Share your AIDA 64 cache and memory benchmark here

[Deleted member 185158]

8086 processor is no joke. I've seen quite a few in the benching world doing top clocks all day. I don't expect my 8700K to be as good, but it's pretty decent I think.

5.5ghz now at x4 cores 1.5350v. PiMod which the OS is not set up for 6.63s 1M runs pretty decent, I'd have to install XP for better (W7 for 3D)

So I fired up the rig, going to try this out here. Later I'll come back with some results.
May keep it as a quad for the time being. I'm not in need of thread count in any way. Except 3DMark06 where I'd want all 6 cores (does not scale with HT anyways).

Will start at my max bootable 5.6ghz x4 core. Will lower Core #0 (for posting stability) and ramp up either core #1 or #2 or #4. Will test which core feels best at 5.6ghz while leaving others at 5.2ghz.
If benching, I'll set affinity to the highest core clock for testing. Temps I'm not worried about, always under 70c while benching. Water delta 8.8c today.

BTW my de-lid is re-pasted, not installed with LM, which I have intentions of purchasing. Just feeling out the Xmas stuff here.

Your pointer is spot on, no apologies needed. The screen shot said 1000 words, learned something new!
I remember doing this in the past with AMD chips. Not sure why I hadn't thought about it lol. So good job there.

But yea, thanks again. Trial stages are begun!

Ok On mobile the pc is not on the net at this time so....
Update.

I can only put core 0 at the highest clock. Left vcore at 1.535v to support the clock speed.
I have post!! In windows can do some clicking around.
But getting an issue.
Core 0 set 5.6ghz this gives max boost for Any core it looks like.
5.2, 5.2 and 5.1ghz are the remaining set speeds.
The issue is 3 of 4 cores are boosting at the same time instead of just one core. This is resulting in low volt hang ups 2 so far...

Do I need to enable power saving features? I am definately in performance mode...lol

 

[E-curbi]

8086 processor is no joke. I've seen quite a few in the benching world doing top clocks all day. I don't expect my 8700K to be as good, but it's pretty decent I think.

5.5ghz now at x4 cores 1.5350v. PiMod which the OS is not set up for 6.63s 1M runs pretty decent, I'd have to install XP for better (W7 for 3D)

So I fired up the rig, going to try this out here. Later I'll come back with some results.
May keep it as a quad for the time being. I'm not in need of thread count in any way. Except 3DMark06 where I'd want all 6 cores (does not scale with HT anyways).

Will start at my max bootable 5.6ghz x4 core. Will lower Core #0 (for posting stability) and ramp up either core #1 or #2 or #4. Will test which core feels best at 5.6ghz while leaving others at 5.2ghz.
If benching, I'll set affinity to the highest core clock for testing. Temps I'm not worried about, always under 70c while benching. Water delta 8.8c today.

BTW my de-lid is re-pasted, not installed with LM, which I have intentions of purchasing. Just feeling out the Xmas stuff here.

Your pointer is spot on, no apologies needed. The screen shot said 1000 words, learned something new!
I remember doing this in the past with AMD chips. Not sure why I hadn't thought about it lol. So good job there.

But yea, thanks again. Trial stages are begun!

Sounds like an outstanding plan bro, see you know more about the fine details and subtle nuances of overclocking than I do, lol.

Drop into my 2nd build log Batman Returns and 1st page, you'll find some 8086K 5.6Ghz single-core benchmarks, and some fairly high ST bench values in my System Specs.

I ran all the 5.6Ghz ST benchmarks with a 6C6T profile and not the 5.6Ghz single-core boost. Must turn off my 8086K hyperthreading to boot and run benchies at 5.6.

Although 5.5Ghz runs all day no issues at 6C12T Hyperthreading Enabled...

Single and Slightly-Threaded performance is what I concentrate on since out of all my work apps, the highest threads utilized continuously is only 8, and the 8086K has 12 to offer.

So why not run those 8threads as fast as possible?

Also, in my 2nd build log there's lots of solid build tips and efficiency tweaks for running an inaudible work pc.

Sorry for moving off-topic, we better stop here before we get into trouble. Back to those AIDA benchmarks.

My delid, I cannot take credit for, it was performed by Siliconlottery.com using LM Conductonaut, and he did a fine job, all cores idle within + or - 1C (2C spread) of one another.

 

[Deleted member 185158]

Lol. I dont mind getting in trouble if readers are learning something new today. No offence but the mods can get a little over serious sometimes, but Im sure for a good reason.

I figure if I can run a core or two much higher frequency, I might get better AIDA scores

Ps I did see your 5.6ghz batman adventures.

 

[E-curbi]

Lol. I dont mind getting in trouble if readers are learning something new today. No offence but the mods can get a little over serious sometimes, but Im sure for a good reason.

I figure if I can run a core or two much higher frequency, I might get better AIDA scores

Ps I did see your 5.6ghz batman adventures.

Better AIDA scores are possible. I remember the time when a White Wizard paid a visit to my home office.

Single-Thread Benchmark scores at 5.6Ghz:

Cinebench R15 ST - 249
CPU-Z ST - 676
PassMark CPU ST - 3389

 

[Deleted member 185158]

Thats impressive. That Apex board kicks butt.
I see a lot of subs at HWBot running 12-12-12 4000mhz.
My ram and board combo isnt the best however. Can do cas 14-14-14 at 4000mhz though.

 

[E-curbi]

Thats impressive. That Apex board kicks butt.
I see a lot of subs at HWBot running 12-12-12 4000mhz.
My ram and board combo isnt the best however. Can do cas 14-14-14 at 4000mhz though.

Fairly certain that AIDA score is only an artifact, but who knows maybe it is and maybe it isn't. lol

Those guys/gals as HWbot members certainly - like yourself - have a much broader and deeper overclocking knowledge base to work with than myself.

I’m more into the extreme decibel and silicon and thermal efficiency that perfects a work computer, than the ultimate overclock, although a deep dive into former eventually leads to the latter. lol

I don’t even have a solid grasp of 2ndary and tertiary memory timings. I just use a preset from Luumi the overclocker (as a template) that arrives within the Z390 Dark bios.

One day I’ll spend the 80hours or so needed to fine tune some ddr4 sticks to perfection, I just never can find that time.

My daily work oc is only:

5.4Ghz 6c12t 1.36volts VSA and VCCIO 1.250volts

4400Mhz 16-16-16-34/28 1.500volts

...can run the 8086K at 5.5Ghz 6c12t 1.380v to 1.40volts all day long (remaining inaudible) with slightly higher thermals, yet the 100Mhz increase while working is not so noticeable, and the memory OC and Optane 280GB 10microsecond SSD OS drive super-low latency more than make up for the 100Mhz loss in clock speed.

Daily OCs (below) are not as much fun as benching.

-not certain why HWiNFO64 gives (2) thermal sensor values for the Z390 Dark PCH, the 2nd probably more accurate since the heatsinks are so EXTRA LARGE.

 

[Deleted member 185158]

Nice IMC too. I'm good at 4300mhz same Cas 16-16-16, but run a 1.610v instead of 1.515v.
My board posts vCCIO at 1.410v on auto!! I usually set it 1.25 - 1.30v seems good there.

 

[E-curbi]

Nice IMC too. I'm good at 4300mhz same Cas 16-16-16, but run a 1.610v instead of 1.515v.
My board posts vCCIO at 1.410v on auto!! I usually set it 1.25 - 1.30v seems good there.

I was told/advised to maintain VSA and VCCIO at 1.250v - something about easily shortening the life of the processor if you allow the AUTO settings do as they please. lol

Yea, ran a little experiment with no OC on the CPU and the Gskill 4600Mhz ddr4 sticks clocked up at 4700Mhz, 4800Mhz, and booted yet super unstable at 4900Mhz - so the IMC isn't too bad.

The Gskill 4600Mhz CL18 XMP 2.0 no issue for the IMC, yet 4400Mhz 16CL and 17CL feels much zippier - much more responsive (to me at least) than the XMP - and not by a little bit.

Have also tried 4200 and 4000 with even lower CL and "overall" I'd say for my apps, 4400Mhz 16CL seems to be the sweet spot, although it's NOT the lowest AIDA 64 latency value score.


In other news:

New mouse mat just arrived, moving away from the white and gray input setup to black and gunmetal to blend better with the Batman build.

Setup still looks a bit silly with the white keyboard, yet the black sku to replace won't be available until probably early January. Unless I go with a Japanese proxy.

Still a good start, need to change things up in setup every 3years or so.

Yep, that keyboard is looking TIRED. 3years of extremely heavy (ab)use. lol

The new black sku HHKB Hybrid Type S just announced Dec 10th, 2019, currently only available from PFU-Fujitsu Japan.

8086 processor is no joke. I've seen quite a few in the benching world doing top clocks all day. I don't expect my 8700K to be as good, but it's pretty decent I think.

I can only put core 0 at the highest clock. Left vcore at 1.535v to support the clock speed.
I have post!! In windows can do some clicking around.
But getting an issue.
Core 0 set 5.6ghz this gives max boost for Any core it looks like.
5.2, 5.2 and 5.1ghz are the remaining set speeds.
The issue is 3 of 4 cores are boosting at the same time instead of just one core. This is resulting in low volt hang ups 2 so far...

Do I need to enable power saving features? I am definately in performance mode...lol

That's a lot of voltage, can you easily boot and run a mild to moderate load with a 54-53-52-50-50-50 profile? I'd start low and move in steps.

Also, if you want all the background settings I'm using in the ROG bios just ask, and I'll post them here.

For Win10, I'm running in Balanced Power Mode, not performance or power saving... I get much improved oc results with a fresh OS install on a separate not my work drive - SSD. Fresh (not partially corrupted) bios also can help achieve brand new OCs. I always use a separate bios on the selector, seems to help when you are really pushing your hardware. lol

Yes, all your cores will/should eventually move to your Core 0 (highest) clock speed, yet not simultaneously. I believe the bios rules are setting each remaining core to either an equal or lesser value by 100Mhz.

I would give 54 a go stepping down by 100Mhz the first three core speed values, then drop the last three cores to a much lower value.

Example: 54-53-52-50-50-50.

Then try 55-54-53-51-51-51 etc.

OR try any combination that will boot like 56-50-50-50-50-50, if it gets the job done.

Then if you are good, increase every clock value by 100Mhz and voltage for a second attempt rinse and repeat etc.

If you're still having some difficulty, let me post all the ROG background bios settings for you. I've put them up about (6) times for other enthusiasts, they seem to help in some instances...

 

[Deleted member 185158]

This chip is a big time leaker I think. Loves lots of voltage.
5ghz all core 1.360v ok 5.3ghz we go over 1.4v for any stability.

Ill play around with it. See how it goes.

Btw cas 12 at 4000nhz needs big ram volts

 

[E-curbi]

This chip is a big time leaker I think. Loves lots of voltage.
5ghz all core 1.360v ok 5.3ghz we go over 1.4v for any stability.

Ill play around with it. See how it goes.

Btw cas 12 at 4000nhz needs big ram volts

An engineer buddy of mine told me my specific ddr4 kit would not benefit from any voltage above 1.500v, but then maybe he was only referring to the newer Trident Z B-die version 2.0 PCBs that are more suited for higher clock speeds and not so great at super low latencies.

Have also heard some overclockers who love the 4000 12-12-12 profiles do indeed prefer the earlier version Trident Z PCB layout and go with up to 2.000v dimm voltage.

My kit from my experience 1.6v has made zero difference in oc performance, either high clock attempts or low low latency profiles. Pretty certain my kit is the Trident Z PCB version 2.0.

F4-4600C18D-16GTRS - Overview - G.SKILL International Enterprise Co., Ltd.

Trident Z Royal is the latest addition to the Trident Z flagship family and features a crown jewel design. Meticulously crafted to display just the right amount of light refraction, the patented crystalline light bar scatters the RGB colors in a magnificent display of LED lighting. Mirroring the...

www.gskill.com

Good Luck with your oc's bro!

 

[Xynt]

Hello all

A friend on discord said i could post my aida thing here, so i do.

I have to say that this was the first time I tried overclocking and without the help of said person i would have failed horribly
After he helped me with the walls i encountered, this was the result.

 

[Arctucas]

Fairly certain that AIDA score is only an artifact, but who knows maybe it is and maybe it isn't. lol

<SNIP>

View attachment 139448

Is it repeatable?

<SNIP>

-not certain why HWiNFO64 gives (2) thermal sensor values for the Z390 Dark PCH, the 2nd probably more accurate since the heatsinks are so EXTRA LARGE.

View attachment 139448

The one under eVGA EC seems to move a lot less on my Dark.

 

[E-curbi]

Is it repeatable?
The one under eVGA EC seems to move a lot less on my Dark.

That 20.3 nanosecond latency? It probably is reproducible within the range of instability - I really think it's more of a software artifact and not a true score. I wish it was, although from three engineers I've consulted on the matter, two in Florida one in Texas 20.3ns they all agree is a score still deemed impossible with our current desktops.

It is a real screenshot though, really did happen, check the image with pixel analyzer or whatever those devices are called...
------
I agree with you and the EVGA Z390 Dark thermal data, the EC PCH temp of 29C is the value I believe to be actual. Also amazing, absolutely love the super-cold thermals the board runs at with high CPU clocks both PCH and VRM.

Will be interesting to see what EVGA does with the Z490 Dark next year.

Plan on skipping Z490 and Z495, but will update my platform when Z590 arrives. Will also be watching Asus ROG Maximus 12, 13, and 14 Apex 2dimm boards - and AMD's single-thread performance.

 

[heky]

Played around a bit, this is the result...

 

[E-curbi]

Played around a bit, this is the result...
View attachment 140151View attachment 140152View attachment 140153

Nice scores.

That single thread Ryzen 3900X CPU-Z ST score of 561.7 is impressive at only 4.6Ghz. Ryzen 4000 series may overtake Intel ST next year.

 

[miuen4o]

Got two individual sticks (single rank) e-die recently for my 3700x. Budget went on video and cpu
I'm new to memory overclocking so simply copied the ryzen calculator without playing too much with it. Seems to run stable at 1.47v.
Not sure if I can squeeze more out of them, but I'm pretty happy with 18gb/s improvement on my old Haswell.

 

[miuen4o]

Managed to run it stable at 3733 cl 16. My previous result probably a one off, never went below 68ns the next few benchmarks I ran at 3600. Can't get it to post on 1900 fclk, but I'm guessing I've reached the limit on my 3700x? No idea if I have a bad or a good chip, read mixed stuff online. Some people struggle with 1866 and only run stable on 1833.

 

[Durvelle27]

Managed to run it stable at 3733 cl 16. My previous result probably a one off, never went below 68ns the next few benchmarks I ran at 3600. Can't get it to post on 1900 fclk, but I'm guessing I've reached the limit on my 3700x? No idea if I have a bad or a good chip, read mixed stuff online. Some people struggle with 1866 and only run stable on 1833.

View attachment 140931

Managed to hit 3733MHz myself but at CL18

 

[UsAs]

Old X79 DDR3 QuadChannel still going strong

 

[Deleted member 185158]

Managed to hit 3733MHz myself but at CL18

Memory voltage is too low.

 

[Durvelle27]

Memory voltage is too low.

How so

 

[Deleted member 185158]

How so

I assume cas 18 at 1.35 to 1.40v as gou have manually set.

If going for a quick accomplishment, youd be safe at 1.5v actively cooled.

 

[mb194dc]

Both these are on the same machine at 4ghz, without touching the ram speed. First one is via bios overclock and the second using Ryzen Master 1.0.0.0227 in windows 7. Have tried with newer version of RM, which requite reboot to clock, but don't have the same results. Only on 1.0.0.0227.

Why does memory bandwidth increase so vastly using this old version of Ryzen Master? Also see vast increases in CPU queen and other benches.

 

[smaraux]

Here is a result OCing Corsair Dominator Platinum 2666 C15 into 3200 C16, Hynix M Die (MFR). Not specially good, but was at a discount price (99 euros) ! And dual rank helps. (2x16 GB)
DDR4 voltage should be 1.39V / 1.05 SOC

 

[smaraux]

Here are some tests with Corsair dominator platinium 2666C15 (die Hynix MFR)

Result : OK H24 @ 3200C16
1.36V set & 1.385V Read on memory, 1.05V set on SOC
At 3333 Boot OK, but was not stable (I did not increase the Volts to do better, I intend to sell those)

With 2x8 Go Ballistix LT gray 3200C16, reached 3800C18 with 1.4V set (1.425 read) but 3600C16 is reached with a much lower voltage (1.36, 1.385 read) and was more stable

And with 2x16 Go Ballistix LT gray 3000C15, achieved the same 3600 without worrying, same 1usmus settings than my 2x8 GB 3200C15 kit. Lost some latency, probably won some write (because of dual rank) I am not looking further than 3600 : it starts to icnrease voltage too much, and reaches the limits of the sync with CPU