Is X58 X5850 overclocking with high QPI voltage safe?

[Thrax]

I have been experimenting again with overclocking my system to get smooth 60FPS in modern titles. After some tinkering, I have settled on 4GHz with turbo boost up to 4.6GHz using a 200mhz bclk.

I am using an MSI X58 Pro with an X5650 Xeon and watercooling.

Unfortunately inorder to get to 4ghz stable, I have had to increase QPI voltage to 1.46v. It seems I cant lower QPI ratio on this board to use a lower voltage hence the need to increase for overclocking.

I have read numerous posts citing going above 1.3v is bad for the chip's IMC but never really found what this is based on. Maybe its not bad afterall? I'm not too bothered if the chip eventually dies as will replace with an X5675 with higher multiplier or upgrade to Ryzen etc.

I was wondering what other peoples experience was using higher QPI voltages and if anyone has run their chip at high levels for any sustainable period?

 

[savior02]

I have been experimenting again with overclocking my system to get smooth 60FPS in modern titles. After some tinkering, I have settled on 4GHz with turbo boost up to 4.6GHz using a 200mhz bclk.

I am using an MSI X58 Pro with an X5650 Xeon and watercooling.

Unfortunately inorder to get to 4ghz stable, I have had to increase QPI voltage to 1.46v. It seems I cant lower QPI ratio on this board to use a lower voltage hence the need to increase for overclocking.

I have read numerous posts citing going above 1.3v is bad for the chip's IMC but never really found what this is based on. Maybe its not bad afterall? I'm not too bothered if the chip eventually dies as will replace with an X5675 with higher multiplier or upgrade to Ryzen etc.

I was wondering what other peoples experience was using higher QPI voltages and if anyone has run their chip at high levels for any sustainable period?

What are your hardware specs?

 

[silentbogo]

QPI is how older CPUs communicate with each other and PCH.
x58 is hot as it is, and raising QPI voltages higher than recommended will make it run even hotter, and since most people don't bother with additional cooling for chipset - it may even cause long-term damage. That's why back in those days we had waterblocks for ASUS Rampage II/III and other expensive trinkets.

I have read numerous posts citing going above 1.3v is bad for the chip's IMC but never really found what this is based on.

Well, 1.3V is an arbitrary "safe" number not really backed by anything.
All you need to remember is this:
- The "stock" value is 1.2V.
- anything below 1.4V is considered safe
- anything in 1.4-1.6V range is fine for OC, assuming you know what you are doing and you did all the prepwork(e.g. proper hub cooling, which on that MSI is a must)
- anything above 1.6V is a danger zone
These voltages were pretty universal across all x58 boards, so I assume they are based on Intel spec and recommendations.

In your case, you have to add some sort of active cooling to your hub, otherwise it'll either become unstable soon and affect your OC, or in a worst case scenario - start degrading and die.
Also, try to keep your voltages as low as you can, cause that MSI x58 Pro doesn't deserve that suffix in the slightest. It has the shittiest VRM you can possibly imagine for LGA1366 platform, and past certain point those Westmere CPUs become super-power-hungry. My old X5650 would trip an alarm on my 800VA UPS when I hit 4.7GHz at 1.45V Vcore. Considering it was paired with stock GTX 750 Ti at that time, it means the CPU was pulling upwards of 250W on its own during benchmark. Your board will eventually explode with wattages like these.
You should look for other reasons why your CPU can't hold 4GHz (e.g. drop QPI link speed, loosen RAM clocks/timings, play with UNCORE etc.) and start lowering your vcore to the lowest possible value.
Otherwise, you can assume you lost a silicon lottery or some other dude already clocked the shit out of that poor little Westmere.

 

[freeagent]

I ran 1.4v qpi on my x5690 for years. Don’t forget about the sag either, it may say 1.4 in the bios but the actual voltage may be down around 1.375. My ud5 rev1 was terrible on all fronts for saggy voltage. My R3F has a jumper that I have enabled so it doesn’t sag so low on the qpi.

 

[Thrax]

What are your hardware specs?

X5650 CPU
MSI X58 Pro
48GB DDR3 - Corsair Vengeance 1600 C9
INNO3D GTX 1080
Inateck PCI-e to USB 3.0 (4 Ports)
Samsung 840 EVO 500GB SSD
Corsair AX1200i Power Supply

Also, try to keep your voltages as low as you can, cause that MSI x58 Pro doesn't deserve that suffix in the slightest. It has the shittiest VRM you can possibly imagine for LGA1366 platform, and past certain point those Westmere CPUs become super-power-hungry. My old X5650 would trip an alarm on my 800VA UPS when I hit 4.7GHz at 1.45V Vcore.

System is drawing up to 500W fully loaded according to Corsair Link. There seems to be 5 phases for the VRMs , do you know how many amps they are rated for? If they are say 40A each, I can imagine I am at the limit for cpu power delivery (e.g 40a x 1.35v x 5 phases = 270 watts)

I ran 1.4v qpi on my x5690 for years. Don’t forget about the sag either, it may say 1.4 in the bios but the actual voltage may be down around 1.375. My ud5 rev1 was terrible on all fronts for saggy voltage. My R3F has a jumper that I have enabled so it doesn’t sag so low on the qpi.

I was wondering if this might be a possibility.

 

[silentbogo]

System is drawing up to 500W fully loaded according to Corsair Link. There seems to be 5 phases for the VRMs , do you know how many amps they are rated for? If they are say 40A each, I can imagine I am at the limit for cpu power delivery (e.g 40a x 1.35v x 5 phases = 270 watts)

Can't tell you with 100% certainty (only ~99% ), but at that time MSI's favorites were Renesas R2J20602, which is rated at 40A max continuous. If it's R2J20655, then it's only 35A continuous.

 

[basco]

did ya try Cputweaker to lower qpi ?

CPU-Tweaker Download

Eine spezielle Software für Hardwaretuner. Freeware, kostenloser Download!

www.computerbase.de

 

[Thrax]

Can't tell you with 100% certainty (only ~99% ), but at that time MSI's favorites were Renesas R2J20602, which is rated at 40A max continuous. If it's R2J20655, then it's only 35A continuous.

Corsair link for my psu shows that, I am drawing 13 Amps on the cpu 8 pin power connector when the cpu is fully loaded. Occasionally it spikes up to 17 amps when the CPU turbo boosts up to 4.7ghz.

13a x 12v = 156 watts
17a x 12v = 204 watts

Originally I had turbo boost disabled for overclocking but I turned it on just to see if the system would still be stable. Given 17 amps is possible, I guess I still have some additional headroom with the VRMs?

 

[silentbogo]

Originally I had turbo boost disabled for overclocking but I turned it on just to see if the system would still be stable. Given 17 amps is possible, I guess I still have some additional headroom with the VRMs?

It's not just about how much current can you put through that VRM, but also about how much that VRM can handle. Basically, it's a heat problem.
As a reference, read up through R2J20602 datasheet:

https://cdn.datasheetspdf.com/pdf-down/R/2/J/R2J20602NP_RenesasTechnology.pdf

I'm not sure if your board has a VRM temperature sensor, but if it did - it'd be scary to look at with that puny heatsink.
For reference, at 25A and in ideal conditions each phase has a 4-5W power loss(e.g. heat). That's already 20-25W total, which is already sketchy for a tiny heatsink even with decent airflow over it, but within reasonable range. These are your worst-case conditions for any stock 130W CPU of that time. Even if VRM heats up to ~110C - it'll work, but anything past that is when things get sketchy.
At 35A output and a slight output voltage bump your power loss nearly doubles, that joke of a heatsink must now dissipate somewhere around 40-50W from all 5 phases! If you go through datasheet(safe operating area graph), you'll see that past 100C the "safe" output starts to drop drastically. Also at those temps the IC package cannot transfer that much heat to heatsink anymore and will heat up even more. Best case - thermal shutdown. Worst case - blown VRM. Basically it was designed to be an absolute bare minimum for Nehalem i7's, so that kids can play Crysis.
There is a reason why on x58 Pro-E the VRM heatsink is almost twice the size comparing to regular x58 Pro. Similarly priced P6T SE had 8 phases and similar heatsink for each 4-phase block. Even chinese knockoff x58 boards with similar phase count have larger heatsinks on VRM and NB.

As I said before, if you want stable overclock past 4GHz - stop increasing Vcore and QPI voltages and look elsewhere for issues(RAM speed/timings, Uncore, QPI speed, Spread Spectrum etc). And disable turboboost.

 

[Thrax]

I know what the problem is, basically I cannot lower QPI ratio sufficiently unlike other x58 boards.