|
Post by ShrimpBrime on Aug 8, 2020 23:38:51 GMT -5
A bit of work, it'll do the job. 5.1ghz is a good place to be. v-core will also depend on the cache frequency demands. So if it gets a little shakey at 4.9ghz, try 4.7ghz 4.6ghz ect. Might help on therms too.
|
|
|
Post by stormchaser on Aug 9, 2020 13:54:09 GMT -5
Didn't do much in the way of overclocking last night. Spent a lot of time just cleaning up the rest of the build and getting everything neat and ordered. Having a problem with my memory. Keep in mind this is the same memory (and CPU) I've had in this rig before. Except now I cant seem to run xmp mode, which 4133mhz. No matter what I do it reverts to 2133 MHz, and the ram is in the correct slot. I also loaded the OC profile I had saved, which was also done with the same memory, and nothing. Just doesn't want to move from 2133. So this is as far as I got last night... the pump seems very quiet and very powerful. I will upload a video later.
|
|
|
Post by MainFramR on Aug 9, 2020 17:45:14 GMT -5
Its up! W00t!
Wondering what temps were at that voltage even in the short cpuz test.. though the money shot is something much longer...45-60min+.
Do you only have one stick of ram?
|
|
|
Post by stormchaser on Aug 9, 2020 21:46:40 GMT -5
Its up! W00t! Wondering what temps were at that voltage even in the short cpuz test.. though the money shot is something much longer...45-60min+. Do you only have one stick of ram? No it's a dual channel patriot viper c 19 set @ 4133 MHz... Reseating the ram did the trick. Now I can do XMP no problem. Unfortunately, Im having some trouble with this board. It seems to be BSODing even at stock speeds. This board was the one I was talking about earlier. I lost the primary GPU slot and a couple other issues. Now it seems it really need to be RMA'd. Which is a shame because it's only 6 m old. So for now, here is where we are at. And yes mainframeR, the money shot will be an hour or so at AIDA64 torture testAll I can tell you right now about performance of the loop at this point ----> after a baseline torture test for about 20 minutes, with the turbo set to 5GHz on a single core, and other wise stock settings across the board, the highest temperature reached was 118*F (that's about 86*C). Not an accurate representation, but an indicator that this cooling system may get us in the ballpark for 5.1 or 5.2 GHz....(If it doesn't, I'm adding a miniature 80mm alphacool rad in the front area of the case, (near the Noctua fan in the front) I can cut a hole in the back panel and that way I wont be introducing hot air directly into the case. I am confident with both rads I should be able to do 5.2 and That test was run by AIDA64 and for options I selected CPU only. But yeah, it's not very meaningful at this speed. So I hope to have a money shot for your guys a little bit later tonight, if the mobo cooperates
|
|
|
Post by stormchaser on Aug 10, 2020 2:40:15 GMT -5
|
|
|
Post by stormchaser on Aug 10, 2020 3:25:54 GMT -5
All fans dialed in and fine tuned via PWM. Seriously, it's very difficult to tell if the rig is running from 4 feet away. This is a 3.6w fan. If I ever got into extreme overclocking this would be a handy little fan to have around... basically a server grade part with lots of horsepower, but can also be tuned down to near silent, which the way I have it set up right now. Gotta represent! MSI fanboy? Hell yeah!
PWM control in bios. What's crazy is despite being pegged at 100% for a quick torture test, the fans don't even need to spool up very much to hold temps. Granted it's mostly standard settings just a slight OC... working my way up to 5.0GHz but don't hold your breath. Pretty sure I'm going to have to RMA this board... I easily hit 5.1GHz just to test the waters. By no means is this proof of reliability at that clock speed.
|
|
|
Post by MainFramR on Aug 10, 2020 5:29:46 GMT -5
Spoiler tags! Why do the pics have a weird yellow haze to them? Is that a filter you've added?
In Aida64 you'll want to run cpu, fpu, and cache for your stress test. 86C with one core at 5ghz and cpu only is pretty warm (if you didn't touch voltage, raising one core won't make much of a difference at all). Wait until all cores and threads are there and you run the test with the proper options. 1.4v is a lot for 24/7, so be careful.
In looking at the system images, it look like there is a lot air in a couple of those lines or is it an optical illusion?
|
|
|
Post by stormchaser on Aug 10, 2020 6:40:44 GMT -5
Not enough rad. It will heatsoak pretty quickly. Rule of thumb is a single 120 for every 100w. Flow will not help, nor will multiple reservoirs. And don't tell me that it's only a 95w processor, because it's under rated as hell. It's more like 160w. Warning: *Heads will explode* I was having some problems with the motherboard earlier which were preventing me from getting a decent overclock dialed in. Happy to report those issues with the motherboard have been resolved. We can now get down to business. This was actually straight forward enough once I tuned my fans and pump via PWM. There are a total of four 120 mm Noctual fans in this machine. It's rock solid stable at 5.0 GHz, which was my predicted outcome since day one... (and make no mistake, I will make another effort to push it harder from here and go for 5.1 or 5.2) Apparently, this so called "heat soak" problem does not apt pear to be a hindrance in this rig, at least not at this juncture in time. Moreover, we can also determine that there was very little, if any, evidence for heat soak at all because if you look at the temperature chart on the torture test snip, you can see a near perfect plateau along horizontal plane, from start to finish as temps hold at approximately 160*F for the duration of the torture test. Had my cooling system overheated or reached it's limit, the temperature graph would show a progressive increase from the start of the torture test to the completion of the torture test. So it would not be a smooth plateau, it would more closely resemble the leaning tower of Pisa. Next thing to do is test some intensive games and see how that goes. The rig is exceptionally quiet for most everything until you really push it hard. I'm literally two feet away from it and its still difficult to pick up on the standard noise of a rotating fan. Because they (the fans) are all spinning around 500 rpm at idle and ramp up from there. This is by far the most quiet machine I've ever built and ironically the highest clocked machine at that! And I've built a number of PCs using sound dampening material. I have to thank Noctua their absolute genius in consistently crafting the best fans money can buy. Now keep in mind, this 1 hour test run was done with the case sealed. Meaning all side panels in place. Believe it or not, putting the side cover on actually reduced motherboard temps by about 2*F More to come
|
|
|
Post by MainFramR on Aug 10, 2020 8:03:03 GMT -5
Can you list temps in C moving forward? That's how it has been done in PC computing forever. F means nothing to most in this context and I'm too lazy to convert. Lol. Dont be different. Be easy. EDIT: What are you going to do with all that tubing when you get your mobo back and use the primary GPU slot? I wouldn't want that tubing sitting on the backplate of the GPU personally.
|
|
|
Post by stormchaser on Aug 10, 2020 11:58:37 GMT -5
Can you list temps in C moving forward? That's how it has been done in PC computing forever. F means nothing to most in this context and I'm too lazy to convert. Lol. Dont be different. Be easy. EDIT: What are you going to do with all that tubing when you get your mobo back and use the primary GPU slot? I wouldn't want that tubing sitting on the backplate of the GPU personally. Normally I would agree with you on that. The blackplate is no place for coolant lines to rest on. However, the MSI RX 5700 XT is uniquely qualified for this role. The gaming x edition has a completely re-designed back plate that sits at least 1mm further away from the PCB. On top of that, MSI has made the backplate itself much thicker (almost double) with better thermal dissipation properties. This combination is stellar GPU / blackplate temps, regardless of clock speed. In all honestly you can game for hours and actually put your hand on the back plate without getting burned (try that with an RX 580 you might end up in the hospital with 2nd degree burns lol). Matter of fact it runs lukewarm, anywhere from 35 - 40 *C so while it's not best practices, I shouldn't have a problem doing that if needed. EDIT: Yes I have a bad habit of using *F. I'm make more of an effort going forward. I just grew up with *F so it's second nature for me, where I look at *C and I just have no point of reference, really.
|
|
|
Post by stormchaser on Aug 10, 2020 12:13:45 GMT -5
Mainframe... see that black zip tie around the three coolant lines?
I have to be careful I dont kink anything, but I should be able to tighten the lines up right there with the zip tie if needed. So in theory, I should end up with about 1/2" between the GPU backplate and the coolant lines as it stands.
|
|
|
Post by MainFramR on Aug 10, 2020 12:55:08 GMT -5
With everything in PC computing land C, you had to convert when using F, lol. So this should actually be easier for you (and the rest of us) since nothing in PC land is in F in the first place.
|
|
|
Post by stormchaser on Aug 11, 2020 12:19:10 GMT -5
Spoiler tags! Why do the pics have a weird yellow haze to them? Is that a filter you've added? In Aida64 you'll want to run cpu, fpu, and cache for your stress test. 86C with one core at 5ghz and cpu only is pretty warm (if you didn't touch voltage, raising one core won't make much of a difference at all). Wait until all cores and threads are there and you run the test with the proper options. 1.4v is a lot for 24/7, so be careful. In looking at the system images, it look like there is a lot air in a couple of those lines or is it an optical illusion? Yeah the yellow haze is simply the color filter "zeke" from microsoft's W10 photo editing app. I am not concerned about the 1.4v core. That's what I need for reliability otherwise I'm leaving "meat on the bone" so if I roll back on the voltage, I am not going to be able to hit my mark of 5.0GHz. If the processor dies due to overvolt I will simply go out and find another cheap 9600KF to replace it with and continue on my merry way. I am pretty confident that will not be happening and I'll tell you way. Every platform is obviously different, however I have run a number of CPUs well above their max allowable voltage with no ill fated results (including core 2 duo and core 2 quad). In some cases for well over 10 years. While it is true that over-volting my damage the processor in the long run, most people don't realize that it's life expectancy still remains much higher than the next upgrade cycle. So long as you have decent cooling, moderate over volting should not be an issue. It's sort of like the writes to an SSD. Sure, there is a theoretical limit at which it will stop function but in most cases would still take a decade to reach that # for the average consumer. Yet people are still a little paranoid about it. It remains a concern for the general public who see the term "life expectancy" and get nervous like it's going to fail next year. Back on the subject of torture tests. Out of curiosity, I went back to stock settings on the CPU for the moment to see how a stock TDP load would be handled by the cooling system. Granted, I only checked CPU. But it was still pegged at 100% and the fans barley moved off the stop. I just wanted to post this since I've never seen anything like it. Other than from my z820 system, which has one of the best cooling systems ever designed. Im just remarking on the fact that during this stock "test" of sorts, the core temps only went up a couple degrees and the fans hardly went about 20%. It was still essentially inaudible. This is truly insane thermal control and excellent voltage flexibility/efficiency that Intel baked into their coffee lake chips. And yes, you must be seeing an optical illusion (perhaps from the flash?) because all the air is out of the system. I should actually top off both reservoirs today seeing as they are both down to around 70% now that all the air has been bled from the system.
|
|
|
Post by stormchaser on Aug 11, 2020 13:46:16 GMT -5
Best I can do with memory throughput for now...
|
|
|
Post by MainFramR on Aug 11, 2020 18:01:32 GMT -5
We've all run like that. Still a lot for the chip 24/7. The CPU only test in AIDA is weak. Just because your cpu is at 100% doesn't mean its working it hard. Floating point instructions and cache are merely tickled in the cpu only test. It would have been neat to see an hour test at stock and see the difference between that and your overclock (instead of a completely different and random test). If you really want to test out your cooling, run fpu only. Nice job on the mem. Your idle temps are in the 40s?
|
|
|
Post by stormchaser on Aug 12, 2020 0:27:21 GMT -5
We've all run like that. Still a lot for the chip 24/7. The CPU only test in AIDA is weak. Just because your cpu is at 100% doesn't mean its working it hard. Floating point instructions and cache are merely tickled in the cpu only test. It would have been neat to see an hour test at stock and see the difference between that and your overclock (instead of a completely different and random test). If you really want to test out your cooling, run fpu only. Nice job on the mem. Your idle temps are in the 40s? Yeah, but how many people are doing that on a 120 mm radiator? I know CPU only selection is weak, that's why I didn't do a full hour of it in the first place Of course I know types of torture tests handled by the CPU differ, with some generating more heat than others. I was just surprised to see almost no increase in core temps during that particular test. I will try to get you a longer test with all the boxes checked for contrast with the 5.0GHz OC... Interesting times we live it. CPUs have gotten so advanced and refined to the point where overclocking could actually be detrimental. Total paradox, isnt it? Look at Ryzen, for example... Processor runs about 36*C at idle, not much fluctuation here. At this point it's easy enough to get the fans spooled down to about 600 rpm . With the interior sound proofed and the fans spinning at such a low RPM, the case is virtually silent.
|
|
|
Post by stormchaser on Aug 12, 2020 0:42:45 GMT -5
Actually, that core voltage is about spot on for 5GHz.. Here is your test of all three boxes checked on stock CPU settings. You can obviously tell it was nowhere near overwhelming the limits of my cooling system.
|
|
|
Post by MainFramR on Aug 12, 2020 6:12:15 GMT -5
Lol, of course. If it worked at 5ghz it will clearly work at stock! I wanted to see the temp difference, not an epeen thing. Cooling these things at stock isn't really an issue. So, I expect a double thick 120mm to work. You're preaching to the choir to me.
Lol, you listed everything in F again. If you're really going to make an effort, go into each of your applications and flip the temp flag back to the default, Celsius.
|
|
|
Post by stormchaser on Aug 12, 2020 13:11:00 GMT -5
Lol, of course. If it worked at 5ghz it will clearly work at stock! I wanted to see the temp difference, not an epeen thing. Cooling these things at stock isn't really an issue. So, I expect a double thick 120mm to work. You're preaching to the choir to me. Lol, you listed everything in F again. If you're really going to make an effort, go into each of your applications and flip the temp flag back to the default, Celsius. Yeah, I figured as much, but could have worded my reply better. That was my assumption -- That you wanted to see the differential in temps between the two runs. What's insane about this setup is the fact that I can actually run the 120 mm rad fan at less than 300 rpm and still be able to cool the chip without thermal throttling. I tested with 4.6 GHz static overclock (1.200 vcore) and also with stock CPU settings (turbo core = 4.6GHz). You can even do the same on the 5ghz OC but the fan curve has to be aggressive. But at idle, there is no problem, even at that high voltage and clock speed. Must be the low FPI count? Plus the double core rad? Plus very good noctua fan? As for C / F, I don't think I can change at this point. I am going to counseling for the burden my behavior is putting on others.
|
|
|
Post by stormchaser on Aug 13, 2020 1:35:45 GMT -5
I got around to testing GPU performance on a couple resource intensive games earlier this evening. I wanted to see game play performance from a stock vs 5.0GHz OC standpoint. Just to put this in perspective. I tried the exact same thing during the last iteration of this build, which as nearly identical except for an RX 580 (MSI gaming X) GPU and MSI air cooler. With the RX 580 being pushed to the limit in games like BF3 or BF4, it (the high thermals) totally obliterated my chances of being able to game with the side cover closed. There were a number of problems with that build which have been fixed in this "revision". GPU upgrade was very important. This is because the backplate on the RX 580 sits EXACTLY one inch below the CPU cooler and can easily sustain temps upwards of 180*F under full load. Very bad combination (especially in a case like this). The new GPU does not have that problem. I also have two 120 mm fans in place for active cooling on the RAM, waterblock and motherboard. Temps are pretty insane if you ask me... because this screenshot was taken during live gameplay, it is representative and accurate. I did not wait to take the screenshot, I took at simultaneously. Keep in mind this run was done with the CPU at stock settings. I will get the 5.0GHz profile loaded tomorrow and post that screenshot as well. Look towards the bottom and you will see my fan data with most of them at 10-20 % and still able to hold these clocks which seems very impressive to me. Is this about normal temp wise (for BF4)?
Seems both GPU and CPU are pretty close with no real bottleneck to speak of. Although for longer periods of time, I did see the GPU pegged slightly more than the CPU.
|
|
|
Post by MainFramR on Aug 13, 2020 8:00:19 GMT -5
Gpu should be pegged around 99% in most titles (assuming 1080p, ultra and no frame limiting). Are you playing mine craft or something??
Gaming loads in most titles are nothing so im not surprised temps are low. You're still listing F so i have no clue what the temps are. I'm not converting it, lol.
|
|
|
Post by stormchaser on Aug 13, 2020 13:24:32 GMT -5
Gpu should be pegged around 99% in most titles (assuming 1080p, ultra and no frame limiting). Are you playing mine craft or something?? Gaming loads in most titles are nothing so im not surprised temps are low. You're still listing F so i have no clue what the temps are. I'm not converting it, lol. You have access to a calculator or computer, no? It's really easy to translate, don't be overwhelmed by the working out the conversion. No, it was Doom II. I figured that would really push my GPU to the limit, especially with the 800 x 600 resolution going on there. I can confirm that the RX 5700 has officially solved my high case temperatures and is contributing to lower CPU temps as well. MSI really had their ducks in a row when they cooked up the Gaming X GPU. P.S. gaming can put quite a load on the system, matter of fact, some people use games as a torture test and completely forgo the staple programs like AIDA64 and Prime.Well, that pretty much sums it up - another milestone for this build. Funny how all the people who said this wouldn't work had never actually tried it. That's what we call bench racing.
|
|
|
Post by Mr.Scott on Aug 13, 2020 15:23:58 GMT -5
You have access to a calculator or computer, no? It's really easy to translate, don't be overwhelmed by the working out the conversion. Same could be said to you man, as the PC standard has always been in degrees C. It's not too hard to just conform.
|
|
|
Post by MainFramR on Aug 13, 2020 15:28:40 GMT -5
Storm-chaser (...errr NightRanger), Lol, overwhelmed. I could easily do it. But I enjoy busting your balls because you incessantly go against the grain and like to be difficult (it isn't just me, lulz). Gaming can put quite a load on a system, indeed. That said, I've never had a game come close to stress testing (P95 small fft, OCCT, IBT, AIDA w/CPU,fpu,cache) temps. Typically, gaming is AT LEAST 15C less than any stress test (BF4...makes sense... ancient game and alt-tab out will tends to drop temps). If all you use your PC for is gaming, then stress testing with a game would make sense. If you use it for other things, gaming isn't nearly enough and you'll bork on other, mkre stressful situations on the pc. Scotty, lol, It's literally easier for him to use C... but, storm-chaser logic... WTF is bench racing? You're such a weirdo... Now, go hiking and be safe. When you come back... let's see if you're a man of your word (lolololol) and you actually leave TPU after our election in November and the US isn't in anarchy. Nutjob.
|
|
|
Post by ShrimpBrime on Aug 13, 2020 17:02:59 GMT -5
Whoa, let's not throw being classy out the window fellas....
|
|
|
Post by Vinster on Aug 13, 2020 20:57:03 GMT -5
I've had builds that were 100% stable and well temp'd while being on 24/7 and being used for gaming. I've even had system be perfectly fine Crunching WCG (World Community Grid) on 50-75% of the cores available and be just fine.
Toss in Prime95 in a blend or the Intel burn test and in 5 min it was over heating and crashing the OS with some sort of BSOD (Usually vCore).
I'm sorry, but gaming doesn't stress a system like bashing a series of binary calculations and iterations at a CPU and memory like one of the stress test I just noted.
It just doesn't.
I'm happy it's fine for what you want to use it for, as I've had many builds that are perfectly fine for a good OC and 24/7 Gaming. There isn't anything wrong with that. just don't run a 24hr P95 blend test.
Vin
|
|
|
Post by stormchaser on Aug 18, 2020 14:55:38 GMT -5
I've had builds that were 100% stable and well temp'd while being on 24/7 and being used for gaming. I've even had system be perfectly fine Crunching WCG (World Community Grid) on 50-75% of the cores available and be just fine. Toss in Prime95 in a blend or the Intel burn test and in 5 min it was over heating and crashing the OS with some sort of BSOD (Usually vCore). I'm sorry, but gaming doesn't stress a system like bashing a series of binary calculations and iterations at a CPU and memory like one of the stress test I just noted. It just doesn't. I'm happy it's fine for what you want to use it for, as I've had many builds that are perfectly fine for a good OC and 24/7 Gaming. There isn't anything wrong with that. just don't run a 24hr P95 blend test. Vin Great equalizer and good to point out. Now I do think this rig would do slightly better than most home brewed systems due to higher than average fan performance. Do I think 24 hours on prime is feasible given the demands that this system will face in the real world? No, but I will give it a go. Bios is a little tricky as I am learning pre-programmed profiles as opposed to plain tinkering. The board needs to be RMA'd so my airflow though the case is not ideal, but it should still give us a deceit showing.
|
|
|
Post by stormchaser on Aug 18, 2020 15:10:51 GMT -5
Whoa, let's not throw being classy out the window fellas.... "Some people" (as in one in particular) expected it to not work and were a little disappointed when it actually worked, like a Frazier hook to the side of the head, it's "unexpected"... they didn't expect to see the benchmark completed for one hour and didn't expect my "expectations" to line up with reality. Except they did. So one person in particular is a little pissed that it was indeed his "expectation" that was wrong, not mine. Yeah, nice intro, you're really gonna be liked here, being so friendly, non judgmental and all I think I will actually keep my word about the account thing @ TPU. If I am right I will be laughing at you know who. If I am wrong, I will come back with storm_Chaser1 or something. Now, if we can all act our age instead of little children, there is more to be revealed here with this rig and prime 95. Stay tuned for results. As for the RMA, the primary PCIe slot on this MSI MEG Z390 ACE is bad, so I have to have the GPU moved all the way down to the bottom of the motherboard which limits how much of a load I can put on the GPU and effects airflow through the remaining components to some extent.
|
|
|
Post by ShrimpBrime on Aug 18, 2020 15:38:55 GMT -5
Whoa, let's not throw being classy out the window fellas.... "Some people" (as in one in particular) expected it to not work and were a little disappointed when it actually worked, like a Frazier hook to the side of the head, it's "unexpected"... they didn't expect to see the benchmark completed for one hour and didn't expect my "expectations" to line up with reality. Except they did. So one person in particular is a little pissed that it was indeed his "expectation" that was wrong, not mine. Yeah, nice intro, you're really gonna be liked here, being so friendly, non judgmental and all I think I will actually keep my word about the account thing @ TPU. If I am right I will be laughing at you know who. If I am wrong, I will come back with storm_Chaser1 or something. Now, if we can all act our age instead of little children, there is more to be revealed here with this rig and prime 95. Stay tuned for results. As for the RMA, the primary PCIe slot on this MSI MEG Z390 ACE is bad, so I have to have the GPU moved all the way down to the bottom of the motherboard which limits how much of a load I can put on the GPU and effects airflow through the remaining components to some extent. If that's particular to a certain individual from a certain thread from TPU, I'm not aware of either. Not that it matters to me any, just for you guys to have respect for each other at THIS forum here. We are all brothers and team mates, nothing we couldn't iron out here I don't think. Honor and respect. We can all hold a glass up to that!
|
|
|
Post by stormchaser on Aug 18, 2020 15:50:48 GMT -5
I agree, this is a brotherhood, and a positive place to come together... we need to keep it that way. and just for kicks Im going to see if I can do 24 hours of prime with this machine.. this is what I'm looking at right now. This liquid cooling system is good for about 200 watts so I do have a little bit of voltage headroom if I need it. I seem to get a bit more out of offset + override mode vs override (on its own). still pretty new in terms of tuning this processor at the limit.
|
|