increased only incrementally, and they probably have teams at the Mobo places making the Bios's, or at least helping a lot.
Not sure AMD have the cash to do that....
I am bitching at the Mobo makers a bit, but mostly for the default settings that 99.9% are going to use...anyway, rant over, benchs.
So, I can't find base clock anywhere in this Bios, so I decided to enable PBO, and set it to the "Motherboard" settings.
Allowing the Mobo to control the Power limits etc. A bit dicey with everything stock, including the coolers etc, but it is open air, in the garage, where its pretty cool
And i'm only testing the single core R15 for now. (and cracked open a bottle of single malt to get the courage to run the mutlithread bench, it is Friday after all TFI Friday )
edit : Well, opened the bottle and forgot to put the data
With just the PBO-Motherboard setting enabled, the Voltage went up to 1.475V average, dropping sometimes to 1.463-1.450V
Stayed at 4466.6 MHz for quite a while at the beginning of the test.
It likes Cores 0 and 3.
Then they both boosted to 4516.5 MHz when it was each ones turn at least 1-2 seconds at that boost clock, before dropping to 4466.6 again, and then back to 4516.5.
It only did it a few times, maybe 5 during the single core test, that I saw.
I had to blink a bit here and there
4516.6 seems to be the auto limit so far, its not gone above it.
211 cb though, 2 runs.
Quite a big voltage jump for not much, but some boost.
Still with PBO-Motherboard :
full CB R15 run, all cores, it downclocks when the TDie hits 80°C, it maxed at 80.4°C and the fan ramped higher between 79.4 to 80.4°C to bring it back under 80°C.
That was running three all thread runs in a row without any wait time though.
Before, without PBO, I noticed 4.192 MHz all core boost.
With, I got to 4217 MHz all core boost.
Dropped to 4167 MHz when it had to downclock at the end of the 2nd run, and nearing the end of the 3rd run.
2233 cb R15 allcore on first run, with no downclocking.
Now to try R20.
All core cb20 got me 81°C, a very high fan, and 4167 MHz, at 1.325V (vdroop anyone?), then it popped back a bit to 1.337V and stayed there to the end (~15s). I LOL'ed
edit : 4985 CB R20 allcore
edit2 : 509 CB R20 single core
Saw 4516.5 a few times, its a lot longer test to do on single core. I counted almost 7s at that clock sustained on one core, the rest I saw were 2-4s sustained.
Asus rep commented on overclock.net:
Kind of makes me question wanting to overclock these things when ASUS is claiming AMD is concerned about the long term reliability of stock advertised clocks. I don't mind too much on not hitting the advertise clock because overall the CPU is way faster then what I had but it is a little disappointing.
Not a good omen, no.
I'll be under water tomorrow if everythings arrived, so at least on the heat front i'll be safe.
Not at all. Just that those motherboards usually cost three times more than most others.
Only ones I buy.... Still using my Asus REIV and that mobo used today can go for upwards of three hundo.
I think there are many MBs with just single PCIe x16 slot.
And as for @TieSKey 's too many 2x PCIe x8. With PCIe 4.0+ , people may start using PCIe card based dual x4 M.2 sor fast data storage in case they run out of M.2 slots on MB itself.
How prevalent is PCIe bifurcation on the AMD chipsets I wonder...
i think one of the x570 boards has a x8 slot from the chipset, which I think is two x4's anyway.
Motherboards turbo speeds will vary because motherboard manufactures have different voltages MSI and Gigabyte are well renound for overvolting while the likes of ASUS normally stick closely to the CPU specifications imposed by Intel/AMD.
Well, I think he meant primary CPU PCIe lanes for GPU. That practically any larger board has 2 slots that can be used for graphics cards.
And while true, with PCIe 4.0 one can sacrifice 1/2 of bandwidth for GPU and use 2nd slot for something else.
And those PCIe to M.2 cards are rather cheap today. Issue is in finding 2x M.2 to x8 PCIe.
Products like this are kind of rare in good prices.
When looking at the chipset diagrams for the x16 from the CPU, it doesn't look like it can do anything other than x16 or dual X8.
So you might not be able to put a dual M2 card in the second slot.
Thats what I was thinking.
Yes, not easy thing to get exactly as someone likes it.
For PCIe 3.0 there was not big demand in PC world, therefore no products. So there are very few good and reasonably priced products.
I think Supermicro AOC-SHG3-4M2P provides good M.2 slot count for price. And unless someone needs all 4 M.2 drives to pull/push data at same time... Very good.
But what I meant is that with PCIe 4.0+ bandwidths for GPUs will be overkill and people will start to look more often on alternative slot use.
I am sure that prices will go down soonish (tm).