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

Ahaha I love Final reality bench ... my 3dfx memories kick in

 

Interesting, but I still think that x4 bus is a big bottleneck ... you are right also the CPU can pull down. (Lag behind fast gpu’s) .. my crappiest AW is a r15 with i5 ... will check

 

It's not much of a bottleneck. I am not sure where everyone is getting this information from.

I know the AMP is a bit annoying to set up at first....but I've tested AMD Vega 64 and Titan X (Pascal) so both sides of the camp and my performance was within 5% or less of a desktop. All of this using a AW13 R3 with i7-7700HQ CPU. Maybe it bottlenecks people with i5 or older CPU a bit more but shouldn't be drastic (again using external display).

Even a 1080 in TechPowerUp PCI-e scaling saw a delta of about 4% from full on 3.0 x16: https://www.techpowerup.com/review/nvidia-geforce-gtx-1080-pci-express-scaling/24.html

Top of the line current RTX 2080Ti shows a 8% average delta: https://www.techpowerup.com/review/nvidia-geforce-rtx-2080-ti-pci-express-scaling/6.html

My Titan X (Pascal) testing was within 3-5% so the AMP only loses performance based on the PCI-e 3.0 x4 bandwidth and nothing else.

I will compare with Area-51m (9900K) & AW 13 R3 (7700HQ) to see if scaling is improved at all from the extra CPU power.

Kinda wish they wired this thing for X8, then there would be almost no bottleneck.

 

EVGA Precision X1 0.4.8.0 Launched with "Boost Lock" Feature
by Techpowerup.com | Today, EVGA Thursday released the latest version of its Precision X1 software that lets you overclock, monitor, and customize your graphics cards. The latest version 0.4.8.0 adds a particularly useful feature for enthusiasts, which has the potential to become a must-have, called Boost Lock. This toggle locks the graphics card to a specific frequency, and disables boost beyond these clocks.

Power-limit will still be active in this state, the GPU will either hit the specified clock speed or below. The feature is particularly useful for overclockers, as it vastly reduces the uncertainty involved in getting the GPU stay at a desired clock-speed, and truly brings "precision" to the table. One drawback is that the card will no longer return to its idle clock state, leading to an increase in power consumption.

 

That's just the same as locking frequency in Afterburner is it? Or does this stop the automatic lowering of frequency when the GPU heats up every 5 or 10 degC? (I thought that was built into the driver behaviour, and couldn't be bypassed.)

 

I haven’t looked much on the tool (I’m on old 980 with Nvidia Inspector). But I posted it here because I mean it was a proper thread post it Test it.

———————

EK Releases EK-Leak Tester - a Superior Leak Testing Solution
PRESS RELEASE by VSG Today, 16:22
EK the premium liquid cooling gear manufacturer is releasing the EK-Leak Tester. A fast and safe way of testing for possible leaks in your liquid cooling loop. While traditional leak testing works, it requires hours before very small leaks are detected and then things can get messy. The EK-Leak Tester does the job way better.

 

I tested it just now, seems like a waste of time to me...Boost Lock just seemed to lock the GPU to it's default boost clock, not an overclocked boost clock, and also a lower frequency than the GPU would otherwise reach (even when just stock - because the NVidia driver normally allows frequency boosting above the advertised boost clock temperature willing). Yeah, so my card wouldn't go above 1797Mhz core clock when activating boost lock. Maybe I was doing it wrong, but I tried all sorts of permutations, I think it sucks, but if I'm using it wrong then I'd like to hear how to use it properly.

 

Same experience, unless I was doing something wrong, nothing exciting. It was locking my boost clock to the vBIOS "boost" clock which is FAR lower than my cards stable overclocks.

 

Joe from BeardedHardware was testing this version of Precision on his livestream. It sucked, so he quickly uninstalled it and installed an older version.

 

Was bored yesterday

https://www.3dmark.com/pr/129238

 

Nice score!! KingPin?

Debating going and grabbing a bag of ice and going ice bucket challenge on my radiator today to try and push it to around 11K.

 

For sure you will be able to push it 1-2 levels up ... but don’t keep it too much below ambient, condensation is not electronics friend

Ps: yeah is a kingpin board

 

I think I need to go on exotic .. the aircooling/AIO water is not enough by any chance ..

https://www.3dmark.com/pr/129617

My RTX Titan was only about 65 C ... does anybody know how to bypass the power draw (before I hit, most likely a voltage wall)

 

I think the shunt mod is the only way to go with the Titan RTX cards. I don't know of any alternative vbios with higher limits.

https://benchmark.unigine.com/results/rid_3fb7b3771514438ba47bc81b6f519c4d Managed to move up a few tiers. #26 in the world.

 

Stock Radeon VII: https://www.3dmark.com/fs/20052800

Don't mind GPU name, it's in a Alienware Amplifier, Futuremark hasn't fixed their detection issues.

 

Very good I have a GTX 1080 of the KFA2 brand and its maximum speed is 2114, but I don't know what happens in a matter of 5 minutes it goes down to 2088 ... I don't understand it and the temperatures are normal ...

 

What you are seeing is normal behavior. This stupidity has been with us since Pascal and Turing is equally retarded. They also started gimping voltage and blocking/limiting voltage controls with Pascal, and that same stupidity was also inherited by Turing.

Boost clocks are limited based on temperature. This is how people that cool their GPUs with LN2 and dry ice can reach such high clock speeds. There is no way to correct this behavior, even with firmware mods. The only solution is to keep the GPU at or below 50°C for it to maintain advertised maximum boost clocks. Once it reaches 45-50°C it will begin lowering boost clocks and as the temperatures increase the clocks will continue to decrease. Overclocking does not prevent it. The clock stepping is about 12MHz for each step and as the temperature goes up and set trigger point temperatures are met the boost clock will change by 12MHz at each step.

 

Bro, check your EM

 

as always I thank you for the time it takes to answer since others seem to be the most important thing of yours or others and of course here I am nobody .... xd

 

I am glad you are here with us, brother. Here you are relatively new member, but definitely not a nobody. I enjoy our interaction. You are a nice fellow and you do some really great tuning of your CPU and RAM.

I realize there is a language barrier and it takes effort to translate to English. Perhaps including a few words to go along with the screen shots to draw attention to your successful ventures would enhance the engagement of peers.

Considering you have been with us less than 2 years, it seems pretty obvious to me that your rep/like ratio compares very favorable with your post (message) count and others appreciate you a lot as well. You have more likes than posts (which is very positive).

 

Literal first attempt at Port Royal: https://www.3dmark.com/pr/132174

 

Congrats man, great score

 

Thanks. First time in at least a month I could sit down and bench. I honestly didn't think I would score so high on my first attempt. Anyway back to lurking and working on my channel's set. Have a great weekend all!

 

Any love for vintage hardware/extreme overclocking of yesteryears hardware?

 

Hey guys, here is a test video. Please give me feedback (positive or negative). This video is unlisted and it was created for the people in this thread to give me their thoughts. Thank you in advance!

 

Hey Brother Frank. Great video. Nice to finally see your smiling face, friend. Thanks for the mention. Also, nice job on the test video. Talking in front of the camera is a whole lot more challenging than doing a video with dubbed narration, which is why I do not do it often. You seemed really calm and collected, and well composed. When I talk to the camera rather than narrating, I sometimes spend more time thinking and planning what I am going to say than the length of the video.

Go for the chiller, bro. I do not think you will regret it. Maybe Brother @Johnksss and Brother @JoeT44 can prod you a little more. We all love our chillers.

 

Good stuff Sir - keep them coming

 

I forgot to mention (and you asked about this specifically in the video comments); the video and audio quality are very good. What camera are you using?

 

Thanks! I'm using a Panasonic GH5 with a Sigma 18-35mm/Metabones Ultra lens combo. While that combo is widely used on YT, it is somewhat overkill. If I could do it all over again, I would have gone with a slightly cheaper camera, even though I got a great deal on the camera body. I'm going to review this setup along with the kind of computing power it takes to edit 4K video. Haha, we should be doing more benchmarks in Davinci Resolve.

 

Yeah, that video quality was good, the picture quality, with the clarity of the foreground, and the blurring in the background - it looked good. You could maybe plan out what you're gonna say better, maybe not a script that you read, but a mid-detailed outline that you can glance at so you don't lose where you are - and the bonus is that if you're not reading it word for word then it's more natural. Have some clarity about what the purpose of the video is, a specific topic or subject. Kudos to you for starting out with some vids, you're bound to get better at it the more you do it. I don't know what your main goals are for creating the videos, and what you're aiming for in terms of popularity or even money earned, but depending on what you're aiming at I suppose requires more or less learning or dedication/practice - if it's just for fun then just go nuts & do what you want & enjoy (that probably works really well for some people anyway).

The "Winter Coming" bit was a good natural flash.

 

Thanks for the feedback. I am not looking for money or popularity to be quite honest. I am looking to share the voices of the computing enthusiast. Many media outlets cater to the mainstream. I will seek out the tools and some of the secrets of the overclocker and try to shed light in layman's on why we enjoy what we do. As for planning things out, I will stick to bullet points.

Thank you again for your feedback

 

Definitely a good start. I’m looking forward to more.

As @Robbo99999 said, a basic outline or bullet points (aka the Gamer Nexus approach) lets you flow naturally and unscripted which I prefer too.

In the end, the content trumps everything.

 

Sounds good, post them up whenever you've got them!

 

Is anybody testing World of tanks ?

https://wotencore.net/3343_en1

I am wondering what are some of the optimization settings that can be done, as there are a lot of disconnects between scores on the interwebs (ex a 1080Ti with a 5960x for 55k compared to my score bellow)

 

While the "Boost Lock" is a piece of garbage, the voltage curve tool seems to be working better than MSI Afterburner now. I can set the voltage and there seems to be no fluctuation and no PerfCap reasons. And, the GPU returns to idle clocks with no load.

 

You mean you don't get the 13Mhz reduction every time temperature increases by 5 degC, or you're running below 40 degC anyway where that kind of 'temperature throttling' doesn't come into effect?

 

No, I was not referring to that. There is no way to correct that other than keeping the GPU as cold as possible. As retarded as it is, I think the Green Goblin thinks of it as a "feature" and have the cancer that controls that buried deep.

What I was referring to is with most Turing GPU, the vBIOS causes power and voltage limit flags to trigger that show in GPU-Z PerfCaps. The only one (before now) that didn't do that were the Aorus Extreme vBIOS and generally the stock EVGA FTW3 and XC2 vBIOS did not. The other Galax vBIOSes (before this one @Talon found) and even the K|INGP|N vBIOS had those stupid flags. And, so did this better Galax vBIOS until now. Something with this new version of EVGA Precision X1 stopped it on the Galax vBIOS. I still need to test it on the K|INGP|N vBIOS.

 

And you're getting a higher frequency as a result then, and more performance?

 

Haven't benched it yet. But, I expect it will be the same or better. I am not convinced that the PerfCap reasons shown by GPU-Z are 100% accurate on Turing GPUs. If they are accurate, then probably better since those performance limit flags seem to be gone now using EVGA Precision X1. If the flags are not accurate and all this is doing is killing the erroneous PerfCap reasons then it might not change anything. But, I like the behavior regardless. Not being spastic or telling lies to make it appear as though it is spastic is an improvement. This new version of EVGA Precision X1 appears to have the advantages associated with using "L" to lock the voltage using MSI Afterburner except for the fact that it only applies under load and the GPU will downclock at idle, which does not happen when locked with MSI Afterburner. Will report back when I have time to test and find out more detail.

 

That's good behaviour re the downclocking at idle. Do you think I'd be able to run a higher stable frequency by using that "L" lock feature for my aircooled GTX 1070? I know for instance that 2050Mhz at 1.05V is stable, and 2062Mhz at 1.062V - maybe it would be worth locking in 1.062V and 2062Mhz? That's all with the same standard +100Mhz on the core slider, but it can drop to 2037Mhz along with reduced voltage as it heats up.

EDIT: and how do you lock a voltage point on the voltage/freqency curve in Precision X1, got it loaded up now, but can't figure out how to lock it?

 

Yes, I think you probably would be able to run a higher stable frequency with the voltage locked. You'd have to test it to confirm. It worked flawlessly on my 1080 Ti. And, it works great on the 2080 Ti as long as I am using a vBIOS that isn't garbage.

You move the dot slider to where you want it. While it is still selected, press "L" and you will see a new vertical line appear and the voltage will jump up and stay there. The exception is your voltage max. It won't let you set higher voltage that the vBIOS allows. So, if your max voltage is 1.093V, select that dot. Move it up to the boost clock you want, then "L" to lock. The click the apply button in the main MSI AB window to apply the changes. (If you do not the changes will not be applied from the voltage curve tool.) Also, you need to move the voltage slider to the max for this to work. On my 1080 Ti and 2080 Ti if that is set to default the voltage won't lock. Don't worry about any other part of the slider to the left of the dot you select. No need to set a curve. Just set the core clock and voltage you want for the max. If the voltage does not go up, try moving the 1.200V slider on the far right to the top. Then click apply.

I demonstrate how to do it in this video.

 

Thanks man, I'll give that a go. I thought you meant in Precision X1 though rather than MSI Afterburner? I couldn't figure out how to lock in Precision X1 - I tried pressing "L" and Ctrl-L, but still wouldn't do it. I have copied your curve shape though in Precision X1 from your pic and put in the same pattern of peaks & troughs at I guess the operating voltages I wanted to 'lock'. I have managed to replicate the Mhz & voltage I'm after using that method, and I'm testing it now. I'll try MSI Afterburner tomorrow with the lock function, not enough time today, but will finish testing this weird curve I got from copying your "peaks & troughs" lol.

 

Sorry I was not more specific, bro. MSI AB works differently and I thought you were asking about that instead of PX1. The video only applies to MSI AB.

You cannot lock the voltage with Precision X1, but I do not think you need to because it behaves as if it is locked when under load with the latest version of Precision X1. (It does not with versions before v0.4.8.0, so be sure you have the latest.) Boost Lock doesn't lock the voltage, it only locks the boost at the stock lowest level of boost (which is far less than the additional boost clocks that are temperature spastic).

Try copying this. Notice there is only one peak. Ignore the other peaks in the prior screen shot. Only the one on the far right matters and the horizontal voltage slider set to 100. No need to touch any to the left. The dot I moved up to 2205MHz is for 1125mV. Under load it immediately goes up to what that dot is set for on boost and voltage. At idle it goes back to 300MHz on core and 405MHz on memory.

 

Thanks, the downclocking on load is the attractive option of that one. I actually just tested MSI Afterburner (said I wouldn't today because didn't have time, but doing this rather than watching a film, ha) - so far I'm +130Mhz stable on the OC Scanner Test with locked at 1.093V (max 75 degC) - frequency stable during the run at 2100Mhz - that's about 50Mhz more than I've managed before. I'll have to test in 3DMark to see if it's stable. OC Scanner Test passes it with 90% confidence, which is the max confidence level.

 

Hmm, I'm getting the same or slightly lower GPU score in Firestrike & Timespy even though I'm at least 50Mhz higher on the core clock than my previous 'normal everyday' overclock. That's very strange. I've seen the phenomenon of reduced scores but stability when locking to really low voltages & ridiculously high frequencies - but I'm locked at 1.093V and 2100Mhz, that's not in that insanely low voltage/high frequency territory, so that's strange.

EDIT: done some testing with both Precision X1 and MSI Afterburner for locking a voltage (& frequency), and I'm getting something like 2-5% less GPU score when locking it any given voltage. If I just use an offset overclock without locking a voltage I can score higher than a locked voltage even when the locked voltage frequency is higher than that achieved with the conventional overclocking. @Mr. Fox , you might want to check that you're not getting less performance from locking your voltage - you might want to compare it against standard offset overclocking. It may look like you're stable at a higher overclock (on my card it's 50Mhz difference/increase in max stable overclock) when locking the voltage, but I think it's a bug like I described here, and not showing the reality. This seems to be the case for my GTX 1070 at any rate.

 

I have seen that behavior before. I varies by vBIOS for me.

 

Strange, but interesting! Why one vBIOS would display that behaviour and another that doesn't, strange. I'm using a Gigabyte vBIOS on my Zotac card, so it's not the original vBIOS. One curiosity with this vBIOS is that it under reports power consumption by about 50W, whether or not that is related I don't know. But either way, locking the voltage on my card/vBIOS combo results in an extra 50Mhz of overclock stability but less performance. (When I talk about frequency I'm talking about the actual frequency being measured from second to second in GPUz sensors during the testing/benchmarks.) Yeah, so I'm gonna go back to my standard overclock - locking voltage actually lowers performance on this card.

 

Yes, it definitely varies widely by vBIOS. It did on my 1080 Ti as well. You might want to try locking the voltage at different values. Perhaps the voltage it was locked at wasn't ideal for the clock.

I really hate how things are now. Pascal and Turing are both spastic crap due to buggy firmware. The stupid algorithms the Green Goblin is baking into the firmware is rubbish and nonsense. Firmware should be simple and should never have a mind of its. It's there to take orders and follow them explicitly, but not give any.