Gadgetory

this is a quick and straightforward piece inspired by a reddit post from about a week ago the reddit post was itself a response to a video where a youtuber claimed to be lowering temperatures and boosting performance on rise in 3000 CPUs by lowering the vcore value in BIOS we never didn't catch that video as it has since been retracted and followed up by the creator and the community with new information even though the original content was too good to be true it was still based on a completely valid idea lowering voltage 50% of the equation for power will theoretically reduce thermals and power load the content ended up indirectly demonstrating some unique AMD rise and 3000 behaviors that we thought would be worth testing ourselves in this video will demonstrate how to know when under volting is working versus not working we'll talk about gains and losses and get some hard numbers for the master and the god-like motherboards and how and the Rison behaves with regard to clock stretching before that this video is brought to you by us and the GN store the best way to support our independent reporting is through store gamers nexus net viewers like you allowed us to recently switch to paying for all of our own flights to product launch events rather than accepting flights from companies this is made possible with your purchases of merch like our GN medium mod mat in stock and shipping now and designed with GP teardown diagrams and grids our 100% custom-made two-tone shirt is also a great way to help and it's currently on sale the shirt uses 95% cotton and 5% to last in for a sporty fit with vibrant colors and was designed entirely by the GN team learn more at the link of the description below or go to store document access dotnet to paraphrase reddit user bossman 90's post which was in response to the original content and a pretty well exclaimed post lowering v core a little bit without being an OC mode locking the clocks locking the voltage and override and everything will cause the requested voltage per core the v ID to increase so this is where it gets kind of interesting the response is that with VI D rising you end up with a result that's nearly the same but lowering v core a lot will like down to one volt will initiate clock stretching and the end result is that monitoring software will report normal clocks so it'll look like it's fine and it'll look like it's stable at the new voltage but in reality the performance will reflect that things are worse and it's just it's similar in GPU overclocking especially with AMD GPU is if you're using well any tool these days where it looks like things are getting better with regard especially to voltage numbers or in the case of GPU overclocking frequency numbers but until you validate with performance it's not necessarily a number you can believe and this doesn't mean that anyone who who does go through under Bolton processes or overclocking on GPUs is doing anything wrong by instantiating worse performance but thinking it's better it just means that there's this really interesting characteristic of the new hardware and the way the software interacts with it and reports on it where you just you can't trust anything anymore until you run some kind of validation test and so for GPU overclocking we use 3d mark to validate that the score is actually going up when the frequency says it's going up and with CPUs you can use something like Cinebench to run through it pretty quickly so ultimately it's up to AMD to explain what their CPUs do and why they do those things what we're attempting to do in this piece is replicate the behavior that was seen originally and then see if we can find a way to undervolt effectively actually have the the benefit that everyone wants to have when they're under vaulting we ran the 3600 and the 3900 acts for this piece we use two motherboards the master and the godlike so we've got gigabyte and msi both represented here for some different bios's and we ran Cinebench our twenty multiple times three times for multi-threaded and then we also ran sync single threaded passes with Cinebench our twenty note again our twenty not ar-15s the numbers are not comparable between the two and then additionally we did this stock we set for a second test one volt and bios we also did a voltage offsets with a negative offset for things like for example - point zero five volts to see if we could under both it successfully and then we also ran blender and logs the frequency but today we're gonna focus on the cinder bench numbers because it's plenty for now and it tells the story just fine and we're all tired and none of us have rested yet so we controlled the core and the test that called for it we apply at XMP we did our custom timing for all of our CP reviews these days and we applied the DRM voltages as necessary we also max out the fan speeds as we always do pump speeds all that stuff and we kept the ambient temperatures and cooling as constant as reasonable plus or minus about 1 degree Celsius for the ambient temperature from a 21 baseline otherwise we use the motherboard Auto settings for both of the boards so that means no LLC overrides nothing like that to flatten out the voltage and the boards will behave differently based on how the manufacturers configured them so all of that aside let's go through the numbers here's our first two charts we'll just take turns putting them on the screen and multi-threaded and single-threaded results for the r5 3600 it's immediately obvious that applying 1 volt V core tank to the performance the system boots fine and it doesn't show any signs of instability but on both motherboards there was a very real decrease in performance at stock the godlike performed slightly better than the master and the best undervolt offset we could manage didn't significantly change either board score 1.0 is a ridiculously low voltage to run verizon 3000 cpus but both boards and both cpus allowed us to set and boot at this voltage at least it looks like that so now we've established the truth which is that setting a V core this low will in fact degrade performance even if that degradation doesn't come in a form as obvious as a typical blue screen it does happen so this is that unique behavior we were talking about where traditionally you set a voltage to low and there's a pretty simple outcome it blue screens it doesn't boot you get a black screen something that's extremely obvious that it hasn't worked so it would be easy to get excited about this result and we did - when we first started doing this testing but of course by the time we had started we already knew what the outcome would be so that's why we also tested voltage offset shown in the next chart pushing an offset lower than - point zero five on the X 570 master actually did cause instability and boot problems this was encouraging because it meant that we were hitting real limits rather than fooling ourselves by setting an unrealistically low manual v core so there was a chance it might actually help performance with that offset it didn't on the 3600 but the 3900 acts was a different story the third 900 acts like the 3600 performed much worse than stock with a V core set to 1 volt the master actually outperformed the godlike board here at stock as opposed to the test with the 3600 so the trend of rising 3,000 related test variants continues the most interesting part of these results however are the scores with a negative voltage offset the godlike board was stable with an offset as low as minus 0.1 volt using the 3900 acts and it gains 2% over the stock multi-threaded score it's not a huge leave but it's a real improvement and the same behavior was shown on the x5 Sony master single-threaded scores were largely unaffected and actually decreased slightly on the godlike board with the voltage offset so the minus 0.05 offset for the 3900 acts on the master is the best combination we've tried there's an unholy number of charts that we could generate from the data we've gathered so we'll just focus on the msi godlike board for most of this piece just because the graphs for it are a little easier to read and they all tell the same story these next charts are all frequency plots we've also zoomed the frequency charts to a 2,000 mega Hertz range in order to make the difference is more obvious so we have intentionally set a nonzero axis here this first one shows the r5 3600 on the godlike at 1 volt stock and at minus 0.05 volts offset from just these numbers it would appear that the 1 volt V core are the best numbers as the r5 3600 how they constant 4.2 gigahertz throughout the three multi-threaded passes as well as the lawn single-threaded passes the next chart shows the same type of data except that the r9 3900 X is used instead on the 3900 X clocks were less consistent but still reported holding higher clocks then stock with a reported 42 25 during the multi-threaded tests and a single-threaded tests it logged boosts to 4.5 gigahertz and beyond but the stock results went out here with a fairly consistent 45 75 mega Hertz for most of the single threaded test it's clear that the 1 volt V core results isn't actually performing better on either CPU even though it kind of sounds like it might be this is a simply shown by rough comparison of how much longer the test took them to complete which you can see by looking at this chart where the line takes longer or by looking earlier scores these next two charts show V Corps as logged by hardware info it's admittedly not the best way to do things it'd be better to measure in ml/cc cap or something like that but this is good enough the 1 volt bio setting works according to software logging although the next batch of charts might contradict that at the very least this proves that bios as an option is doing something even if the end result isn't what we want which is lower performance here as an unwanted by-product stock feed core for the 3,600 shown here and this chart was reported at about one point three nine volts under all core load or about one point three four four volts with a negative offset applied both fluctuated between one point three eight one point four volts in single threaded tests the 3900 acts showed the inverse and we can pop that one up to the stock and offset results both managed to stick to about one point two seven volts in the all core load and a gap showed itself instead and the single threaded pass with stock v core 18 one point four seven one point four eight and the offset voltage holding at one point three nine these charts show v ID instead that's the next one the average core v ID numbers stack up differently than the reported v core it's the 3600 set to one volt that requires the highest voltages during the e multi-threaded test at about one point four versus one point three six one point three seven stock and offset moving into the single threaded test has the offset results averaging the highest of eid with the 3900 x chart there's a more clear division during the multi-threaded test with the stock the lowest at one point two three volts than the offset at one point three two then the one volt results at one point three four to one point three five single threaded test aden push them closer together with all the results landing in the one point four seven one point four nine range as mentioned earlier we've chosen not to include the results we've gathered on the XY of so many master board but they back up these results that we've seen on the godlike so far for our final charts we've stuck some results from boards together because the points is simple t die the temperature of the die is vastly lower on both the our five thirty six hundred and the our 939 hundred acts with v core set to one volt and as the previous charts have shown a clocks are higher but doesn't mean the performance is better much like GPU overclocking must to be validated with performance tests CPU volt frequency changes at least with these CPUs have to be validated with performance a lot of the time in the past and also with most of the Intel processors it's pretty easy to just see if it blue screens or not and that's your validation of if it worked but it's not so easy this time it's very easy to trick yourself into thinking that all the numbers are better but actually sacrifice performance by accident the only other notables are the 3600 with a minus 0.05 volt offset on the X 570 master which was slightly cooler than the stock setting in multi-threaded testing with an extremely small bump to its average score the 3900 acts to the point one volts offset on the msi god-like ended up much cooler than stock and single threaded testing but with a minor performance degradation sorry conclusion is that yes obviously there's a pretty easily reproducible bug we suppose AMD might call it a feature but we're setting V core too low on the rise in 3000 CPUs will cause a significant decrease in performance while hardware info appears to report better everything better at least better clocks and temperatures and this all does make logical sense there's every reason that I mean you lower the voltage which of course the temperatures gonna go down you lower the voltage of course the power consumption should go down and the frequencies should stabilize on these new CPUs because they're so bound by temperature as we demonstrated in our Allen to positive 84 to minus 80 or so degree testing so everything here is logical it's just that once you actually test it the scoring is does not match what you would want it to it's worse in most cases so what we also saw is it's possible to tweak the voltage down on our 3900 X anyway for lower voltages than what you might run otherwise you can get some minor benefits and score and temperature which causes the score benefit things like that we were also able to run as we showed in our original review a higher all core clock of about 4.3 gigahertz at just one point three four to one point three five volts which is really pretty damn good so it's going to depend CPU to CPU this test and obviously doesn't show you every CPU and then there gonna have their own unique characteristics per unit not just per model or SKU so when leaving everything else stock setting a small negative V core offset seems like it helped more and yielded more predictable behavior than just setting a straight flat V core number a manual v core override the benefits buried by CPU by motherboard my workload but in the very least it was an interesting item to follow up on we had a lot of you asking about it we've been completely buried the last few days and haven't had a chance to look over any other content so we appreciate you giving us a heads up because at this point viewers are like our boots on the ground because we're completely buried and stuff to do but anyway yeah it's it's a valid idea and there is a way to kind of get under volting to work but just make sure you validate each time and like we said earlier this you know it's very easy to end up in a situation where things look like they improved so there's absolutely no no harm and ending up in that situation and we will give a quick shout to I believe it was the optimum tech we didn't get to catch the original video we only heard about the content but you know huge kudos as always for following up quickly and posting the new information there's absolutely no shame with new hardware to discover something that you think is good and then I actually it didn't work out the way you thought it did because this stuff's new and spoiler alert CPUs are extremely complex so it's cool it's cool to see the whole community poking at these things with sticks and throwing rocks at them and seeing what happens because that's what we have to do to understand today CPU at this point they're very complicated these days especially as they begin to adopt more and more GPU like behaviors anyway thanks for watching subscribe for more go to strode I cater sexist on that that helps head directly or patreon.com slash gamers Nexus we'll see you all next time