*Official* NBR Desktop Overclocker's Lounge [laptop owners welcome, too]

Yeah, must be something like that. Definitely not a middle-ground test for me.

 

Congratulations, Brother. If you want to join Team @Prema at HWBOT let me know and I will give you a secret password. I thought you had already joined, but I did not see your name in the list.

http://hwbot.org/team/premamod/#Members

 

Cool, congrats, looks great! You had an air cooled GPU before right - did you convert your air cooled GPU to liquid cooling or just buy a new GPU, or did you have a 580 before - can't remember?! GPU temperatures good?

 

Yeah send me a password and I will get on the team. I actually have an account but never did anything with it.

 

Sold the 580, bought a 1080, sold the 1080, and bought a Ti. Then bought the copper cooler from evga. Superposition max temp I have seen is 38c. It's helped my overclock slightly, as I still can't run anything at 2100mhz core. Close, at 2088 though.

Sent from my SM-G935T using Tapatalk

 

2088 is about right. I'm only able to run 2100 on a good day. Most of the time the core drops to 2088.

 

My core is 2025 on air. It starts at 2050 on SP but after a few scenes drops and holds at 2025.

 

OK. I'm just in the infantile stages of my understanding of what's going on here. I'm getting my feet with with first trying to come up with a stable undervolt. I'm not 100% positive of what I'm doing right now, but trying to map what I see in the ASRock's BIOS vs. what you guys say is the first step.

If I'm understanding their UEFI correctly, I can change voltage in the "FIVR Configuration". "Core Voltage Mode" is set to adaptive. I *think* I need to change the "Core Voltage Offset", which "specifies the offset voltage applied to the IA Core domain" with a negative 'millivolt' component. Does that sound about right?

Also, (assuming I've done this correctly), I'm stable with -88 mV setting in Adaptive for an i7-7820X. Anyone have any idea if the silicon for this CPU is going to be a poor performer? I don't know if expected voltages from architecture to architecture can be compared. My guess is they cannot be compared, so I don't know where -85mV falls in the grand scheme of things for the i7-7820X.

Thx

----

Edit - Apparently undervolting in Linux is going to require a lot of research. Setting an undervolt in BIOS and then launching a Linux Mint installer causes it to crash with a 'machine check exception'. Anyone have any suggestions on where to start?

 

Try running it again, but this time leave NV Control Panel setting at stock. and see if your test one score starts to match your test 2 score.

 

Looks really nice, bro.

Sent via PM.

 

It is only there "in case you want to" rerun any of those tests individually instead of running the entire benchmark from start to finish. The scores are still valid and you are not required to rerun them.

 

Well, that test is designed for low end stuff and may not be taxing enough on your system. You can try AFR1 and AFR2 to see if either one helps. If the GPU core is running 1632, the test is too light to cause enough workload for the GPUs to hit boost clocks. Also try setting K-Boost on as well as testing AFR1 and AFR2.

 

I am not sure what is going on. Did you also try AFR1? As I mentioned, this benchmark is for low-end garbage hardware, so it probably does not know how to deal with your monster with SLI, CPU cores up the ying-yang and quad channel memory, LOL. That's probably why your Alienware laptop gets as good or better scores. It actually has to work harder and your desktop is simply getting slightly annoyed by it.

Here is what you read about the benchmark in the 3DMark application:

Description
Use Ice Storm to compare smartphones, tablets and ultra-portable notebooks. You can compare Ice Storm scores from Windows, Android and iOS devices. Ice Storm is a DirectX 11 feature level 9 test.

In the 3DMark Technical Guide, it has this to say about it on page 85:

Ice Storm is a cross-platform benchmark for low cost, basic smartphones and tablets and older mobile devices. We recommend using Time Spy, Fire Strike, and Sky Diver for testing modern PCs. Ice Storm includes two Graphics tests focusing on GPU performance and a Physics test targeting CPU performance. On Android and iOS, Ice Storm uses OpenGL ES 2.0. On Windows, Ice Storm uses a DirectX 11 engine limited to Direct3D feature level 9. Ice Storm's test content, settings and rendering resolution are the same on all platforms and scores can be compared across Windows, Android and iOS.

• Cross-platform benchmark for older mobile devices.
• Includes two Graphics tests and a Physics test.
• Compare scores across Windows, Android and iOS.

 

That's quite a trip Raiderman! Good to see those low temperatures.

One thing that is amazing about Pascal, they often top out at the same clocks, almost regardless of card, 2100Mhz is pretty close to the absolute limit of what is achievable, but a lot of people seem to get there, I can get a max of 2113Mhz, you guys are at 2050 to 2100Mhz top, which is also noteworthy because you have 1080ti with a lot more silicon than my GTX 1070 which traditionally makes the probability of higher overclocks lower yet it's effectively the same as my 2113Mhz (only 3% difference in clock from 2050Mhz to 2113Mhz - nothing). My air cooled GTX 1070 holds 2050Mhz as a sustained gaming boost figure on Default Auto Fans without added voltage, but holds 2075Mhz with added voltage and slightly more aggressive manual fans as a sustained gaming boost figure. I normally just game without adding voltage as it's quieter, pretty much silent, and the difference between 2050Mhz and 2075Mhz is nothing. That's the thing, water cooling would get my card to run at 2113Mhz sustained gaming boost clock, but what's 2113Mhz vs 2050Mhz - significantly less than 3% performance difference in games (because performance doesn't scale equally with core clock), and both solutions are quiet - it's nice to see the aesthetics of a water cooled GPU, and it's satisfying to see the low temperatures and higher boost clocks, but at the end of the day we're only talking about less than 3% performance difference between air cooled and water cooled Pascal GPUs, this generation of NVidia GPUs really doesn't respond to water cooling with a practical advantage. I wonder if future generations of NVidia GPU will respond better to water cooling.

 

http://hwbot.org/submission/3790590_

 

Windows 7 FTW... still the best at almost everything. Not being ugly helps as well.

Here's this... most local servers are slow with higher ping times.

 

I know how, but I got rid of it on mine a week or two ago. It reduced the CPU benchmarks by a little bit having RAID0. The CPU has to manage the RAID array, so it takes a little bit of juice for that. My wPrime and Cinebench scores immediately went back up to where they were before after I stopped using RAID0.

All you have to do is change the BIOS settings and create the RAID0 membership. No need to buy anything.

 

Not my best score, but close...

 

Perhaps 8700K upgrade is next.


The MXM mod (and CPU power delivery mod posted below) are the good work of @Khenglish.

Guys like @Khenglish, @Prema and @Johnksss are heroes.

 

Cool, so GTX 1080 (laptop) power delivery mod, how much TDP are you at now for GTX 1080, can you use it all?

 

After I get the heat sink built and I can find its functional limits, will post more info. I expect it to be identical to the P870DM3 with single 1080. Should be able to push the 1080 to 300W or higher with no problem using the Prema 1080 vBIOS mod. It has the external power and the additional MXM power (which is required for it to post) on the previously unused (non-reference design) pins in the MXM slot that Clevo jerry-rigged to make upgrades a pain in the ass.

The CPU power delivery is also beefed up now. Maybe will be able to supply a full 1.500V (hopefully) and push the 7700K to 5.3 or 5.5GHz. Stock on the left and mod on the right. All of the empty capacitor and FET pads are filled now, as well as 6 additional capacitors (more than a dozen capacitors added). Look closely... many changes. Credit goes to @Khenglish for this.


@Johnksss

 

Wow, 300W, that's a lot for a laptop GPU! Yes, would be good to see how you get along with it once you've got it fixed up.

 

If you recall, I was pulling 300W each with the 1080s in the P870DM3 on both GPUs. In this case, we're only talking about one, so it should be pretty easy.

 

I noticed the EVO 960's under use get quite hot. Stock 45C or so at idle and under CDM benchmark about 59C for just 2 runs of 100mb. So I figured to try some heatsinks. I went passive with https://www.amazon.com/gp/product/B078PBK76J/ref=oh_aui_detailpage_o00_s00?ie=UTF8&psc=1

Under Idle temps were the same but the same CDM run only got too 52c. Now multiple runs of CDM brings it up more so something active would be needed for extreme usage. For normal to low usage these are fine though.

 

Nice! Those are very similar to the NVMe/M.2 heat sinks I made for myself.

I bought these in bulk and cut them to the right length:
https://www.ebay.com/itm/5PCS-Silve...-Heat-Sink-Aluminum-Cooling-Fin-/201673666122

Then I used these rubber watch bands (20mm width) to hold them on:
https://www.ebay.com/itm/Pack-of-10...-Loop-Hoop-Holder-Retainer-Ring-/391549495400

The fins are a little taller (might not fit on some laptops due to that) and they work really excellent. Just add a 0.5mm thermal pad.

Below is a photo of it. The upper NVMe drive (near the CPU) has an ASUS heat sink that came with the mobo. My custom heat sink is below the PCH heat sink.

Spoiler: Large Photo Inside

 

You could do Linux, but there's no real benching or overclocking fun to be had with Linux.

Windows 8.1 can be used for Fire Strike and some other things.

 

Would you need a motherboard to try and do something like this, or will you try looking into a way to make the 8700k compatible with the current board/socket?

 

We are already working on making 8700K compatible with it, just as we @Prema already made 7700K compatible with it. This has already been done, in fact. There are one or two modders out there that have 8700K running on a Sky Lake Z170 motherboard. I just need to source an 8700K for a good price for testing. I am not willing to use the CPU in my desktop for experiments.

 

Forgive my Intel ignorance, but does the 7700K and 8700K use the same socket?

 

These;

 

Se if you can get higher temp with ATTO Disk Benchmark v3.05 (run 2 -3 rounds).

Yeah same, but I think Intel maid some chages with the pin paths for Coffee. Aka not plug and play

 

Speaking of 960 EVO, I think NewEgg has the 512 GB NVMe SSD on sale today for $200.

 

Thats what I got mine for. I think they are 500gb though. At least thats what mine is.

Edit: Evo's are 500gb, Pro's are 512

 

About how much for the heatsink material did @TANWare and @Mr. Fox spend? I found this on newegg.

https://www.newegg.com/Product/Prod...m_re=nvme_heatsink-_-9SIAC8W6SF4126-_-Product

My MB came with some kind of metal cover for M2's
https://www.newegg.com/Product/Prod...x370_gaming_pro_carbon-_-13-144-017-_-Product

 

About the same. But, I got enough material to do about 8 or 10 NVMe heat sinks for that amount of money. Look at the eBay links I posted.

 

Ya, but does it have the cool EK logo on it

 

No. But, the GPU pretty much covers it up anyhow. GPU also covers up the fancy ROG PCH heat sink as well. I know you are just joking around. The red ones are a nice fit for an AMD machine.

 

I clicked you link for the heatsink material, and decided to do a search for M2 heatsinks on ebay. Seems there are many making their own, and selling them for about 7-15 bucks.

 

Have you seen these, with active cooling...lol
https://www.ebay.com/itm/XPG-Storm-...341719&hash=item4d6d172f8f:g:RToAAOSw86Jadjc~

https://www.newegg.com/Product/Prod...01&cm_re=xpg_storm-_-9SIADF16UR8101-_-Product

Oh man, thats really got my geek juices going!!

 

Wow, that's pretty crazy. My 960 Pros do not get hot enough to need active cooling.

 

Ya, but its so damn cool! Which means my arm is being twisted really hard!

Having watercooled GPU's probably helps them to not be so hot also

 

Mine was 10.99 for the pack of two, including bands and thermal pads. A bit more expensive but for intel people; https://www.amazon.com/EKWB-EK-M-2-...834&sr=1-2-spons&keywords=nvme+heatsink&psc=1

For the best passive cooling; https://www.amazon.com/Aluminum-Hea...87&sr=1-25-spons&keywords=nvme+heatsink&psc=1

 

Ya, Newegg has all of the colors, except silver. For 11.99 free shipping

 

Atto was starting at 47c idle after the first run it went to 63c with a back to back run on the second it went to 69c and on a third run it went to 71c then on the fourth run 73c.

 

Yeah, i expected a bit higher temp. ATTO is a good tool to test max ssd temp. It will trigger ssd throttling in thin and flimsy chassis

 

I did not expect CDM to max out the temp. I do know there is just about no way real word I'll ever subject the drives to the torture I just put them under. It is good to know that they seem to handle that torture well though.

 

One round Atto benchmark pushed the Toshiba NVMe ssd's Dell used in Alienware's up to 83C (far outside the ssd's max allowed temp spec ). Aka fully throttling.

 

Samsungs site lists 0-70c operating temp but it seems not to throttle.

 

Some info www.hwinfo.com/forum/Thread-Drive-Temperature-2-on-Samsung-960-Evo

 

Atto Benchmark right before finish. Maxed out 45C with forced fans in my Clevo laptop (Samsung 950 Pro).