*** AMD "Zen" thread (inc AM4/APU discussion) ***

[Panos]

How long before WCCF tech pick this up?

Actually this is old news. I remember a discussion back in 2014 talking about it.
Since W7 the OS runs tests on the underlying hardware to see which hardware is best used for timekeeping. Then makes a decision based on the hardware it found if it will use TSC or HPET or ACPI PM. If TSC is not compatibles with the hardware, uses HPET and if that is not found uses ACPI PM. However HPET & ACPI PM have serious overhead tax on the IPC, hence they are the backup generators.

If somehow MS has some bug in the code where on Ryzen is using HPET or ACPI PM because it cannot see the CPU as TSC compatible one, could be the case of the perf differences.
I remember back then, many reported huge improvements on the FX8350 (8320, even 1090T) by disabling HPET in Windows 7 and 8.
But this is too technical and everyone jumped on the bandwagon "FX8350 sucks".

 

[Safetytrousers]

I've just tried turning HPET on and off and it definitely makes a difference because the difference was repeatable. With it off my performance is slightly worse.
Crosshair VI.

 

[LtMatt]

I've just tried turning HPET on and off and it definitely makes a difference because the difference was repeatable. With it off my performance is slightly worse.
Crosshair VI.

That's what i see. Small performance advantage to having it on, but Firestrike does not pick my run up as invalid like it should when it's off.

 

[labbby]

That's what i see. Small performance advantage to having it on, but Firestrike does not pick my run up as invalid like it should when it's off.

My results are a bit all over the place... when i first turned HPET off i got a small decrease in the firestrike result.. I just tried it again and now i have a tiny gain... tbh its such small margins it could be down to anything such as a program i didn't close down before i benched.

I am still getting a vaild result in firestrike which makes me think HPET is still running.. I have just double checked in cmd and it gets the "error element is not found" which suggests it is off but when i check in device manager HPET is still showing enabled.

i've tried it with disabling through device manager with only a small decrease in performance.

That physics score gav got is crazy.

Matt what kind of physics score, on firestrike, did your 5960x get ?

 

[CEUOTC]

New 1.32 beta BIOS up for MSI B350M Tomahawk on MSI forum.

 

[Spyhop]

@Spyhop, I received the bracket/mount from Thermalright. It's exactly the same as the one that came with the LGM. Haven't tried it as well I can't be bothered to take the Noctua off and see if this one fits properly . May give it a go at some point, if I'm bored. The TY-147B's that came as well, are nice and quiet though.

Cheers. Maybe they didn't get any pre-release Ryzen/AM4 boards and designed their mount off of blueprints alone without actually testing it?!

 

[nashathedog]

These videos should be of interest.

 

[teppic]

Interesting that JayzTwoCents has managed 3200MHz RAM and 4.1GHz after a BIOS + microcode update, when it wouldn't get anywhere near that before.

 

[DragonQ]

Pretty informative videos, thanks for that. Definitely interested in whether the IOMMU groups get patched later on, would be really nice to use that.

 

[gavinh87]

Interesting that JayzTwoCents has managed 3200MHz RAM and 4.1GHz after a BIOS + microcode update, when it wouldn't get anywhere near that before.

What board? I can't see above videos, still at work.

 

[RavenXXX2]

 

[Panos]

FOR THOSE WHO HAVE NOT DISABLED HPET - DO IT NOW!


My gaming performance has rocketed.
Nothing has changed with my setup same clocks same ram the only thing I did was disable HPET.

3Dmark firestrike (now this isn't a verified score due to the timer being disabled but even so.)

Before

3DMark Score
15869
Graphics Score
18738
Physics Score
19904
Combined Score
6471
Graphics Test 1
89.29 fps
Graphics Test 2
74.91 fps
Physics Test
63.19 fps
Combined Test
30.1 fps

After
3DMark Score
20320
Graphics Score
23813
Physics Score
25511
Combined Score
8449
Graphics Test 1
112.93 fps
Graphics Test 2
95.59 fps
Physics Test
80.99 fps
Combined Test
39.3 fps


Cinebench score did not change at all but gaming wise its definitely a difference.
Overwatch is higher too but I did not measure that, just believe me when I say my fps was around 225 for most of the time compared to 190 to what I was getting before.

*I have not tried to replicate this by re-enabling HPET but its worth a look for any gamers.
I have no clue what happened with the firestrike benchmark but the geforce overlay was reporting the exact same fps as the benchmark was. The physics scene was the most impressive for me, usually it hovers around 60 and on a 144hz screen 60 vs 80 fps is noticeable*

If anyone else wants to post their finding that would be helpful to see if HPET does cause a performance hit or its just a case of the Mrs pouring redbull into my computer when I'm not here.

On my i7 6800K by default the HPET is OFF on Windows 10 apparently. By activating it saw no difference on the performance on both Spy & Firestrike.
So turned it off again and run one more Firestrike downclocking the CPU at 4Ghz, for comparison, and here are the results.

http://www.3dmark.com/fs/12112181

Score 15144
Graphics score 17355
Physics 17335
Combined 7055

 

[teppic]

What board? I can't see above videos, still at work.

The CH6.

 

[gavinh87]

On my i7 6800K by default the HPET is OFF on Windows 10 apparently. By activating it saw no difference on the performance on both Spy & Firestrike.
So turned it off again and run one more Firestrike downclocking the CPU at 4Ghz, for comparison, and here are the results.

http://www.3dmark.com/fs/12112181

Score 15144
Graphics score 17355
Physics 17335
Combined 7055

Your results don't show an event timer error indicating that it is still active. I don't know what happened to mine. I'll have another play over the weekend.

 

[CAT-THE-FIFTH]

That is getting close to a certain magical number,lol!!

 

[Panos]

Your results don't show an event timer error indicating that it is still active. I don't know what happened to mine. I'll have another play over the weekend.

On my system apparently the HPET was turned off by default. And is a fresh installation of windows 10. Did it on Sunday when I got the 6800K.

Maybe MS switches it off on intel CPUs? Is there anyone else to confirm?

 

[Nelly]

In this video below by Chew on the Xtreme Systems forum, he's using the ASUS PRIME X370-PRO, multimeter direct from the board is showing 1.4v, but in CPU-Z it's reading between 1.472 to 1.504v at load lol.

Solder pads. Left SOC +/- Right VCORE +/-

Grounds on right side of pads. Positive on left sides of pads.

Soldered to a spare molex extender for ease of use. YMMV.

Source

 

[DarrenM343]

So how's the Beta testing for AMD going guys?(I joke) Is the bug I've been reading about something that happens often or only specific software/uses?
Hoping to make a decision in the next few months, whether to go for an 1800 or wait for the next 8 core from Intel but system needs to be rock steady from day 1 as it'll be a work PC.

 

[Yaayuh!]

So how's the Beta testing for AMD going guys?(I joke) Is the bug I've been reading about something that happens often or only specific software/uses?
Hoping to make a decision in the next few months, whether to go for an 1800 or wait for the next 8 core from Intel but system needs to be rock steady from day 1 as it'll be a work PC.

What bug is that?

 

[DarrenM343]

What bug is that?

Read the second section - "AMD confirms problems with Ryzen chips".
http://www.valuewalk.com/2017/03/amd-stock-short-interest-soars-record-high-ryzen-chip-problems/

I can't find the other article but saw something on a financial site about users having issues with their PC freezing/crashing.