Intel 9th Gen CPUs revealed

[Kek700]

ACC

 

[Sim Strata]

ran it for you @Andrew_WOT 5.1 with HT and uncore at 4.7ghz

View attachment 285628

Thing is, single threaded performance is not much above my 3770K@4.5ghz which scores just under 29something.

 

[RasmusP]

Funnily when racing in the heavy rain, I can set my 3440 x 1440 to 70% resolution and get 20 fps more without really noticing any meaningful reduction in graphics quality. unfortunately this does not apply to clear weather.

Have I become delusional

I use 83% even for clear weather. It's simple: at your view distance, fullhd would probably still do the job. Similar to my old setup. It was fine but a bit muddy when reading texts.
Now 1080 to 1440 is 0.75 to 1.0 so at 75% resolution scale you're exactly at a full HD view.
But the second clue here is, that only the internal accuracy will be "1080p" but you still keep the finer pixel matrix of the monitor.

Or in other words, but I have no source to prove my statement: it's a difference for our eyes to have let's say a printed picture with a bit too few pixels in front of us or a glowing pixel matrix that has too big gaps between each "light spot". Active and passive light emitting etc are terms that come to my mind.

No idea if that's making sense for you

But 1440p with resolution scale at 75%, aka 1080p looks definitely better than 100% on a 1080p monitor. Because only the actual image is "blurry" but the Pixel matrix your eyes are looking at is fine enough to not cause additional blurriness.

 

[Deleted member 197115]

Yeah, it doesn't work like that. On Ryzen2, it was supposed to be a feature from the software Ryzen Master (the highest freq core was supposed to be always chosen). But as of today windows will chose any core to run @ you max speed, Ryzen or Intel.

So, if you have the time and patience, you will need to test each core individually with prime95, and apply the vcore of the worst performing core to all. To test each core individually with prime95 or lock AC thread to a specific core, just use windows affinity in task manager.

May be I was misinformed then. From what I've read it does assign higher ratio to specific core, not randomly based on core load like TurboBoost.
And this post with quote from ASUS OC Guide seems to confirm that. Again, I could be wrong.
https://rog.asus.com/forum/showthre...locking-changed-the-last-5-6-years#post678987

Per Core: Allows ratios to be applied to each core independently. In this scenario, when non-threaded applications are run, they can be assigned to cores that are running at a higher frequency to improve performance. However, current versions of the Windows operating system are configured to balance loads across all available cores, which results in all available cores reverting to the same ratio as the slowest core when faced with a workload. The workaround is to assign processor affinity for non-threaded workloads manually via the Windows Task Manager.

We recommend using the Sync All Cores setting in association with the AVX Instruction Core ratio Negative Offset setting, or with the ASUS CPU overclocking temperature control features to get the best performance from the Kaby Lake architecture.

 

[Loregan]

Yeah, it doesn't work like that. On Ryzen2, it was supposed to be a feature from the software Ryzen Master (the highest freq core was supposed to be always chosen). But as of today windows will chose any core to run @ you max speed, Ryzen or Intel.

So, if you have the time and patience, you will need to test each core individually with prime95, and apply the vcore of the worst performing core to all. To test each core individually with prime95 or lock AC thread to a specific core, just use windows

Do you use process lasso to save your CPU affinity or there is another way to do it ?

 

[Kek700]

@RasmusP
it makes a lot of sense, at ( 70% ), if i go to a clear day, i am mainly looking way down the road.
Then the lack of pixels density becomes a problem, because i am loosing information,
even if those actual pixels are much smaller and better detailed.
( it just does not look good either )
Close up, ie dashboard and near road , even though i am still loosing information, i have a
large image with fine detailed edges, so the edges are crisp.
( in the wet, at 70% ) the scene in front is full of spray, so the car in front loses its
requirement for detail. and the hazy back ground also does not require the same detail either.

????????????

 

[Dan Costa]

May be I was misinformed then. From what I've read it does assign higher ratio to specific core, not randomly based on core load like TurboBoost.
And this post with quote from ASUS OC Guide seems to confirm that. Again, I could be wrong.
https://rog.asus.com/forum/showthre...locking-changed-the-last-5-6-years#post678987

First time I'm hearing of that for Intel. I'll look into it thanks

Do you use process lasso to save your CPU affinity or there is another way to do it ?

lasso or the affinity option in the default windows task manager. I dont know if you can saved the setting. I'm not using either because I haven't tested each core yet.

 

[RobertR1]

I'm not sure how feasible it is in Intel land to bind processes to cores due to the task scheduler in windows and how intel cpu's behave.

I've seen some very weird behavior with core hopping/swapping of workloads which would give me pause of binding workloads. You'd really want to test this out per game extensively and not just do a set and forget.

I also can't notice any notable performance gain in VR performance in AC/iracing by having a 5.0ghz vs 5.2ghz outside of the 9900k heating up my house of course. Funny enough, enabling HT actually reduced my fps dips which again goes back to the way intel/win10 handle scheduling. If I had to guess the HT comes in when a particular load gets overloaded and the process decides it's faster to use a HT thread then split the the workload onto a diff physical core.

I have all energy savings disabled so it's a not a core parking issue.

 

[RasmusP]

I'm not sure how feasible it is in Intel land to bind processes to cores due to the task scheduler in windows and how intel cpu's behave.

I've seen some very weird behavior with core hopping/swapping of workloads which would give me pause of binding workloads. You'd really want to test this out per game extensively and not just do a set and forget.

I also can't notice any notable performance gain in VR performance in AC/iracing by having a 5.0ghz vs 5.2ghz outside of the 9900k heating up my house of course. Funny enough, enabling HT actually reduced my fps dips which again goes back to the way intel/win10 handle scheduling. If I had to guess the HT comes in when a particular load gets overloaded and the process decides it's faster to use a HT thread then split the the workload onto a diff physical core.

I have all energy savings disabled so it's a not a core parking issue.

Yep, due to the swapping HT gives a bit more headroom. It's just logical like you said.
That's why I'm not sure if forcing processes to specific cores with a slightly higher per core oc will result in more fps than just activating hyperthreading, deactivating core parking, putting all cores to the maximum stable oc and just let it do its thing.
I honestly won't bother with it. Only thing that I'd do would be to put each core to its own specific maximum oc to gain a little overall performance. Not sure if the instability or the difference in the clock speeds would actually slow things down though...

 

[Dan Costa]

There's 2% between 5.2 and 5.1ghz. It could be just enough to keep VR above 90fps or activate ASW 45fps in certain situations. I'll accept any percentage I can get. But to get 5.2ghz stable I can only do it with per-core/load overclock like I suggested earlier and honestly is the way to go.

 

[RobertR1]

Going to leave RAM tuning at this. I can run CL14 but voltage required isn't making me comfortable.

 

[Dan Costa]

Are you testing the ram with HCI Mem test ?

 

[RobertR1]

Intel burn test mainly.

There's a bios bug. That speed should be 3700mhz. I had to roll back to 3600mhz using those timings and it's stable. Will need to see why i'm hitting that bug.

 

[Dan Costa]

IBT is ok for catching cpu and memory controller instability, but to test the ram itself HCI comes rated highly above IBT and even the Bootable memtest86 iso.

The free version has a limit of 2GB of ram, but you can open as many instances as you need until if fills up the ram. Just make sure you stay under your max free ram or else you begin using the page file. Disk usage should be zero while testing for ram.

 

[RobertR1]

https://www.3dmark.com/3dm/32809219

end result of all my tuning.

 

[David Turnbull]

Just got a 9th gen 9600k and overclocked it to 4ghz base clock and 4.8ghz turbo per core, this is what I'm getting with passmark.

Vcore at 1.275v and temps are great with a Corsair H100x cooler

 

[Kek700]

i5 6600k “ 2751 single core score”, i have down graded from 4.7 to 4.6 ghz.
I think if i push voltages, i could get 4.8ghz maybe 4.9, not really worth the
instability risk.?
So there must be some degredation in the score due to that.
thanks for posting, i had thought about changing to i5 9600k, but now
having second thoughts.
At the moment i am GPU and CPU bound on ACC, and processor bound on AC.
This is making an upgrading difficult for me.
i have just changed from gtx980ti to a water cooled gtx1080.
This has not exactly been a major improvement, but a good improvement in ACC.
I will halt it there now, limited fps in ACC to 75 fps which it comfortably achieves
with little variation. ( 3440 x 1440 )
I run AC at 110 fps, again which is also achievable. All at max setting or almost max.
I am at a cross road now i9 9900k and rtx2080ti , now i’m probably into £2000,
Personal circumstance make this a silly amount of money. i’m now done.

 

[RasmusP]

Just got a 9th gen 9600k and overclocked it to 4ghz base clock and 4.8ghz turbo per core, this is what I'm getting with passmark.

View attachment 290183

Vcore at 1.275v and temps are great with a Corsair H100x cooler

You should monitor your clock speeds during the single thread part of the test and then check if locking them all at 4.6 GHz for example and see if you get a better result.
Your score isn't bad at all but I would've thought it would be higher!

 

[David Turnbull]

so i disabled turbo and used the cpu ratio and set to 48x and ran the test again, all cores are deffo getting to 4.8ghz and its only 180 marks behind a 8086k at 5.1ghz in single core performance from earlier in this thread so id think its roughly where it should be for an i5.

 

[Kek700]

i would like to know the real world results from your i5 9600k . This is always the real test of a
system performance. Okay, i realise that when all the varibles add up, trying to get a meaningful
comaparison is difficult.

this is what i get....

In ACC, settings, generally are at max, 10 cars , nurburg GP, 3440 x 1440, i5 6600k , gtx1080
running at 2ghz, 16Gb of 3000 memory. CPU at 4.6Ghz.

90 fps, fluid ( checked fps and it is reasonably constant ( will recheck this again though )
( i do have G sync) , with cpu and gpu running on end stops, gpu 52 deg C , cpu 45 deg C.