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

You're absolutely right on that. A person can get tons of hardware points with old hardware, and have less competition with fewer people benching the older hardware any more. I was speaking mainly about the benching the current CPU rather than the system as a whole. I would have to spend money on cooling if I wanted to go further with the current hardware, or spend money for better CPU if I wanted to continue seeing the numbers (scores, not points) go up. But, there are older GPUs that can be purchased at garage sales, Craigslist, etc. that rack up lots of points by downgrading. I could also downgrade the CPU to an i5 or something like that, but I'm not really interested in downgrading the CPU. Unless I could get it almost for free or something, then I might do that just to rack up some points, and then sell it, or even throw it away if it's not worth enough to bother selling, after I have extracted everything it has to offer in terms of points.

 

This is very true.

I don't think you fully understand.... You are still trying to bench a desktop setup like it's a laptop setup. Which you can if you like, but to expand past your current levels and stay within the Enthusiast league, (No LN2/Dice)(No sure where water chiller falls) You need to take that motherboard and exploit it. You are not limited to just a 8700K. It will bench a variety of cpus if you can find them. That Motherboard will bench any gpu ever made that is set for a PCIe slot. You can get old cards from a lot of places for dirt cheap. If you go and look at some of the scores, You will notice people using newer hardware to revamp old scores with older hardware posted scores. This is the line of thinking that should be used. Because technically. If you aren't using LN2/Cass/SS/LHe/Chilled water on single card benching....You will never catch anyone of real value. And all these side categories that we get the golds in are only because the upper class benchers don't care. That's why you will see easy categories you can get in at the top at. Now if they decided to throw scores in there. You will see how fast you drop. Now take firestrike for example. See how hard that one is to get into the top 10/20/30 with a single card. And that's because that category is worth points on the bot. And all top 20 leaders need to have a valid url. (That was some what a bit of an exaggeration, but you get the idea)

 

A 10+ year habit is not easily broken. But, I am glad to no longer have to deal with those limitations. Kind of nice having more and better options. What you are advocating should definitely be lots cheaper, and that part is a good fit for me at this point. Making obsolescence your friend is smart. It doesn't even need to be any good by today's standards to get hardware points.

 

I've got threadripper 16-core 1950X. I need to do a tear down analysis for my loop as others are getting better temps than I am while using monster amounts more voltage, although I am getting no throttling in power delivery, etc., so am keeping up with many scores of people using faster speeds than I am. But, I've got some ideas on changes to my loop to help, as well as replacing a fan I manhandled, but that was on me. Meanwhile, still betting on Zen 2 with 7nm (not to be confused with Ryzen 2 which is Zen+ on 12nm). But, I posted a roadmap leaked for Intel showing 14nm++ for all of 2018, so we will have 10nm+ versus Zen 2 7nm in 2019. Same socket compatibility will help on Zen, but I'd need to upgrade for the new PCIe features and potentially DDR5 compatibility, and the socket is only promised until around 2020, so maybe Zen 3.

But I pull over 400W on 4050MHz on the CPU alone, and that may only get worse with better speed scaling with AMD chips (like how much Intel chips can pull).

 

Yeah, mental! So, are you into benchmarking & getting the points, tweaking for the fun of it, or do you need to use all your CPU power for work/leisure?

 

Need the power for hobbies and greater multitasking. But, yeah, also have some benches on the bot (same handle).

 

@4050 on my 1950x set to 1.394v with Prime 95 I pull max 328w on the core. This with my EVGA 1000GQ. I noticed in HW monitor the vid drops to 1.350 and Vcore was 1.360 at the lowest. Got too 71C on the CPU and 55C on the VRM.

Overall I am happy, not a silicon lottery winner but it will do. I'll never use up the CPU here, so I doubt I will be looking to upgrade until way down the line when it is cheap to do so.

 

Yeah, I'm pulling close to 425W with 1.2875V @4050. P95 causes too much heat, shooting to the 80s. But, I think part is the block mount for the CPU. But I bought an SL chip for their highest binning, so...

So once I find the heat problem in the tear down, I should do even better.

 

I hear ya!
And besides...with 10 and 20 dollar parts. You can man handle them and not worry about losing your shirt in the process. .

Get an addon AIO that can be fitted on a majority of cards. Get some fans to blow across the cards parts that do not have protection and bench away. Once done. Either keep the card for your next motherboard upgrade or sell it for a few dollars.
Just don't expect to get rich off it though.

Try that with XTU if you want to really find your max watts at that voltage and clock point.

 

BTW, is there a way to use XTU on amd platforms to bench their bench? I thought it was exclusive to Intel chips (although I haven't tried it). I'd love to submit some numbers for XTU on an AMD chip....

 

You know what. I completely forgot about that @ajc9988

 

That is my biggest gripe on using XTU, it is so points heavy, but now that there is a competitor, that part of the points is blocked off by choice of platform. It would still lose to equal core intel chips, but just from a points perspective, it really sucks not being able to get them for the bot.

 

How come ajc is pulling siginificantly more Watts in the same test and same frequency while at a significantly lower voltage? TANWare is running cooler, is that the difference?

 

Apparently then with those SL chips the AMP draw is higher. I am sure if mine were to draw 400w on the core I would be in the 80's as well. Since the AIO is rated at 500w it probably does so but at 100c. Not something I really want to test.

Simplistically if the amp draw is higher then chips internal resistance is lower. Usually then meaning better circuitry paths on those binned chips. That same note then the heat generated is higher.

 

I remove a lot of the power limits and run at level 1 Vdroop, which with Asrock does keep the voltage relatively flat. I also water cool the VRM, meaning that heat is dumped in my loop, which effects cooling capacity. So there are a mix of reasons, not least of which is found in bios settings. You can even turn off tdp throttle or set to a specific desired wattage. Even with that, there are differences in everyone's preferential setup. I have virtualization turned on because of some programs I use. I don't know if tanware does. I also believe part of my issue is the mounting posts were too tight and I didn't get enough pressure to make real good contact between the IHS and the block. This is why I mentioned needing to do a tear down, as with 3x480s, even with one fan not in commission, you should have better temps. But it shouldn't just be the settings and wattage pulled, although that is part of it.

 

Power draw is a function of voltage & frequency & temperature isn't it. To me the voltages & power draws are flipped the wrong way round for you guys, as in TANware should be seeing the higher power draws - mainly due to the voltage difference but this following stuff might explain it....

Ah, so you're intimating that TANWare is getting vdroop when he runs Prime 95, therefore he's not really at a higher voltage than you during that test? That could explain why TANWare draws less power for a 'seemingly' higher voltage. Hey, vdroop is better than brewers droop TANWare! [shrug shoulders]

 

Here are some more 5.3GHz fun runs...

Sky Diver: https://www.3dmark.com/sd/4912852 | http://hwbot.org/submission/3727273_

Ice Storm: https://www.3dmark.com/is/4279997 | http://hwbot.org/submission/3727225_

Ice Storm Extreme: https://www.3dmark.com/is/4280002 | http://hwbot.org/submission/3727227_

Ice Storm Unlimited: https://www.3dmark.com/is/4280008 | http://hwbot.org/submission/3727242_

 

Watts is a mathematical function of Volts x Amps. Even two identical circuits can have different AMP and or Volt draw (droop from total power draw). This is why all components have a +/- or tolerance. With CPU's this is why they are underclocked when stock, it allows a greater successful yield as the chips fall within tolerances without failures. This is also why binning works so well, under all other hardware being the exact same one will clock better than the other under a certain vcore and we do not set AMP limits.

All other things being equal is what SL has to do in order to determine if one chip is better than another. They care about what vcore is needed with GHz set and it can be cooled by their solution. Nothing is limited by AMP or Watt draw.

While I can run 4050 I prefer for everyday use just 4.0 GHz. I almost never like running a chip to the bleeding edge. To be truthful since I do not have anything to fully push all the cores stock speed with XFR runs great too.

 

You and @ajc9988 can run Cinebench R15 with same clocks, but your own power settings, then compare power numbers from Hwinfo Maybe with 4050MHz.

 

Here is mine @4050 & 10 runs of CBR 15 one rapidly after another. Really warm here tonight for December.

 

I really hate how HWBOT web site is so unresponsive later in the evenings when I do most of my benching. Sometimes I have to wait to the next day to see my ranking and scores on new submissions. "Waiting for calculation..." LOL. Anyhoo... here is my highest Aquamark3 run so far.

http://hwbot.org/submission/3727466_

 

Metro Redux Benchmark

 

@Mr. Fox have you given the XOC vBIOS a run with your 1080 Ti? If I remember correctly it removes the power limits and allows up to 1.2v. Might get that 1080 Ti up to a stable 2100Mhz.

http://forum.hwbot.org/showthread.php?t=169488 -- for more information, digging around the forums it seems it works on EVGA GTX 1080 Ti cards as well.

Also for any of those out there that thought Intel was readying up yet another flagship CPU so soon may just be severely disappointed and can go ahead and safely jump on the 8700K train. It should remain the flagship consumer CPU for at least another year. That rumored 8 core z390 chip is nowhere to be seen anywhere on the latest Intel road map. It makes sense if you think about it. Why released a chip that would cut into their HEDT sales (looking at you 7800X) and the 8700K is already selling like hotcakes. Milk milk milk.

http://www.game-debate.com/news/241...eaked-no-new-mainstream-processors-until-2019

 

What is/could be the next exciting step in PSU design/performance?​

Antec has a live-in Rep at JonnyGURU.com like Meaker from Sager here, Antec guy is GI Joe, he was saying (similar to Haswell Sleep States relative to a Haswell-ready power supply a few years ago) that Intel's given an unprecedented 2.5years head's up to the industry that there will be a new sleep mode in 2020 that may mean if you want a new CPU motherboard that's Win11 compatible you'll need a new power supply for it, the type of which may not be compatible with our old stuff, meaning our new stuff right now

Sapphire Ridge

Ruby Ridge, same difference

We're at ATX Specs 2.4 now I think, an intermediate move to 2.5 soon, then the dreaded 2.6 in 2020

With Win se7en already dead in 2020, this is the Poweracolypse we're headed for

Bintel MacroHard gonna ruin the industry over some blink-of-the-eye sleep crap

Head's up / stay tuned --> RubyRidge is coming

Here's a copy/paste, part1 is a reply to warranty claims from clueless people that try to pair old power supply to new motherboard and neither will work

"Yes, people will claim it and not get it. They bought 2.4 or older PSUs, not 2.6. Same with anything else in the world. You don't get future compatibility. So the costs of this is on the DIY-component companies to tell end users they gotta buy new. The Apple-way, ya know?

And for all the other points by you and others here:

- Of course simplifying output voltages is better, but your "dreams" are just CHAOS. Compatibility is of utmost importance.
- Also, your "dreams" are ignoring space on boards, the transfer of costs from one component to the other (You guys do not realize that no one wants to "complicate & cost up" his own products to "simplify & cost-down" someone else's products. All companies will fight like hell to prevent this.

Just to make it as simple as possible:

ATX12V 2.5 is just the early message to the ecosystem of the PC to prepare for ATX12V 2.6/2020 when Intel plans too release 12th gen Sapphire Rapids. They give 2.5 years headstart, which is a first for Intel. They never gave such a long headstart for major changes.
Microsoft (and end users) want to have systems "awakening" as fast as possible. For that they need ASMs, which keep a system in a state similar to "S0" (not S3 - S7) but powering off as many components as possible while the OS is actually still ON with minimal processes/tasks running.

ASMs are Microsoft's " Modern Standby" & Google's " Lucid Sleep" (DuckDuckGo them )
Just the Legacy Standby Model of S0-S7 way does not allow this.
https://docs.microsoft.com/en-us/win...tandby-over-s3
For these "Fast Wake-up" ASM modes power needs to go up FAST. Much faster than before. And the crackpoint is especially T3 for this.

So in 2020 for:
Non-ASM systems (Legacy) systems with Sapphire Rapids Intel wants 100-250 ms T3
And Nicrosoft+Intel want for ASM systems even 100-150 ms T3.
So:
100-500= 300 ms
changes in 2020 to
100-250= 175 ms
and later to
100-150= 125 ms
But Windows 10 already can do the ASMs, so it is now for the ecosystem to release the hardware to make that happen = The age of incompatibility has started. The SOLE purpose of ATX12V 2.5 is to announce the new T1 & T3's 2.6 years in advance to soften the "bang".

And please don't complain to me about this. I am just the guy who can add up 1 + 1 And no, changing even more of the ecosystem by removing DC voltages would cause even more and unnecessary chaos"

 

After looking closer at the vBIOS, I see that I have already tried that one. It bricked my 1080 Ti SC2 and I had to switch to IGFX to flash back the stock vBIOS. Glad I saved the dump of the stock vBIOS.

 

Sorry for continue this OT..

So whats the Intel's evil plan?
even C7 and deeper states are still bugged (with no plain to fix) on many Intel Cpus (also on 8th Gen. CofeeLake) and ErP(S4+S5) fight on many combination of motherboard/PSU, and now in 2,5 years dream about superdeeperduperlowfuckin state and 100% compatibility?
Dream Intel, dream.

 

Coffee have already cut into their HEDT sales. All depends on next move from AMD.

 

Not much has changed since you tested Windoze 10 RTM first time (almost 2.5 years ago). Can you see exactly same difference or has it become even worse?

 

Here's mine:
This was with 3200MHz on the ram and 1.1 SOC.

Edit: Highlighted version

 

Same weird differences in CPU power consumption between yourself and TANware - as in in you'd expect both of your power consumptions to be flipped with each other given the voltages you guys are running at - oh well!

 

As I said, it comes down to different settings in the BIOS, I'm sure. Nothing wrong with that. Also, if you look, I'm water cooling the VRM and those temps (VR Temp) are roughly related to the CPU temps due to water as well as staying relatively low for pushing 33% more wattage comparing package temps.

Also, the two temps I highlighted on the CPU (look at die for the actual temp) refer to the two dies on the chip. But, what I want to do, since the BIOS would kick an error reading on an older version when using HWInfo with the VRM, is to try the VRM block with a 120mm blowing on the main one and the VRM for the ram sharing with the ram fan on that side (which that alone dropped 12-15C on temps on VRM). I think that would remove a significant amount from the loop allowing for higher speeds, but the VRM heatsink that came with the board raised the ram heat on that side. Also, that could pull a fair amount over 400W, so there is a question whether that will effect it. I was hitting 60s and 70s at 3.95 before using the ram fan to cool the VRM sink previously when using the AIO. So, if the 120mm would allow to keep the VRM under load at 70s running 4.2 in this loop, then I may make the switch. I also need to reseat the block, but I mentioned that, for the CPU. I may also use a silicon insulator pad rather than thermal pad between the chokes and the VRM water blocks which may lower the loop temp under load. I also want to try to move the two pumps from a shared pwm splitter (they are powered by the sata power connectors) to see if that will help with control, otherwise I'll need to get a controller and set that up for manual control. The two high speeds (4000 rpm) are the pumps (chassis 3) and ram fans. The near 3000 is the Nactua iPPC3000 fans and the 2200-2300 are the EK Vardars. So, do with that info what you will.

 

Hmm, nevermind.

 

You already had the explanation from TANWare, on W=VxA. So, obviously he is pushing lower amps. That is, in part, controlled by bios settings, as well as his other explanation. You wanted to prove something you can't with our runs. If he ran my settings, both his heat and results would be different. Also, with the thermal pads acting between the chokes and vrm blocks, it removes more heat lowering resistance. This can effect things, but usually lowers wattage as the amps have less to overcome. We could speculate as to every aspect there is, but why?

Also, @Rage Set - I think bitspower just released a monoblock for the Zenith Extreme X399!

 

I was curious about the massive difference in Watts that went against normal logic. I listened to your explanation, but I have doubts, but I don't really want to debate with you about it, which is why I settled for a 'nevermind' short reply, but thanks for the reply though.

 

Why do you doubt my explanation? Except for a VRM throttle event, which the temps on his dispel, what else is there? Bios settings that allow for higher amps. That is logic. This doesn't go against normal logic, you just either will not examine other factors or refuse to accept your eyes. W=VxA is basic physics. Period. So, he is pushing less amps. Now, aside from settings, there is some minutia in the differences in our chips, mine being SL, his not. We all know I remove a lot of the normal power limits in my BIOS settings. So what is there you are not getting?

 

Lordy lordy, nevermind!

 

Funny you deride and belittle the explanation without a proffer of an alternative theory. Do you even...?

 

That's pretty much it, bro. There has been no meaningful change or improvement in W10 performance. It's essentially every bit as screwed up today in terms of performance as it was on launch day. The CPU crippling started with W8.X and here we are today with the same botched up piece of trash OS.

This is simple and straightforward logic and it applies to all CPUs. As voltage goes up, so does the watts. There is a range of stability. Being stable with lower voltage produces less heat, but it often also produces lower benchmark scores. If you take any two CPUs of the same model and similar bin quality, the one pulling more watts (and needing more cooling assistance because of it) will almost always be the one getting higher benchmark scores. Being stable is not always good enough for benching. You can be stable and not be as powerful or efficient. The wildcard in the mix is thermal management. That has to be taken into account as performance increases.

 

@Rage Set - https://bitspower.com/html/product/pro_show.php?products_id=5413

 

Here are a few Benchmarks I did today...I7-8700k @5.0Ghz and GTX 1080ti @2025 core 6000 memory.

Not too bad for a cookie cutter MSi ...Plus I delidded my CPU today and used CLU under the IHS. I used Thermal Grizzley Kryonaut between the heatsink and my Evga closed loop cooler. Dropped my temps from 12-15c...Glad I did this but still have trouble getting to 5.1ghz. Not temp related, It's settings related in the bios. I will eventually figure it out.

 

Few more


Scored P30242 on 3d Mark11
Scored P73463 on 3D Mark Vantage

Couldn't get the screen shots to load on these 2.

 

Something is wrong with your Cpu CBR15 score.

 

@Papusan First time I've ran it. I'm on windows 10 and I hear that it's not very compatible with each other. Is my score too low or something?

 

All too low.
Stock [email protected] all cores more like 1400-1420cb. What clocks? And what ram speed?
Edit. A few scores from stock clocked 8700K in P870TM

 

Running 5.0 ghz at 1.4v
Ram is G-Skill 16bg dual channel 3200mhz overclocked to 3500mhz

I will run it again. Anti virus or something might've been running in the background.

 

Still with AV, the score is all all too low. See my edit above.

 

Anyone here picking up a titan volta?

 

Congrats on the delid & choice of thermal paste, I did exactly the same & with the same pastes (CLU & Kryonaut) - interestingly we both see the same decrease in temperatures with delidding, seems like we both did a good job or an equally bad job (probably the former!). One thing I've noticed from your CPU scores, you don't get near to 50% added performance by adding two more cores to the CPU (6 vs 4 cores) - comparing Physics scores to my own 4 core CPU. I'd still like an 8700K though, and supposedly it's gonna be the best gaming CPU all the way though 2018!

EDIT: could be you could improve CPU & Physics scores by tweaking RAM & CPU cache frequencies/timings. I see you've overclocked your RAM, perhaps your timings are too loose, it doesn't always pay to increase the frequency at the cost of loosening timings, make sure you're not running with a too loose tRFC RAM timing, it's not one of the main timings, but in my testing it makes a big difference if it's too loose (I'm running 240 on tRFC, details in spoiler of my signature).

 

Seems I have ironed out all the issue and will post my 5820k OC + my RX 480 (vBIOS mod to 580) reference mod with liquid cooling.

Will post asap

 

@Papusan ...Ran it again got 1460 on R15 but now I'm locked out of my bios with the CPU debug light on. Greeeeeat! Will post once I get it figured out.