My new build PC not stable with XMP1 profile

[tranthephongpq]

I have a set of PC specifications as follows:

• CPU: Intel Core i5-13500
• Motherboard: MSI B660 Mortar DDR5
• RAM: Kingston Fury Beast DDR5 5600MHz 16GB (8GBx2) installed in slots 2 and 4
• GPU: MSI RX6750 Mech 12GB OC
• Storage: Samsung 980 Pro 1TB NVMe SSD
• PSU: Corsair CV750 80 Plus Bronze

My PC currently runs stably with the XMP2 profile enabled, which sets the RAM speed to 5200MHz without encountering any errors in memtest86. However, when I switch to the XMP1 profile to achieve a RAM speed of 5600MHz, I experience a Blue Screen of Death (BSOD) while gaming.

I'm wondering if the issue lies with my RAM. Can you please provide some assistance and guidance?"

 

[badboy2k]

for your Intel Core i5-13500 its memory controller specs are:

Up to DDR5 4800 MT/s <<< 4800MHz
Up to DDR4 3200 MT/s

the cpu you have are a lower Grade CPU with a much worse Memory Controller in them as the i7 has a rating Upto 5600 MT/s (5600MHz) and is the reason its a lower skew as i5 cpus tend to be from the same Die as the i7's came off but with Defects that stop them being a i7 as they are not exposed well.

so its your CPU with a Weak Memory Controller that is the Issue here.

I5 - 13500

Intel® Core™ i5-13500 Processor (24M Cache, up to 4.80 GHz) - Product Specifications | Intel

Intel® Core™ i5-13500 Processor (24M Cache, up to 4.80 GHz) quick reference with specifications, features, and technologies.

www.intel.com

i7 - 13700K

Intel® Core™ i7-13700K Processor (30M Cache, up to 5.40 GHz) - Product Specifications | Intel

Intel® Core™ i7-13700K Processor (30M Cache, up to 5.40 GHz) quick reference with specifications, features, and technologies.

www.intel.com

 

[tranthephongpq]

for your Intel Core i5-13500 its memory controller specs are:

Up to DDR5 4800 MT/s <<< 4800MHz
Up to DDR4 3200 MT/s

the cpu you have are a lower Grade CPU with a much worse Memory Controller in them as the i7 has a rating Upto 5600 MT/s (5600MHz) and is the reason its a lower skew as i5 cpus tend to be from the same Die as the i7's came off but with Defects that stop them being a i7 as they are not exposed well.

so its your CPU with a Weak Memory Controller that is the Issue here.

I5 - 13500

Intel® Core™ i5-13500 Processor (24M Cache, up to 4.80 GHz) - Product Specifications | Intel

Intel® Core™ i5-13500 Processor (24M Cache, up to 4.80 GHz) quick reference with specifications, features, and technologies.

www.intel.com

i7 - 13700K

Intel® Core™ i7-13700K Processor (30M Cache, up to 5.40 GHz) - Product Specifications | Intel

Intel® Core™ i7-13700K Processor (30M Cache, up to 5.40 GHz) quick reference with specifications, features, and technologies.

www.intel.com

I thought that many RAM modules had speeds exceeding 5600MHz, and I did not possess RAM with such a high speed.

 

[badboy2k]

I thought that many RAM modules had speeds exceeding 5600MHz, and I did not possess RAM with such a high speed.

there is RAM with much Higher speeds as they go upto like 7800MHz but for that what you need is what is called a Unicorn CPU that has come directly from the Center of the Die that all the CPUs came off that is exposed Perfectly without a single Flaw on it.

most i7 CPU's out there may do upto 6200Mhz but beyond that is you need to de getting into the most high quality CPU's out there as not all CPU's are the same as you can get like 50 of them and some will stop at just on spec and then you may have out of the 50 one that hits way above all the others. its called the Silicon Lottery where some are just at spec and a odd one here or there is way better and can do more.

 

[tranthephongpq]

there is RAM with much Higher speeds as they go upto like 7800MHz but for that what you need is what is called a Unicorn CPU that has come directly from the Center of the Die that all the CPUs came off that is exposed Perfectly without a single Flaw on it.

most i7 CPU's out there may do upto 6200Mhz but beyond that is you need to de getting into the most high quality CPU's out there as not all CPU's are the same as you can get like 50 of them and some will stop at just on spec and then you may have out of the 50 one that hits way above all the others. its called the Silicon Lottery where some are just at spec and a odd one here or there is way better and can do more.

Is there any method to operate my PC at its maximum speed of 5600MHz?

 

[badboy2k]

Is there any method to operate my PC at its maximum speed of 5600MHz?

the SA Voltage on your CPU if you can Increase it may make it work at a higher speed without errors for a while but its not recommended as what will happen is over time it will start to degrade and then it may cause it to no longer even work at the 5200MHz you have it at currently.

SA = System Agent and is the Voltage the Memory Controller is on but like i said its not recommended unless you only plan on having the PC you have for a few years tops as it will Rapidly Degrade the CPU itself

 

[tranthephongpq]

the SA Voltage on your CPU if you can Increase it may make it work at a higher speed without errors for a while but its not recommended as what will happen is over time it will start to degrade and then it may cause it to no longer even work at the 5200MHz you have it at currently.

SA = System Agent and is the Voltage the Memory Controller is on but like i said its not recommended unless you only plan on having the PC you have for a few years tops as it will Rapidly Degrade the CPU itself

I've come across some posts discussing the possibility of increasing the DRAM voltage. Is this feasible?

 

[badboy2k]

I've come across some posts discussing the possibility of increasing the DRAM voltage. Is this feasible?

its possible to increase the DRAM Voltage but may not work if your Memory Controller can not cope with it still as its more dependant on the SA Voltage to the CPU to make the memory Controller work faster.

if you do play with the DRAM Voltage i would suggest you not going much above 1.45V as then your getting into the territory of potentially damaging the memory Itself so go in 0.025V steps if you do that till you find a stable voltage as you do not want to be too aggressive with it so try 1.375V and see if that works if not bump it up to 1.400V and so on

 

[citay]

I'm wondering if the issue lies with my RAM. Can you please provide some assistance and guidance?"

The very first thing to try is updating to the latest BIOS version, if you're on an older one.
Your updates are here: https://www.msi.com/Motherboard/MAG-B660M-MORTAR/support
Update via M-Flash in the BIOS, then try again with XMP.

For stability testing, waiting for a bluescreen is not the proper way. Memtest86 is a good start, or check in my RAM thread under 5) for some stress testing tools in Windows.

RAM: Kingston Fury Beast DDR5 5600MHz 16GB (8GBx2) installed in slots 2 and 4

Not a big fan of the 8GB modules (see here), for DDR5 i would've gone 2x 16 GB. But we have to deal with what you got. Same for the PSU, the Corsair CV650 is a budget PSU, it's not outright catastropic, but still classed as "Tier C • Low End" on the PSU tier list. Could be better, could be worse, i guess.

I've come across some posts discussing the possibility of increasing the DRAM voltage. Is this feasible?

DRAM Voltage is very rarely the limiting factor. The RAM itself is tested at the factory to work fine at their XMP speed with the XMP voltage. The sticking point is usually some other IMC-related voltage, this would mainly be CPU SA, CPU VDDQ and CPU VDD2 Voltage. You should first check what the BIOS sets for those when you enable the faster XMP profile and report them here. I personally don't believe DRAM Voltage will get you anywhere, or if it does, it would be indirectly, because above a certain DRAM Voltage, the BIOS can automatically raise IMC-related voltages as well. Anyway, first try the BIOS update.

 

[tranthephongpq]

The very first thing to try is updating to the latest BIOS version, if you're on an older one.
Your updates are here: https://www.msi.com/Motherboard/MAG-B660M-MORTAR/support
Update via M-Flash in the BIOS, then try again with XMP.

For stability testing, waiting for a bluescreen is not the proper way. Memtest86 is a good start, or check in my RAM thread under 5) for some stress testing tools in Windows.



Not a big fan of the 8GB modules (see here), for DDR5 i would've gone 2x 16 GB. But we have to deal with what you got. Same for the PSU, the Corsair CV650 is a budget PSU, it's not outright catastropic, but still classed as "Tier C • Low End" on the PSU tier list. Could be better, could be worse, i guess.



DRAM Voltage is very rarely the limiting factor. The RAM itself is tested at the factory to work fine at their XMP speed with the XMP voltage. The sticking point is usually some other IMC-related voltage, this would mainly be CPU SA, CPU VDDQ and CPU VDD2 Voltage. You should first check what the BIOS sets for those when you enable the faster XMP profile and report them here. I personally don't believe DRAM Voltage will get you anywhere, or if it does, it would be indirectly, because above a certain DRAM Voltage, the BIOS can automatically raise IMC-related voltages as well. Anyway, first try the BIOS update.

I overclocked the memory and ran the memory test using the mainboard's functionality with the following profile: 5600 MHz, timings set to 36-36-36-77, and voltage at 1.35 volts (compared to the XMP1 profile, which is 5600 MHz with timings at 40-40-40-135 and a voltage of 1.25 volts). After conducting four passes with memtest86, the system appears to be stable.
I've observed that the voltage is higher than the XMP1 profile. Is this elevated voltage safe for my CPU, or could it potentially lead to degradation?

 

[citay]

Did you find that 1.35V was necessary for stability? Usually you want to determine the lowest voltage that is still stable, then add a tiny bit on top for additional headroom. Adding a full 100 mV is a rather big step that might not be necessary for a timing adjustment without raising the frequency. If you find that 1.25V is not sufficient, try at 1.3V first, don't go right to 1.35V. This goes for all voltages. First you orientate yourself on what is set for XMP and raise it just slightly above that, not 100 mV higher right away.

 

[tranthephongpq]

Did you find that 1.35V was necessary for stability? Usually you want to determine the lowest voltage that is still stable, then add a tiny bit on top for additional headroom. Adding a full 100 mV is a rather big step that might not be necessary for a timing adjustment without raising the frequency. If you find that 1.25V is not sufficient, try at 1.3V first, don't go right to 1.35V. This goes for all voltages. First you orientate yourself on what is set for XMP and raise it just slightly above that, not 100 mV higher right away.

I applied a Memory Try It profile, allowing the mainboard to handle all the configurations, as I aimed to assess the system's stability at 5600MHz. My apologies for any inconvenience; I'm still relatively inexperienced when it comes to overclocking memory.

 

[citay]

Ok. Well, i would activate the first XMP profile (which was unstable), and take a screenshot of how the voltages are set there, then compare it to what you have set now.
Maybe it's not that much of a difference. But i would try DRAM Voltage at 1.3V for example, it might be enough for what you changed with the timings.

 

[tranthephongpq]

Ok. Well, i would activate the first XMP profile (which was unstable), and take a screenshot of how the voltages are set there, then compare it to what you have set now.
Maybe it's not that much of a difference. But i would try DRAM Voltage at 1.3V for example, it might be enough for what you changed with the timings.

How should I increase the CPU SA, CPU VDDQ, and CPU VDD2 voltages? Is it possible to replace the RAM sticks with ones from the motherboard's supported list without needing to replace the CPU, and still have the XMP1 profile activate without errors? Thanks.

 

[citay]

Look, if you just have to tune the voltages slightly to make it stable, that is perfectly ok. I had to do it for my kit too, and that's DDR4. So it's nothing completely unusual, sometimes there's just one voltage that is a tad too low for stability, then you tweak it and Bob's your uncle.

You already achieved stability at even better timings than XMP, that's quite a success, so don't think about replacing your RAM kit now. You have all the tools to make this 100% stable at XMP and even a bit beyond.

All i'm saying is, let's compare the voltages, how XMP sets them, and how they ended up for you on the screenshots you posted, and let's see how different they are.

 

[tranthephongpq]

Last night, I overclocked my system, and the motherboard detected a failure. In response, I reset all the settings and opted for XMP1, with the only change being switching the cooler to a 65w fan. Surprisingly, I couldn't figure out why my system remained stable during memtest (4 passes, twice). I've examined the Event Viewer, and it's showing numerous warnings and errors.

 

[citay]

Usually it doesn't make that much sense anymore nowadays to overclock your system. How did you do it?

Selecting a different cooler setup in the BIOS is really just MSI's way to set the power limits for the CPU:

The only way to enforce the Intel power limits (if you want that) is to select "Boxed Cooler". "Tower Air Cooler" already means that the limits are effectively removed, because PL1 and PL2 (Long and Short Duration Power Limit) are set to 288W in this example. A power limit that is above what kind of power most CPU models (except i7/i9) will draw, and thus basically identical to the "Water Cooler" option.

On some other boards, it might be slightly different, like 255W for Tower Cooler. Still, "Tower Cooler" is a wasted opportunity to select a good middle ground, like 150-200W, which would be the maximum for all but the best tower air coolers to handle. Anything over 200W, you either need one of the best tower air coolers, or you're in water cooler territory.

But with a 13500, which is the same CPU i have, you can select Tower Air Cooler or Water Cooler no problem, because the maximum CPU-only power draw is around 135W in Cinebench R23, which represents the maximum load you can create using normal workloads. Only in Prime95 can it go higher, but that's not normal load anymore. So, 135W can be handled by a good tower cooler. Only if your cooler is not very good, you have to set lower limits, but 65W Long power limit is a bit too low, it restricts performance a bit too much under sustained full load.

And overclocking and setting the official Intel power limits makes no sense at all, your overclock will never come into effect. Because at that low of a power limit, the CPU will be throttled (clocked lower), never mind keeping its original boost clocks or even beyond with OC! So what you are doing cancels each other out, more or less. But as i said, i wouldn't OC anyway.

 

[tranthephongpq]

Usually it doesn't make that much sense anymore nowadays to overclock your system. How did you do it?

Selecting a different cooler setup in the BIOS is really just MSI's way to set the power limits for the CPU:

The only way to enforce the Intel power limits (if you want that) is to select "Boxed Cooler". "Tower Air Cooler" already means that the limits are effectively removed, because PL1 and PL2 (Long and Short Duration Power Limit) are set to 288W in this example. A power limit that is above what kind of power most CPU models (except i7/i9) will draw, and thus basically identical to the "Water Cooler" option.

On some other boards, it might be slightly different, like 255W for Tower Cooler. Still, "Tower Cooler" is a wasted opportunity to select a good middle ground, like 150-200W, which would be the maximum for all but the best tower air coolers to handle. Anything over 200W, you either need one of the best tower air coolers, or you're in water cooler territory.

But with a 13500, which is the same CPU i have, you can select Tower Air Cooler or Water Cooler no problem, because the maximum CPU-only power draw is around 135W in Cinebench R23, which represents the maximum load you can create using normal workloads. Only in Prime95 can it go higher, but that's not normal load anymore. So, 135W can be handled by a good tower cooler. Only if your cooler is not very good, you have to set lower limits, but 65W Long power limit is a bit too low, it restricts performance a bit too much under sustained full load.

And overclocking and setting the official Intel power limits makes no sense at all, your overclock will never come into effect. Because at that low of a power limit, the CPU will be throttled (clocked lower), never mind keeping its original boost clocks or even beyond with OC! So what you are doing cancels each other out, more or less. But as i said, i wouldn't OC anyway.

I've updated the BIOS settings to use a Tower Air Cooler. My primary objective is to maximize RAM performance and avoid unnecessary expenses. However, despite passing memtest, it seems that overclocking isn't providing the stability I hoped for.

 

[citay]

What and how did you OC exactly?

 

[tranthephongpq]

What and how did you OC exactly?

I attempted to set my memory to 6000MHz (just try), but it failed. Next, I opted for XMP1, but I had to adjust the voltage from 1.25 to 1.3. Surprisingly, the system reported an overclocking failure, which hadn't occurred previously when I tried setting the memory to 5800MHz, and it still booted up to Windows. Eventually, I decided to reset all the settings and simply selected XMP1, and it has been working without any errors ever since.