Be careful overclocking

Im not specifically against over clocking but i advise caution doing so as it generates so much more heat and can lead to early failure of your cpu.
depending on the need for processing speed a lot more can be accomplished with the use of smp configurations. (multiple nodes linked and running smp software) are quite fast even on cat 5 and sheer lightning on fibre network

if you do overclock make absolutely sure your cooling is sufficient to handle the job.
I wont advertise any specific water cooling system as there are many out there. but an on screen temperature monitor is a must.
positive pressure ventilation can keep dust levels down providing intakes are filtered and filters cleaned frequently and regularly.
you can control individual case fans using relays and a programmed arduino or raspberry pi. to run as needed when temperature rises beyond certain levels.
so you can experiment as much as you like.

you should also consider that cpu’s and components today use so much smaller amounts of materials that they are not as robust os the older equipment was and much less forgiving.

3 Likes

Advise that is at first “aaa no shit” BUT!!! is of much importance. Especially now with b350/b450 mobos using VRM that is questionable at best for overclocking. It’s not made for it but stupid advertisement will make people do terrible things. No airflow over the VRMs and you can hit thermal runaway fast.

2 Likes

I am really wondering why someone like Arctic or Thermaltake does not make custom VRM heatsinks.

3 Likes

Probably because if your VRMs are getting that hot, heat sinks are a band-aid for which the proper solution is an adequate power delivery configuration…

With the money you’d spend on custom heatsink for the VRM you could just buy a proper board…

1 Like

30€ don´t get you far when you are on higher-end platforms

also, not many motherboards available if yours is several years old and you’re trying to extend the life of it by overclocking.

Time and effort plus postage plus thermal compound plus heat sinks… not worth it.

30 euro could easily be the difference between a b350/450 board that has garbage for a vrm and one that doesn’t.

on x370/470 you can get good VRMs without breaking the bank.

extending life a bit with overclocking is imho a waste of time. if the machine is too slow because its old, overclocking won’t get you THAT much. we aren’t dealing with celeron 300As any more.

Put the money away that you’d be spending on coolers, etc. into the piggy bank for a proper upgrade.

2c.

Heatsinks for electronic components are fairly cheap (80ct a pop), compund or retention stuff is not that expensive either.

Concerning VRMs on mainboards: It does not get you anywhere without overspending.

1 Like

It’s platform specific. I have my P7P55D Deluxe 7 years going strong. Combined 5 Years OC 4.2GHz on 760 and now 4.0GHz on current x3440. Because any CPU from 1. gen is unusable garbage if left at stock. But overclocked, there is no reason to buy new. Even games are fine in my scenario because I run them at 3440x1440.

I’d like to dispute this. If you are overclocking without touching voltages as is very often the case, the stock cooling is usually more then good enough.

Overclocking (not overvolting) at the worst can cause transistor latch up, or skew timing beyond limits causing erroneous readings (both scenarios do not cause damage), it doesn’t however cause more then a marginal temperature increase over stock clock speeds.

If you want to run faster your goal is to remove the heat and keep the temps down to minimise transistor leakage. As the CPU runs faster it will require more current, which is where the voltage boost helps, but again, until you start messing with voltages extra cooling is not that critical.

I have been overclocking CPUs since the 386SX was new. There is tons of misinformation on how to overclock and what you need. For example, heat spreaders on DDR are laughable. For RDRAM and FBDIMMs (both obsolete) it’s important and makes a difference, but DDR, SDRAM, etc… it’s just marketing wank.

2 Likes

I concur with this.
Overclocking without over-volting has negligible thermal and power implications.

The reason it gets hotter at all is that switching transistors faster causes them to be ‘on’ more often than at lower clocks so slightly more power is used. This is also why overvolting can be useful to stabilize higher clock rates.

Take these two waveforms for example, the bottom waveform has more peaks at a higher cyclic rate than the top waveform.

Transistors switching at a higher rate will dissipate more power and have more heat loss, but it is rather marginal with the sorts of clock increases you can achieve with computer hardware and heat is of limited concern.

With modern CPU’s automatic thermal management they will slow themselves down to a crawl and undervolt and downclock to cope before finally shutting down.

However without an accompanying voltage increase the difference in heat and current draw from overclocking is very slight.

Much rather it can affect the stability of your system in difficult to notice ways if not subjected to exhaustive verification testing. Such as Increasing the chances of memory or mathematical errors.

RAMBUS On a Socket 423 Now that’s some hot stuff! :sunglasses:

That said darwinian evolution strongly applies to overclocked hardware and those who overclock with all the volt modding that entails. If you get it wrong your hardware gets toasted real fast and you learn from it, if you get it right you’re going to keep doing it and probably get better at it, with many broken pieces of hardware littered in your wake :smiley:

1 Like

I thought so until I built a server with 16x4GB FBDIMMs in it… without fan forced cooling, after about 60 seconds there was enough radiated heat coming of the board to feel it 50cm away. Even with cooling the DIMMs averaged 70-80C.

See the issues that Mac user’s had with overheating when Apple in their infinite wisdom decided that general users should use FBDIMMs

http://www.xlr8yourmac.com/systems/Mac_Pro/mac_pro_ram.html

1 Like

completely disagree. Overclocking something like a 2600k gets you a lot. Also, if you don’t have ddr4 ram, it is super fucking expensive to replace the cpu, mobo, ram, possibly cooler.

2 Likes

Something that kind of drives the point home:

image

The advanced memory buffer (AMB) chip has an integrated Thermal Sensor. The AMB dissipates most of the heat on an FB-DIMM system. :smiley:

AFAIK the AMB’s ran at ~4Ghz to handle the the data rates?

1 Like

Not sure, I do know however it was a very effective space heater for the winter.

2 Likes

Yes it was. Burnt my fingers on those sticks at some point or another…

@Gnuuser

Nice content, but your OP opening seems a bit random. Kind of like drive by posting since you’ve not returned to the thread you have started.

Now time to fess up, what did you destroy to prompt you to learn these things and then pass on your new found pearls of wisdom? There must be a story behind your post…

…for me it was a dark night long ago (well, November 1998) and my newly built PC was humming along quite nicely with an AMD K62 350MHz. My one gripe however was that I still could not turn on all the effects (clouds etc) in Janes WWII Fighters, unlike my smug bastard friend who had a Pentium II 333Mhz cpu. So I thought about over-clocking…

Long story short …I had to use some of my student loan to buy a new mobo and a Celeron 300A, that chip was amazing. Once I installed a 3rd party cooler it was a simple case to set the multiplier to X4.5, and yes it was stable at 450MHz, no voltage tweaks needed, and now I had max effects in WWII Fighters, pure awesomeness :slight_smile:

2 Likes

actually never destroyed any equipment myself but seen it happen when guys did it at the shop at work too many times!
my job didn’t pay much so having to buy new mobos was quite out of the question when your raising a family unless you could make a tidy profit from it.

being industrial I’ve also seen the damage that excessive heat does to electronics all too often.
replacing a plc processor for instance when you add the cost of downtime, labour, and hardware it can easily exceed $1500.00 to $3000.00

think about it your hardware is an investment and yes you can tweek the performance with a little judicious over-clocking but you have to ask your self how much is too much? and what will it cost you if you go too far?

thats why i stated Im not against it but be careful doing it

1 Like

In my honest opinion it’s a personal choice best left to each individual. :slight_smile:

I agree you dont have to buy the most expensive board out there,
to get a decent vrm.
There are also options in the midrange segment.

Here is a topic i made on vrm’s for Am4 boards.
Maybe i should add b350 and X370 boards info’s aswell.
But i’m fairly sure that i have posted spec’s and details on those aswell on review topics that wendell did.