5800x3d wont boost up to 4.5

[No_dice]

hello,

 

i bought a 5800x3d recently and did not really have the time to fiddle with the settings up untill now.

i tested the chip in cinebench "stock" and it did not boost higher than 4.1 with max temps of 72°C

 

i tried undervolting the cpu with an offset of -25, and it does boost higher but it will not get past 4.25Ghz , alltough my temps are 56°C and wattage is about 60-65W

 

something seems off, im not sure what is wrong here as i am far from a tech wizz. 

i enabled rezisable bar, enabled PBO and enabled DOCP. did not touch other settings apart from trying the negative offset for undervolting.

 

i am probably missing something but i tought it would boost higher, even on on stock settings.

Not sure what to think here, is it the motherboard that prevents the chip to run at full speed, or was i very unlucky to get a bad chip.

 

specs:

 

Asus Tuf plus B550 (latest bios and chipset drivers)

5800x3d

32GB 3200 Ram

Arctic freezer 2 (240)

Rm750x PSU

RX 6800XT

 

 

 

[Shimejii]

All core workloads are going to be in the 4.1-4.3 ghz, and when gaming it should be 4450, unless you do something more in bios.

 

These cpus pretty much wont ever hit 4550 at stock, maybe for a second as soon as you boot up but besides then it doesnt get that high.

[Somerandomtechyboi]

https://www.tomshardware.com/news/overclocking-bug-5800x3d

 

Just use whatever oc software your board got (maybe an older version incase bug has been patched) and manually set multiplier and voltage, just refrain from going much past 1.35v though thermals will be the main limiter

 

On non x3d 4.7-4.8ghz allcore is about whats expected so if you can run around 1.3v under load with a medium llc (set like 1.32 or 1.34v) without hitting thermal limits you can expect to hit the same

[WereCat]

17 minutes ago, Somerandomtechyboi said:

https://www.tomshardware.com/news/overclocking-bug-5800x3d

 

Just use whatever oc software your board got (maybe an older version incase bug has been patched) and manually set multiplier and voltage, just refrain from going much past 1.35v though thermals will be the main limiter

 

Overclocking on 5800x3D is locked, you can't change multiplier or voltage.

 

Something is definitely wrong there. I can confirm that my CPU does boost to 4.45GHz with curve optimizer offset set to -20 and uses around 90-120W.

 

I would suggest resetting the BIOS settings to default and checking what kind of power plan you run in Windows, updating Chipset drivers as well.

 

EDIT:

It's possible the latest BIOS uses the new AGESA version, I've read they have fixed the EDC bug and is now set to 120 by default instead of 140 so you may be running in some sort of power limit if it's indeed EDC = 120. I'm on previous version as I did not find a reason to update yet.

[Somerandomtechyboi]

9 minutes ago, WereCat said:

Overclocking on 5800x3D is locked, you can't change multiplier or voltage.

The article says otherwise atleast voltage wise

 

might aswell give it a try, i mean youll problably hit thermal limits quite abit before 1.35v so its safe anyways

[WereCat]

11 minutes ago, Somerandomtechyboi said:

The article says otherwise atleast voltage wise

 

might aswell give it a try, i mean youll problably hit thermal limits quite abit before 1.35v so its safe anyways

You can't. You can use Curve Optimizer to move the frequency / voltage curve but it has limit at -30 and there is no reason to use positive offset as the CPU by default can already hit over 90C even with a 360mm AIO at 1.25V. So going to 1.35V even if it was possible would be absolutely unrealistic.

 

Yes, it was possible to adjust voltage due to a bug. But there is a reason why it's locked.

However there is no reason that a multiplier should be locked and even with the bug that gives you control over voltage you still can't adjust multiplier. The only way to clock it higher is with BCLK which has another set of very different limits and issues.

[Somerandomtechyboi]

5 hours ago, WereCat said:

The only way to clock it higher is with BCLK which has another set of very different limits and issues.

Ever even tried bclk?

Literally everyone spreading this has never tried bclk oc themselves problably because too fucking scared or outright laziness

 

ive done abit of bclk on 1155 and heavily depends on cpu but no issues other than that, trivial as shit you just clock the bclk and make sure the rams and cppu arent being clocked too much, best cpu i had was a 2100 that can do 107.5 but it just randomly up and died and i honestly dont know why and i really dont wanna make bullshit assumptions relating to bclk cause wayyy too many ppl make such bullshit assumptions with no solid proof and mislead others

 

 

goodness in the general overclocking scene either everyones a pussy or too lazy to bother, and problably the former being most common. I am neither and i am willing to test this shit myself cause i simply have no fear and if something degrades whatever ill just resell and get another, dies? Whatever just replace it or fix it, though to be fair i dont have any responsibilities so even if i lose all my money it wouldnt matter and when i get money i have no responsibilities on what to spend it on, same thing applies to rich folk that can afford top of the line stuff easily

[WereCat]

2 hours ago, Somerandomtechyboi said:

Ever even tried bclk?

Literally everyone spreading this has never tried bclk oc themselves problably because too fucking scared or outright laziness

 

ive done abit of bclk on 1155 and heavily depends on cpu but no issues other than that, trivial as shit you just clock the bclk and make sure the rams and cppu arent being clocked too much, best cpu i had was a 2100 that can do 107.5 but it just randomly up and died and i honestly dont know why and i really dont wanna make bullshit assumptions relating to bclk cause wayyy too many ppl make such bullshit assumptions with no solid proof and mislead others

 

 

goodness in the general overclocking scene either everyones a pussy or too lazy to bother, and problably the former being most common. I am neither and i am willing to test this shit myself cause i simply have no fear and if something degrades whatever ill just resell and get another, dies? Whatever just replace it or fix it, though to be fair i dont have any responsibilities so even if i lose all my money it wouldnt matter and when i get money i have no responsibilities on what to spend it on, same thing applies to rich folk that can afford top of the line stuff easily

Get off of your high horse...

 

BCLK Overclocking on old systems is easy vs new ones where you have to deal with PCIe 4. If you use NVMe you also risk data corruption even if it's unlikely but is possible. It's not really something you can push too far anyways as the PCIe 4 tends to run in some hard limits before you can gain any significant gains. It's absolutely worthless to do it on a system you run daily vs just doing it for chasing benchmark scores.

At least on Intel (heck even on Haswell) you could do BCLK that was decoupled from rest of the system and only tied to DRAM and CPU. You don't have that option on Ryzen.

[SorryBella]

27 minutes ago, WereCat said:

BCLK Overclocking on old systems is easy vs new ones where you have to deal with PCIe 4

Even in gen 3 and no nvme era BCLK overclocks are basically last resort.

[Somerandomtechyboi]

4 hours ago, WereCat said:

BCLK Overclocking on old systems is easy vs new ones where you have to deal with PCIe 4

So? Just drop it down to gen3 or gen2 atleast thats what i see any ryzen bclkers doing above 104. If thats the only thing i gotta deal with again trivial just set pcie link speed to a lower version

 

4 hours ago, WereCat said:

If you use NVMe you also risk data corruption even if it's unlikely but is possible

Still no proof of either sata hdd/ssd or nvme corruption via bclk anywhere on the internet so still dont belive that one bit. Like atleast extreme ram overclocking in certain unlucky cases where its particularly unstable can corrupt your os (think this happened to buildzoid on one of his videos) though this usually doesnt happen unless you are targetting really high freq (5000+) and also just luck factor

 

If it is true that bclk can corrupt os its problably gonna be similar to ram where you set an unstable bclk and get unlucky. Every chance i get to raise pcie freq (pre 1155) or bclk (1155 and up) i raise it because its free performance, and no issues so far even with 120 pcie freq on x58 though i need to get myself a working 1155/1150 board to test out bclk oc stability and this pentium really sucks at bclk topping out at a mere 105.4 and the 2100 is dead for whatever reason or maybe i lost the thing since i have a bunch more dead ones

 

 

Again fearlessness prevails and i enjoy performance benifits when overclocking with basically no real drawback cause i dont constrain myself with garbage arbitrary volt limits, i could care less if i had to run ddr4 at 2v or ddr3 at 2.5v if that gets me the overclock results i desire. I rarely ever see anyone actually bother to test voltages with a trial by fire method like i do so no shit voltage reccomendations have been conservative for literally forever cause everyones too scared to degrade a cpu thatll be resold or returned anyways

 

if you want examples of real overclockers look to @freeagent and @RONOTHAN## cause they arent scared to shit when it comes to voltages even if its only for benching, though i dont think anyone on this forum is insane enough to actually run high volts for daily aside from me but fair enough since at most i game and thats about it dont need maximum stability and i dont mind reselling stuff that has beed degraded

[WereCat]

8 minutes ago, Somerandomtechyboi said:

So? Just drop it down to gen3 or gen2 atleast thats what i see any ryzen bclkers doing above 104. If thats the only thing i gotta deal with again trivial just set pcie link speed to a lower version

How is this even relevant to this topic? Yes, you can drop PCIe link speed to achieve better overlocking at the cost of PCIe speed so if you have fancy NVMe 4.0 drive you may as well just use SATA... because this is not relevant for system you want to use daily and OP is not looking to OC his CPU...

 

10 minutes ago, Somerandomtechyboi said:

Still no proof of either sata hdd/ssd or nvme corruption via bclk anywhere on the internet so still dont belive that one bit. Like atleast extreme ram overclocking in certain unlucky cases where its particularly unstable can corrupt your os (think this happened to buildzoid on one of his videos) though this usually doesnt happen unless you are targetting really high freq (5000+) and also just luck factor

 

If it is true that bclk can corrupt os its problably gonna be similar to ram where you set an unstable bclk and get unlucky. Every chance i get to raise pcie freq (pre 1155) or bclk (1155 and up) i raise it because its free performance, and no issues so far even with 120 pcie freq on x58 though i need to get myself a working 1155/1150 board to test out bclk oc stability and this pentium really sucks at bclk topping out at a mere 105.4 and the 2100 is dead for whatever reason or maybe i lost the thing since i have a bunch more dead ones

 

 

Again fearlessness prevails and i enjoy performance benifits when overclocking with basically no real drawback cause i dont constrain myself with garbage arbitrary volt limits, i could care less if i had to run ddr4 at 2v or ddr3 at 2.5v if that gets me the overclock results i desire. I rarely ever see anyone actually bother to test voltages with a trial by fire method like i do so no shit voltage reccomendations have been conservative for literally forever cause everyones too scared to degrade a cpu thatll be resold or returned anyways

 

if you want examples of real overclockers look to @freeagent and @RONOTHAN## cause they arent scared to shit when it comes to voltages even if its only for benching, though i dont think anyone on this forum is insane enough to actually run high volts for daily aside from me but fair enough since at most i game and thats about it dont need maximum stability and i dont mind reselling stuff that has beed degraded

Again, none of this is relevant to this topic. The only reason why we're talking about this is because you suggested OP to use a bug in old BIOS to overvolt a CPU which can easily die with a bit more than stock voltage. IDK why the hell are you talking about fear from overclocking... you just keep suggesting things to OP he can do which he probably has no idea about and I'm trying to explain the risk so he does not just blindly do what you're suggesting. You're being malicious.

 

I do OC my GPU, DRAM and CPU in my system and I do tinker with BCLK but it's just not relevant for this topic at all. Nothing to do with fear of overclocking, just making OP aware of some risks involved.

 

There were Intel SSDs that could be overclocked and you could actually corrupt the data on them if you overdid it so it's definitely possible to achieve that with BCLK OC as well. Yes, same for DRAM OC which is definitely more common to corrupt data especially if you tinker with unstable RAM long enough.

 

IDK why do you keep talking about how you don't care about how you run your HW. I mean, good for you.... if you don't care about running RAM at 2V or CPU at high voltages and can just afford to buy a new one... what does that have to do with anything in what OP asked?

 

You don't need to tell me how to OC, I did a fair bit of it myself, you're not being special for doing so as there are plenty of people on this forum that do OC. IDK why you keep boasting about it as if you did something revolutionary by increasing voltage...

[Somerandomtechyboi]

8 minutes ago, WereCat said:

There were Intel SSDs that could be overclocked and you could actually corrupt the data on them if you overdid it so it's definitely possible to achieve that with BCLK OC as well

What ssds specifically?

Ngl overclocking ssds sounds pretty interesting so if i can find these myself might aswell try just for the fun of it and if there hasnt been any solid proof yet i may aswell provide some findings

 

12 minutes ago, WereCat said:

I'm trying to explain the risk so he does not just blindly do what you're suggesting. You're being malicious

and i explain the risks are minimal and almost always exaggerated, besides i mentioned thermal limits so op will hit a thermal wall before even hitting mildly unsafe voltages

 

13 minutes ago, WereCat said:

Yes, you can drop PCIe link speed to achieve better overlocking at the cost of PCIe speed so if you have fancy NVMe 4.0 drive you may as well just use SATA

This is only relevant for owners of gen4 drives that actually use their bandwidth and not a gen4 in a gaming system or gen3 owners

 

14 minutes ago, WereCat said:

You don't need to tell me how to OC, I did a fair bit of it myself, you're not being special for doing so as there are plenty of people on this forum that do OC. IDK why you keep boasting about it as if you did something revolutionary by increasing voltage...

I dont think im special i just really really hate conservative garbage volts ruining overclocks

 

And i have demonstrated that i can push well past it and still have no degradation if running 2v+ ddr3 is anything to go by on a platform where 1.65v is considered safe, though newer hardware is abit more fragile so volts cant really be pushed that far anymore aside from ram but thats purely ic dependant where some ddr4 will go well beyond 2v while others dont even scale past 1.35v, still have no safe voltages to reference cause its niche so i set a sanity limit of 1.7v for ddr4 (most will not scale much past 1.7v or hit diminishing returns/thermals) but id personally throw safety to the wind so i can gather data for others to use and enjoy performance benifits

[WereCat]

21 minutes ago, Somerandomtechyboi said:

What ssds specifically?

I don't remember exactly but it was either the 730 or the ones before. And I think they did not do a full release but there should be some going around.

22 minutes ago, Somerandomtechyboi said:

 

 

and i explain the risks are minimal and almost always exaggerated, besides i mentioned thermal limits so op will hit a thermal wall before even hitting mildly unsafe voltages

 

They are not minimal. The X3D CPUs can't really handle voltage increases. As I said, even on stock 1.25V the CPU already can hit over 90C even with a 360mm AIO, increasing voltage does not help at all.

Also, the CPU is power limited, you will actually hurt the performance by increasing voltage.

Also, "undervolting" by setting the Curve Optimizer offset reduces the voltage during load and increases the clocks and performance because the CPU is no longer hitting the power limit and because it's running cooler so it maintains higher clocks easier.

 

I also think the bug that allows the voltage to be controlled is for the 7000 series not the 5000 series. Also, der8auer instantly killed his 7000 X3D CPU on LN2 by just applying 1.55V in the BIOS, there was no load on the CPU and it was on LN2. This voltage is nothing too crazy for the non-X3D CPUs, especially on LN2.

 

Still, the voltage is pointless without multiplier control and BCLK cant be pushed far enough for that to matter anyways.

 

27 minutes ago, Somerandomtechyboi said:

This is only relevant for owners of gen4 drives that actually use their bandwidth and not a gen4 in a gaming system or gen3 owners

It's unnecessary hassle that is irrelevant outside of chasing benchmarking scores. If you have PCIe 4 capable system you may as well just use it. You get far more performance from tweaking your DRAM than from BCLK overclocking a 5800X3D. You need to be quite a bit lucky to get more than 104MHz BCLK for it to be worth it as 4% increase is nothing of significance for daily use. That's why I keep saying this is not relevant in any way to this topic.

Also it's not just the case of having PCIe but of how many devices you have connected. If you have 5 drives it's way more difficult to push BCLK than when you have just one for example.

 

33 minutes ago, Somerandomtechyboi said:

I dont think im special i just really really hate conservative garbage volts ruining overclocks

 

And i have demonstrated that i can push well past it and still have no degradation if running 2v+ ddr3 is anything to go by on a platform where 1.65v is considered safe, though newer hardware is abit more fragile so volts cant really be pushed that far anymore aside from ram but thats purely ic dependant where some ddr4 will go well beyond 2v while others dont even scale past 1.35v, still have no safe voltages to reference cause its niche so i set a sanity limit of 1.7v for ddr4 (most will not scale much past 1.7v or hit diminishing returns/thermals) but id personally throw safety to the wind so i can gather data for others to use and enjoy performance benifits

No, I just think your methodology of overclocking by brute forcing voltage is really bad. Often increasing voltage above certain point may result in worse stability due to thermal leakage, power limits or other factors. If increasing voltage is what makes OC good then we would have many professional overclockers.

[DarkSmith2]

14 hours ago, WereCat said:

However there is no reason that a multiplier should be locked and even with the bug that gives you control over voltage you still can't adjust multiplier

oh, actually there is a reason, the 3DV-Cache is tied to the CPU core Multiplier. If you ever wondered.

[DarkSmith2]

9 hours ago, Somerandomtechyboi said:

Ever even tried bclk?

yea but its bad. IT doesnt beat this:


https://www.userbenchmark.com/UserRun/61153405


EDIT: for those that are wondering..
https://www.igorslab.de/en/and-saying-goodbye-quiet-servus-ryzen-7-5800x3d-with-msi-center-overclocked-and-executed/

[Somerandomtechyboi]

31 minutes ago, WereCat said:

You get far more performance from tweaking your DRAM than from BCLK overclocking a 5800X3D

Isnt dram tweaking on a 5800x3d pointless? all the vids ive seen dont show it scaling past 3200

 

32 minutes ago, WereCat said:

No, I just think your methodology of overclocking by brute forcing voltage is really bad. Often increasing voltage above certain point may result in worse stability due to thermal leakage, power limits or other factors. If increasing voltage is what makes OC good then we would have many professional overclockers.

I determine thermal limits and set my voltage near or at the thermal limit, say 90c 1.45v, in that case i set 1.45v and see what freq i can get, i lower or raise if needed

 

i completely ignore safe voltages unless there is solid evidence of degradation reported by others and not bullshit some non overclocker says, like 1.45v on zen3 backed by der8aur video on degradation testing zen3, dont need arbitrary unproven garbage getting in the way of my overclocks

 

 

with this logic if i had say a 5500 based system id either look at thermal limits or best v/f though i usually prefer the latter in terms of cpu freq, say 4.8ghz 1.35v vs 5.1ghz 1.5v id prefer the former. Then when i look at ram say i got some decent 16gb 4000c18 bin klevv, i determine voltage target according to how good the cooling is so 1.9v then try to find max freq, with some references from other overclockers alongside cooling solution (fan), maybe ill hit 5800c24 1.9v, then ill run that ram oc after ensuring stability, but if imc cant actually do much more than 2600fclk even with vsoc at 1.5v i detune them to 5200c22/24 and adjust voltage accordingly maybe 1.7v, then ill run that instead cause desync fclk makes no sense on cezanne. 

 

Vcore safe seems to be around 1.45v so ill just run at most around 1.5v if i dont just run into thermal limits

Vsoc and the others in the cpu seem to be undocumented so ill stick with vcore as a safe volt, if not i just use as much as i need

Vdimm depends on ic, good ics generally dont care so ill just run as much as the ic will scale with and cooling allows, in this case around 1.9v

 

Now i can enjoy my speedy system with basically no drawback aside from the inherent volatility of the rams due to high freq, rip 5600 when i can crush it with an overclocked 5500 with some equally fast rams. If something degrades again just resell and replace, and i have never seen any reports of ram degradation

 

 

When it comes to pcs i find that i should not overrely on others and use my own methodologies, and so far ive only seen good things like having overclocks that arent crippled yet not suffering from any degradstion issues or not overspending on power supplies and still getting high quality functional 700w used psus at a mere 16$. I have not seen a single bad thing about using my own methodologies and other opinions as mere references, nor have i ever seen any bad things come from me not being a pussy and willing to try things out for myself and not just immedeatly assume "ohno 16$ used 700w unknown korea psu bad, imma waste 5x more on a branded unit with no noticable diff irl in terms of functionality" or "ohno over 1.65v vdimm bad for x58, ill just run absolute snails pace overclocks that can barely get over 1600 when a completely safe 2v could get me to 2200+"

[freeagent]

58X3D is only good for some games. If you bought this cpu, there is not much you can do with it. You turned your rig into a Dell.
 

You can bclk them on certain boards, they will never ever be fully stable though. I have one too, but it’s a love/hate relationship. I love how easy it is to cool, game performance is good, but everything else about it sucks.

 

Long story short, at stock setting, even with curve you will never see 4550 on more than just a few cores. 4450 all day long with a heavier load.

[Somerandomtechyboi]

2 minutes ago, freeagent said:

You can bclk them on certain boards, they will never ever be fully stable though

Wdym by never fully stable?

like crashing in stresstests or something else?

 

i have very limited bclk experience past 1366 cause i dont have any 1155 and up platforms currently usable and my pentium is garbage at bclk

[WereCat]

Just now, Somerandomtechyboi said:

Isnt dram tweaking on a 5800x3d pointless? all the vids ive seen dont show it scaling past 3200

 

no, it still scales well enough but gains are not as great as with non-X3D but they are still worthwhile

 

I am 17th place on HWbot with just CO and memory OC for R23 (which is not really known to scale well with memory either) and almost everyone above me uses BCLK OC just to get few points more. Also my placement is with a 24/7 stable system, I could probably push it to top 10 again if I tried with pushing more aggressive clocks and timings that are not fully stable.

[freeagent]

49 minutes ago, Somerandomtechyboi said:

Wdym by never fully stable?

On 58X3D. The chip has a hard limit of 143w PPT. In order to be stable at 4600 and above, you need access to more PPT TDC and EDC. So yes, you can run a benchmark at 4600-4700, but it will never ever be all load stable because of its hard power limits. Benchmarks can be deceiving. And Cinibench is a benchmark, not a stress test.

[No_dice]

i am afraid that i might have opened a can of worms here 

but as @Shimejii posted, i do have normal boost clocks ingame or when testing in 3d mark (4.45)

 

what i do not understand is the low boost clocks in cinebench, occt and other synthetic cpu loads.

it must be a thermal wall i am hitting.

 

furthermore, when i undervolt like described in many video's using PBO2. my cpu does the exact oposite of what i dail in.

for example, -20 does not lower but increses the voltage. drawing more watt and running hotter.

same goes for using optimizer curve.

 

to be clear, i am not trying to OC the chip, i just wanted to try the undervolting with PBO2. just to see if there was a benifit to be had i terms of temps and power consumption. 

this is when i noticed i dont hit the boostclocks in syntethic loads that i expected.

and that the "negative offset" gives me the opposite results.

 

 

[Somerandomtechyboi]

1 hour ago, WereCat said:

no, it still scales well enough but gains are not as great as with non-X3D but they are still worthwhile

huh interesting

 

Do the x3d chips still do the same fclk the normal ryzens do? Ive heard that 7000x3d chips can run higher fclk compared to non x3d

 

14 minutes ago, freeagent said:

On 58X3D. The chip has a hard limit of 143w PPT. In order to be stable at 4600 and above, you need access to more PPT TDC and EDC. So yes, you can run a benchmark at 4600-4700, but it will never ever be all load stable because of its hard power limits. Benchmarks can be deceiving. And Cinibench is a benchmark, not a stress test.

Cant even change the limits in bios?

[freeagent]

6 minutes ago, Somerandomtechyboi said:

Do the x3d chips still do the same fclk the normal ryzens do?

Oh yeah, that part is no problem, my X3D clocks better on the IF than my 5900X does, my 5600X is still my best.

 

7 minutes ago, Somerandomtechyboi said:

Cant even change the limits in bios?

You can lower the limits, but you don't want to do that..

[WereCat]

5 minutes ago, Somerandomtechyboi said:

huh interesting

 

Do the x3d chips still do the same fclk the normal ryzens do? Ive heard that 7000x3d chips can run higher fclk compared to non x3d

Yes. But it seems like the early batches of 5800X3D can more likely reach 2000MHz than the newer ones. I can to 1933Mhz FCLK on mine but it's really at the very edge of stability. I can probably get it stable but it's so much work and voltage that it takes power from the rest of the CPU so it actually scales negatively since the CPU is power limited.

[WereCat]

22 minutes ago, No_dice said:

i am afraid that i might have opened a can of worms here 

 

No, it's all right