|
Its 2021 why in the good motherfuck are you idiots overclocking poo poo
|
#
?
Feb 9, 2021 18:01
|
|
|
|
| # ? Jun 1, 2024 02:58 |
|
|
Crunchy Black posted:Its 2021 why in the good motherfuck are you idiots overclocking poo poo im sorry some of us are still on 14nm and don't got PBO
|
|
#
?
Feb 9, 2021 18:02
|
|
|
Crunchy Black posted:Its 2021 why in the good motherfuck are you idiots overclocking poo poo RAM unfortunately is still all manual clocks and timings, and a big bottleneck for Zen2/3 so. Also people like to play with their toys, relax.
|
|
#
?
Feb 9, 2021 18:15
|
|
|
Do not underestimate how awful pre-Zen 2 CCX latency is.
|
|
#
?
Feb 9, 2021 18:22
|
|
|
Cygni posted:RAM unfortunately is still all manual clocks and timings, and a big bottleneck for Zen2/3 so. didn't we have someone literally in the last page or two who burned out their memory controller on zen1/zen+? I guess it doesn't matter if you're planning on throwing the chip away in 2 years like a lot of AMD people seem to, but obviously the "common wisdom" on what's safe on AMD's memory controller/infinity fabric as far as voltage is far far too aggressive for long-term safety edit: it was actually three total processors, a person with a 1600 and a 2600 and a person with a 1700 Paul MaudDib fucked around with this message at 18:47 on Feb 9, 2021 |
|
#
?
Feb 9, 2021 18:34
|
|
|
IDK if you're counting me in there but I'm not overclocking* but running the stuff at the timings+voltage on the box. I did bump up the voltage temporarily to see if it'd help but it didn't. * apparently DDR4's max spec speed is 2400 and anything above that is technically overclocked, but all I ask out of my parts is stability at the numbers they put on the box
|
|
#
?
Feb 9, 2021 18:58
|
|
|
Sure but playing with speeds and timings to maximize what you get is not the same as overvolting? Yeah a lot of people do both, but you aren’t going to “burn out” a memory controller by playing with the CAS latency and speed settings.
|
|
#
?
Feb 9, 2021 19:01
|
|
|
Munkeymon posted:IDK if you're counting me in there but I'm not overclocking* but running the stuff at the timings+voltage on the box. I did bump up the voltage temporarily to see if it'd help but it didn't. JEDEC standards actually go to DDR4-3200, but only at CL20, 22 or 24.
|
|
#
?
Feb 9, 2021 19:04
|
|
|
Cygni posted:Sure but playing with speeds and timings to maximize what you get is not the same as overvolting? Yeah a lot of people do both, but you aren’t going to “burn out” a memory controller by playing with the CAS latency and speed settings. well, you don't actually know what your motherboard is doing in the background. It's not uncommon that when you turn on XMP or bump clock speed that the board will "helpfully" apply more voltage to the memory controller/IO for you, so yes, "playing with speeds" can in fact burn out a memory controller even if you do not manually touch voltages. It's not gonna set settings that destroy the chip instantly, but there's a reason it's considered "overclocking" and not done that way from the factory. but basically the first step in all these Ryzen memory overclocking guides is "apply more voltage to the SOC" anyway... Paul MaudDib fucked around with this message at 19:16 on Feb 9, 2021 |
|
#
?
Feb 9, 2021 19:06
|
|
|
Listen here, kids: Number Go Up = Better. I shouldn't need to explain this.
|
|
#
?
Feb 9, 2021 19:06
|
|
|
Paul MaudDib posted:well, you don't actually know what your motherboard is doing in the background. It's not uncommon that when you turn on XMP or bump clock speed that the board will "helpfully" apply more voltage to the memory controller/IO for you, so yes, "playing with speeds" can in fact burn out a memory controller even if you do not manually touch voltages. Ok Paul
|
|
#
?
Feb 9, 2021 19:10
|
|
|
I assume people overvolt a bit to try to juice that last couple of tenths of a percentage point out of the stone, but IDK maybe I'm wrong. I don't have the patience for that stuff so I've never tried it.
|
|
#
?
Feb 9, 2021 19:10
|
|
|
Cygni posted:Ok Paul /shrug and here we are with three chips with dying memory controllers on the last page they're your chips, if it dying in 3 years isn't a big deal then go for it. Just be aware that motherboards are often sneaky bastards about what they're doing in the background to hold your hand and get that overclock (memory in this case) to run stable, and that even manual settings are often not physically accurate (what you get is often higher than what you dial in). Those aren't even AMD-specific things. https://www.gamersnexus.net/guides/3251-raven-ridge-soc-voltage-guidelines-how-to-kill-cpu-with-safe-voltage Paul MaudDib fucked around with this message at 19:23 on Feb 9, 2021 |
|
#
?
Feb 9, 2021 19:17
|
|
|
Paul MaudDib posted:/shrug I don't know if you're including my post in this, but I said my 1700 is still running fine. You're right though that the mobo set the SOC voltage higher automatically when I bumped the memory speed. It certainly could be to an unsustainable level, but I haven't observed any degradation yet.
|
|
#
?
Feb 9, 2021 19:22
|
|
|
Again, speeds and timings don’t “burn out” memory controllers. And yes, some boards these days love to auto-OC, but a) that’s not what I said and b) if they do that whenever you touch a memory speed, it’s going to do that just trying to set your ram to its STOCK memory speeds as most (all?) Zen boards default to 2133mhz. So you will be dealing with that regardless.
|
|
#
?
Feb 9, 2021 19:25
|
|
|
Cygni posted:b) if they do that whenever you touch a memory speed, it’s going to do that just trying to set your ram to its STOCK memory speeds as most (all?) Zen boards default to 2133mhz. So you will be dealing with that regardless. running at JEDEC and running at XMP/DOCP speeds are inherently not the same thing and boards that do not alter settings at 2133/2400 will often alter settings once you apply XMP/DOCP. basically on many boards "applying XMP" is not a discrete act that says "apply this clock speed", it is a threshold at which the board can asset "OK the user wants to overclock", the seal has officially been broken on "overclocking" and non-stock settings can validly be allowed to be applied, and it will then go ahead and toggle lots of other auto-OC settings. Sometimes that's limited to the memory subsystem (just to get that memory overclock to run stable without the user having to manually fiddle 27 different voltages and timings) and other times it is not and it will enable MCE (potentially increasing core voltages) or other stuff. For example Gigabyte has been known to automatically enable MCE without prompting when the user enables XMP: https://www.gamersnexus.net/guides/3268-multi-core-enhancement-and-core-performance-boost-testing in short- stock is stock and strictly out-of-the-box, most boards should run stock voltages unless you manually toggle another setting. Once you enable XMP, it is up to the board and some boards do lots of stuff in the background at the same time. Paul MaudDib fucked around with this message at 19:38 on Feb 9, 2021 |
|
#
?
Feb 9, 2021 19:30
|
|
|
Yes I’m aware but again, the out of the box speeds/voltages are 2133 and 1.2v. Good luck finding a non server stick of 2133/1.2v DDR4 these days, so that means in order to get your gamer ram to run at its stock, rated speeds, you need to touch the speed settings. If that triggers the auto-OC dogshit, you will be dealing with it to get to stock operation. Also I like that you’ve somehow argued yourself into a corner that you are basically saying people should run JEDEC speeds or risk “burning out” their memory controller, I love this.
|
|
#
?
Feb 9, 2021 19:38
|
|
|
Cygni posted:Yes I’m aware but again, the out of the box speeds/voltages are 2133 and 1.2v. Good luck finding a non server stick of 2133/1.2v DDR4 these days, so that means in order to get your gamer ram to run at its stock, rated speeds, you need to touch the speed settings. If that triggers the auto-OC dogshit, you will be dealing with it to get to stock operation. no, virtually no board I’m aware of will attempt to bring gamer ram up at gamer ram clocks. If the board couldn’t actually post at those settings it would be impossible to get into bios and set the relaxed settings. Gamers not knowing about XMP and running their ram at jedec settings is in fact a perpetual source of comedy in PC building forums precisely because you are completely wrong on how this works. Cygni posted:Also I like that you’ve somehow argued yourself into a corner that you are basically saying people should run JEDEC speeds or risk “burning out” their memory controller, I love this. now this is the true spirit of “ok Paul”, you’re plain and simple wrong and you can’t bring yourself to just admit that I’m right on this. Just say it, you don’t understand how XMP works, you don’t understand how BIOS settings work in general. I’ve cited GN multiple times and you just keep right on cruising. Yes, in the specific narrow case that you have verified that applying XMP on your motherboard doesn’t change any voltages, then running XMP doesn’t affect anything. That is true. In practice applying XMP can change other settings including voltage. The stock voltages are often not enough for the memory controller and infinity fabric to remain stable at the increased clocks (the stock voltages are specified for the stock clock limits of the processor) so the motherboard will helpfully apply more vSOC. It can even go beyond that and apply MCE or other “auto-OC” stuff without telling you. I cited a GN article on this. And yes, if it’s changing settings then over time the additional voltage can degrade the memory controller, so “setting XMP speeds” can cause processor damage over time. That’s another reason it’s not done automatically and the user has to manually select XMP. It’s not gonna happen instantly but yes, it’s more voltage than officially specified and it’s not guaranteed to hit industrial level lifespans if you enable XMP, it will degrade faster than normal. None of this is controversial and it’s just weird how people get so @me. Paul MaudDib fucked around with this message at 20:02 on Feb 9, 2021 |
|
#
?
Feb 9, 2021 19:53
|
|
|
Paul you are the worst. You try to be technically right all the time, but then your get hung up on thread anecdata which when called out as wrong forces your back to telling us how you are technically right again. Your mental illness that makes you act like a bad programming loop is so tiresome. Can't you just know you are correct in your own mind and not attempt to force us to agree with you?
|
|
#
?
Feb 9, 2021 20:01
|
|
|
I'm pretty sure the smart-enough engineers at AMD have figured out how not to let voltage that's auto-applied kill their CPUs in any reasonable (especially given the historic lifespan of CPUs) amount of time.
|
|
#
?
Feb 9, 2021 20:12
|
|
|
Crunchy Black posted:I'm pretty sure the smart-enough engineers at AMD have figured out how not to let voltage that's auto-applied kill their CPUs in any reasonable (especially given the historic lifespan of CPUs) amount of time. AMD explicitly tells you they won’t warranty the processor if you overclock memory (which enabling XMP/DOCP is considered overclocking). It’s not because they hate free performance. quote:1. WARNING: Overclocking memory will void any applicable AMD product warranty, even if such overclocking is enabled via AMD hardware and/or software. This may also void warranties offered by the system manufacturer or retailer or motherboard vendor. Users assume all risks and liabilities that may arise out of overclocking memory, including, without limitation, failure of or damage to RAM/hardware, reduced system performance and/or data loss, corruption or vulnerability. GD-112 Again, it’s not gonna kill it in a week but depending on how much voltage it applies then yes, even XMP can cause degradation that will become a problem within some reasonable lifespan (3-5 years) of the processor and require more voltage, backing off clocks, etc. Like I said, you aren't guaranteed to hit industrial lifespans if you overclock. Or take it from Buildzoid, the stuff he's talking about with XMP on higher-frequency kits often triggering motherboards to increase VCCSA or VCCIO basically apply to AMD too (although which voltages they change and which thresholds are dangerous are of course different). https://www.youtube.com/watch?v=HLNk0NNQQ8s&t=509s Paul MaudDib fucked around with this message at 20:27 on Feb 9, 2021 |
|
#
?
Feb 9, 2021 20:14
|
|
|
Crunchy Black posted:I'm pretty sure the smart-enough engineers at AMD have figured out how not to let voltage that's auto-applied kill their CPUs in any reasonable (especially given the historic lifespan of CPUs) amount of time. The voltages aren’t applied by AMD, they’re applied by the motherboard manufacturer and they can easily apply CPU killing voltages.
|
|
#
?
Feb 9, 2021 20:16
|
|
|
so we're asserting what, exactly, here? What is the amount of voltage variance that's acceptible to y'all, then? Basically, Cygni posted:
|
|
#
?
Feb 9, 2021 20:28
|
|
|
My system is applying 1.35V to my current four RAM sticks via XMP. It's been doing that for two years now. I figure I can do that to my new four sticks, to get lower CAS than CL22@3200 (which runs at 1.2V with those JEDEC stock settings), without burning things out. Assuming I can even reach 3200. My current DDR4-3000 sticks only run at 2800, altho I only found yesterday that the chips on the DIMM are only certified for 2133 from the manufacturer and  Corsair is overvolting and -clocking the poo poo out of these. I figure proper RAM chips that are supposed to run at 3200 ought to fare better.
|
|
#
?
Feb 9, 2021 20:33
|
|
|
Combat Pretzel posted:My system is applying 1.35V to my current four RAM sticks via XMP. It's been doing that for two years now. I figure I can do that to my new four sticks, to get lower CAS than CL22@3200 (which runs at 1.2V with those JEDEC stock settings), without burning things out. The memory side isn’t the problem, it’s what might get dialed in on VSOC (or on Intel VCCSA or VCCIO). The memory sticks themselves will run 1.35V happily forever, yeah. And 3200 isn’t going to do anything too harmful, that’s not “fast” these days. But if you’re running a DDR4-3600 kit and it dials in 1.25V Vsoc automatically for you, that might be a problem, depending on the board and whether it’s applying further droop prevention on top of the nominal setting/etc. GN’s advice for AMD seems to be not to let it go above 1.2V or maybe 1.3V and to watch the droop factor on top of that because it can push maybe as much as .1V higher than indicated by the sensors. So I suppose if you want to be maximally safe don’t use Vsoc over 1.1V and that leaves lots of headroom for droop correction/sensor error before you bump into the 1.2V threshold. And going above 1.35V Vccsa on Intel is not the best idea in the world either. Buildzoid cites 1.4V as being a Vccsa that will cause damage in the mid term and I would probably keep it substantially lower than that, probably 1.3V or less. Paul MaudDib fucked around with this message at 20:55 on Feb 9, 2021 |
|
#
?
Feb 9, 2021 20:35
|
|
|
I can't know for sure, because no specifics are being provided, but I feel like my anecdote about my 1600 is being used to argue that Ryzen memory controllers are dying in middle age. I'm not wading into that, but for clarity:
* Disclaimer: That said, I'm not putting the machine back together just to try to reproduce the issue.
|
|
#
?
Feb 9, 2021 20:35
|
|
|
https://cdn.discordapp.com/attachments/782315536822173716/808472449019936848/fps_boost-1.mp4
|
|
#
?
Feb 9, 2021 20:44
|
|
|
Uh, this is a joke, right? Hard to tell nowadays.
|
|
#
?
Feb 9, 2021 20:59
|
|
|
Paul MaudDib posted:no, virtually no board I’m aware of will attempt to bring gamer ram up at gamer ram clocks. If the board couldn’t actually post at those settings it would be impossible to get into bios and set the relaxed settings. I "Ok Paul"ed you explicitly cause i knew you would spin out like this if we kept arguing. You are arguing things that I literally never said? Or are assuming I meant things that I clearly didnt? Again timings and clocks are not going to "burn out" a memory controller. Again, I am aware that setting XMP auto-adjusts voltages on some boards and brands, i said as much. I've seen it myself. I've talked about it on this very board. And again, when you first boot your system, it is going to default to 2133mhz (in most cases), meaning you are going to be setting speeds and timings. If that is enough to trigger auto OC on your board (and on some Giga boards you cant even touch speed settings without first going their their OC page), then you are going to have that issue even when setting your completely stock/approved 2666mhz gamer memory to run at its rated speed. Or if you are going to be turning XMP on, on Asus and others, yup that's going to turn on their AI overclocking even to run the ram at its stock, AMD approved speed settings. I am aware. But overclocking is not overvolting, which is what I said earlier. Yes, this means you should watch the voltages and what your BIOS is doing. No, that doesn't mean I think everyone should shy away from touching their speed settings because it leaves a god deal of performance you paid for on the table. Good god man.
|
|
#
?
Feb 9, 2021 21:13
|
|
|
Is there a good guide anywhere to adjusting your ram timings beyond the XMP rated values? I'd like to experiment a little but would need a guide for newbs on things like what values to change first and how to evaluate the results.
|
|
#
?
Feb 9, 2021 21:21
|
|
|
Has anyone w/ a 5800x tried messing with the clock / voltages manually
|
|
#
?
Feb 9, 2021 21:47
|
|
|
Booyah- posted:Is there a good guide anywhere to adjusting your ram timings beyond the XMP rated values? I'd like to experiment a little but would need a guide for newbs on things like what values to change first and how to evaluate the results. https://github.com/integralfx/MemTestHelper/blob/master/DDR4%20OC%20Guide.md#finding-a-baseline
|
|
#
?
Feb 9, 2021 22:00
|
|
|
Cygni posted:I "Ok Paul"ed you explicitly cause i knew you would spin out like this if we kept arguing. https://www.youtube.com/watch?v=a8qTjTpV1Fg
|
|
#
?
Feb 9, 2021 23:04
|
|
|
I meant I was trying to stop that from happening. Bad word choice. Whatever shutting up now sorry!
|
|
#
?
Feb 10, 2021 00:40
|
|
|
in order to use our ram at its rated speeds we must use settings that destroy it; because some mobo's will apply these setting of their own accord, we must must therefore never use our computers.
|
|
#
?
Feb 10, 2021 01:04
|
|
|
B-1.1.7 Bomber, more like B-52 Bomber posted:in order to use our ram at its rated speeds we must use settings that destroy it; because some mobo's will apply these setting of their own accord, we must must therefore never use our computers.
|
|
#
?
Feb 10, 2021 01:17
|
|
|
GRINDCORE MEGGIDO posted:https://cdn.discordapp.com/attachments/782315536822173716/808472449019936848/fps_boost-1.mp4 AM3 got what it deserved.
|
|
#
?
Feb 10, 2021 02:47
|
|
|
LRADIKAL posted:Paul you are the worst. You try to be technically right all the time, but then your get hung up on thread anecdata which when called out as wrong forces your back to telling us how you are technically right again. Your mental illness that makes you act like a bad programming loop is so tiresome. Can't you just know you are correct in your own mind and not attempt to force us to agree with you? Hey, let's not accuse people of having a mental illness because they are annoying us on a forum. It's not cool and also it's hellworld and we're all dealing with poo poo. This is a thread about processors and motherboards. It's okay to have technical disagreements. It's not okay to endlessly poo poo up the thread and be otherwise annoying. That doesn't seem to be happening here, but if you see someone doing that, please report them and provide some context. Thank you. And may the binning be ever in your favor.
|
|
#
?
Feb 10, 2021 02:50
|
|
|
Was memory tuning a thing even during the DDR3 era? And, for that matter, if it was, what was the chipset support like at the time? Was it similar to now where AMD would let you do it on any AM3 / FM2 board, but you specifically needed a Z or X motherboard for Intel?
|
|
#
?
Feb 10, 2021 03:06
|
|
|
|
| # ? Jun 1, 2024 02:58 |
|
|
gradenko_2000 posted:Was memory tuning a thing even during the DDR3 era? Not to the extent it's now became A Thing with Zen, no. I started overclocking during the Duron pencil days. You were mainly held back by how far you could push your CPU and FSB in some cases, which was almost completely dependent on whether you bought an unlocked chip or not and the capabilities of the mobo. You basically pushed stuff as fast as it could go in a fairly crude way, and then stress-tested. You could really wring a lot of extra performance out of your CPU in ways not possible these days. This whole getting in the RAM timing weeds started with Zen, as far as I can tell.
|
|
#
?
Feb 10, 2021 03:37
|
|