Windows 10 update on 16 Mar have partially solved scheduler issue

File: AdR9eqL.png (125KB, 1108x719px) Image search: [Google]

125KB, 1108x719px

>>59495425

Now after the fix

Noice.
18% increase in that benchmark.
Would be nice to see some more relevant items tested and posted.

All that's left is to work on the memory controller. Get that posting with 4Ghz DDR4 RAM, and stable and we're really starting to fly.

>>59495425
lmao poozen can't even hit 144hz in Horizon Zero dawn. Face it, it's

>DOA: The Product

>>59495425
Scheduler might not have been updated, but whatever handles core parking, probably the ACPI stack

>>59495785
That's a playstation exclus-

oh right
bait
lol

Everyone knows the only valid benchmarks are 1080p DX 11 gaymur FPS. Chess? What is that, some shit for nerds? Get outta here with that garbage.

>>59495433
>>59495425
>some blue boxes are filled out in different order than others


AMD's 'fix' everyone

>>59495992
Amusingly, the PS4 is basically a beefed up AMD APU.

AMD is so shit how can anyone deal with them must be retarded and for everyone time is money everyone has no time to waste of fixing AMD crap their busier with their life

>>59496014
>>59496035
You should ask for a raise, shill.

>>59496014

I am afraid of posting anything in these shitty threads because I am afraid to suggest that one can be that dumb and ignorant.

>>59496022
isn't jaguar core weaker than excavator?

File: cpuz_x64_2017-03-20_13-22-39.png (20KB, 469x495px) Image search: [Google]

20KB, 469x495px

I do not know how the fucking HT or Windows scheduler works.

I did some tests here with cpu-z, where in two tests I set the affinity manually and in another I left in the automatic.

First test:

Core 0,1,2 and 3.

File: cpuz_x64_2017-03-20_13-23-38.png (20KB, 469x495px) Image search: [Google]

20KB, 469x495px

>>59496625

Second test:

Core 0,2,4 and 6

File: cpuz_x64_2017-03-20_14-42-18.png (21KB, 469x495px) Image search: [Google]

21KB, 469x495px

>>59496625
>>59496641

Now without affinity. Tests were performed using an Xeon E3-1241 v3.

>>59496525
Yes I think. Jaguar cores are comparable to like K10 cores in IPC I believe but they have less memory bandwidth and lower frequencies. They're have lower power consumption and newer instruction set extensions to make up for it though.

>>59496625

Is it Intel? HT is Intel trademark. If yes then probably they are numbering cores differently: first half of numbers is for corresponding cores and second half of numbers is for complementing logical cores.

File: BillGatesonPhone2.jpg (19KB, 140x140px) Image search: [Google]

19KB, 140x140px

>>59495425
>"activate the update."

>>59496689

Yep, is a Intel. What I am not able to understand is the fact that the test done with the affinity set in the first cores had the score much lower than the test done without defined affinity.

See the results above.

>>59496625
Can you test with process lasso with setting "avoid non-physical cores" enabled?

any benchmarks with new update?

>>59496722

You posted last pic after I hit "Reply". Yep, that shit bugs me.

>>59496828

Use only 2 cores and has the same configuration off my anterior test (core 0,2,4 and 6.)

Test with core 4,5,6 and 7 has the same pontuation off test with core 0,1,2 and 3.

>>59496673
Excavator has barely higher IPC than K10. Jaguar is closer to K9 IPC.

File: 16b.gif (2MB, 425x481px) Image search: [Google]

2MB, 425x481px

>>59495425
>We can't fix the issue in the time but Winblows task manager borders are blue
>Just fill the entire boxes blue

>>59495425
If they fixed the scheduler then it would have fixed Doom and Deus Ex benchmarks which are much better on a 1800X if you disable 4 cores so it's 4+0 than running on all 8 cores.

So where are those benchmarks?

>>59498400
Buy the damn thing yourself if you want instant benchmarks.

>>59498451
Make me, faggot

>>59495425
windows 10 kills drives

>use windows 7
>it werks
wew

>>59495425
>Windows 10 is fine.
- Lisa Su

>>59498451
I'm getting a 1600X.

Considering a brand new, different architecture, blah blah, this has been way, way, way smoother than Pentium D, Core 2 duo, Phenom I, Bulldozer, X99, etc.

The real performance for these CPUs at the price is great. The potential performance is just icing on the cake. Plus an AM4 motherboard that I can use later.

>hyperthreading is shit until AMD does it

>>59498869
>Windows 10 and its scheduler are a huge pile of shi-
>"Hey there now AMD, wouldn't it be awful if none of your products got validated for Windows ever again? And wouldn't it be swell if later down the line at some point a windows update might just increase Ryzen performance for unspecified reasons that have nothing to do with our scheduler?"
>...
>Windows 10 is fine everyone, that's definitely not the problem"

>>59496700
the phone call that saved AMD

>>59498942
It´s CPU Async Compute
AMD Ryzen 8 + 8 (16 Compute)

>>59499114
>Shit, the goyim are going to roll back to Windows 7! We have to stop them.

Thats pretty much what happened. Windows 10 absolutely is handling the two CCXs worse than Windows 7. Its likely because Windows 7 inherited a lot of code for addressing C2D and C2Q chips which was dropped in Windows 10.

>>59499866
>8 + 8
you mean 4x2 + 4x2

>>59495671
The thread juggling between CCX's is still broken, so shit's not fixed yet
There's no reason to disable SMT anymore, that's a great win

>>59496722
The answer is always cache coherence.

>>59495425

Fixing windows in 2017 wow whatever will those mad lads invent next

t. crond

>>59501732
bet this post gets overlooked anon, don't worry at least two people in this thread aren't retarded

>>59495425

This is unfortunate, the only version of 10 worth running is LTSB, which doesn't get these kinds of updates (in fact, running LTSB 2016 since April 2016 has yielded not a single update thus far). So until this fix MAYBE gets rolled into LTSB 2017, Ryzen will remain gimped on the only usable Windows 10 version.

>>59495425
Windows 10 itself is the problem

>>59501895
Wait, I thought LTSB was supposed to get those updates, just not new features.