Overclocking on Ivy Bridge ...while undervolting???

[thegoatman]

I just came across this article on Anandtech, linked from their Ivy Bridge review:

http://www.anandtech.com/show/5763/u...-on-ivy-bridge


It's very interesting. While upping both the voltage and the clock speeds have rapid and averse affects on the temps of the chip... they managed to get a significant, stable overclock at 0.90 volts! Which suggests to me that if we go about it the right way, we can actually do fairly well at overclocking IB. Probably not as insanely awesome as SB, but still.

[eXa]

4.5ghz and 1.1v
Fine by me for a nice low wattage 24/7 OC.

[AFQ]

4.5ghz and 1.1v
Fine by me for a nice low wattage 24/7 OC.

Yeah 4.5 GHz seems to be the sweetspot for Ivy Bridge. After that things start to get nasty.

[thegoatman]

I'd be satisfied with that for $250. Especially coming from an old Athlon64 4000+ San diego single core @2.4ghz...

[thegoatman]

Bump.

If over-volting causes premature CPU aging, will under-volting prolong its life?

[JaD]

Yes, but it's negligible over a life span of 10 years.

[OCX600RR]

I used to undervolt my E8400 Wolfdale and could still manage some decent clocks out of it. Never tried with my 2500K yet but I think I will go ahead and play around with it some today. Sitting at 4.5GHz @ 1.26v

[AliG]

I'd be satisfied with that for $250. Especially coming from an old Athlon64 4000+ San diego single core @2.4ghz...

I recommend swapping out your computer chair with a racing seat for the first month haha. Switching from an AM2 dual core to my 2500k damn near blasted my pants off

[Manicdan]

shockingly similar to a review i did of thuban a few years back

its good to show people what it takes to get different frequencies so they can choose how to use the chip.

almost all cpus can be overclcoked some with less voltage as long as temps are maintained. the stock chip has to work at its set MHz up to its max thermal limit, which is why stock voltages are pretty high compared to what is really needed.

they should have tested 1v though, since the jump from .9 to 1.1 is large and there could be an incredible sweet spot between them.

[AliG]

shockingly similar to a review i did of thuban a few years back

its good to show people what it takes to get different frequencies so they can choose how to use the chip.

almost all cpus can be overclcoked some with less voltage as long as temps are maintained. the stock chip has to work at its set MHz up to its max thermal limit, which is why stock voltages are pretty high compared to what is really needed.

they should have tested 1v though, since the jump from .9 to 1.1 is large and there could be an incredible sweet spot between them.

It's possible they were going for the lowest possible stable voltage? Also can you elaborate on what you mean by it'll overclock with less voltage as long as the temps are maintained? Shouldn't the temps be a nonexistant issue if you're undervolting?

[sin0822]

most ivy OCes are low wattage compared to SB. BTW turn down CPU PLL Voltage, it helps. I also downclock the vtt and sa voltages to help with temps too.

[Manicdan]

It's possible they were going for the lowest possible stable voltage? Also can you elaborate on what you mean by it'll overclock with less voltage as long as the temps are maintained? Shouldn't the temps be a nonexistant issue if you're undervolting?

for your first question, i was looking at the max OC per voltage, they had 3.9 for .9v and 4.5 for 1.1v, but what if they got to 4.3 with 1v. that could offer 20% power reduction over the 1.1v 4.5ghz, while being only 5% slower.

when we overclock, we do it based on our cooling power
when a cpu is made its built for a cooling solution that can fail, resulting in temps of 100c (or whatever the cpu is limited to)

so if you try to oc for a small air cooling your looking at temps above 60c, if your on heavy water its sub 40c (depending on what chip we are talking about)

so now to the overclock itself. on air to get 4ghz you might need 1.05v because of the higher temp, for water you might be able to do it with .95v (all fake examples here). but lets say the pump fails, theres a 100% chance your OC will crash before you reach the peak temp where the cpu throttles (100ish C).

so irrelevent of what voltage you use, or what clocks you want, you really OC based on the cooling solution you have available. water users might be able to get a better low voltage OC because their temps can still be 20c under that of air for the same settings. but we dont see that often because most who use water just throw in lots of volts and try to push it past what an air cooler can do.

and 2 other points about their test:
they did not mention how they measured the power consumption, was it a TDP based draw, or total system draw.
and their PSU has 75% efficiency at 50W and 90% at 250W, so every value they have listed is between that range. meaning the lower overclocks are loosing 20% from the psu, and the higher overclocks loosing only 10%, making it SEEM like power consumption is not increasing as fast as it really did

[Halk]

Is 4.5 reserved for the 3770Ks or is it just as possible on the 3570s? In other words I've seen that the 3570's don't seem to want to go as far as the 3770s, but when it's only a moderate overclock of 4.5 does it matter?

[thegoatman]

It could be that a higher grade silicon made it to the 3770K bin... I read a review that mentioned their chip was. Manufactured in August 2011, which is relatively recent, suggesting Intel had a few bumps.

[PatRaceTin]

how about performance different between IVY & Sandy

at the same MAX. temp when overclocking with same cooling

[sin0822]

Is 4.5 reserved for the 3770Ks or is it just as possible on the 3570s? In other words I've seen that the 3570's don't seem to want to go as far as the 3770s, but when it's only a moderate overclock of 4.5 does it matter?

both as the issue isn't the voltage so much that the voltage pushed up the temps really high even without much heat being outputted.

4.5ghz is easy on every Ivy CPU that is unlocked, as you don't need more than 1.3v for it, and 1.3v is below the extreme temp barrier.

[Eldonko]

Interesting article, Im going to see what I can get at 1.1 as well.

[Leeghoofd]

My 3570K's need 1.23 for 4600Mhz prime95 stable... , but even 4500Mhz at 1.1 results in cores failing or freezing... either a sick CPU at Anandtech or OCCT doesn't stress properly... or I suck monky balls... take your fav pick (no dino you can not vote )

[sin0822]

My CPU that does 5.3ghz for valid on air can't even do 4.5ghz at 1.1v so ya go figure.

but I can do 5ghz at 1.2v at -190C

[Eldonko]

No 4.5Ghz for 1.1v for me, more like 4.3Ghz. 4320 is the absolute max for aida64 stability. Average load temperature measured over an hour of load is 46C. Should be a half decent test to quickly get the overclockability of a new chip.



Full sized image - http://images.hardwarecanucks.com/im...undervolt1.png

[BeepBeep2]

AIDA stress test IMO is a joke...

[Eldonko]

Maybe so, go ahead and prime your chip 24/7 if you prefer. When ASUS and Intel say aida64 is optimized for this platform and recommend using it then that is what Ill be using for all stability testing.

[sin0822]

only thing i don't like about AIDA is that they want to charge you for it. i think AIDA added AVX that is why Intel is in love with them hahaha.

But yea you guys know SVID only works till like 4.2ghz-4.3ghz as well. With Sandy you could get it to work really well, but with ivy not so much.

[Manicdan]

whats SVID?

[sin0822]

SVID is serial VID, it is a digital bus between the CPU and the PWM(which is the control chip for the VRM) and it basically allows the CPu and the VRm to talk and pick the right voltage for the frequency. CPU can select its own voltage on the go, max is 1.52v, but it never picks that high. It was introduced on the Intel side with Sandy Bridge, it is what allowed people to just increase the multiplier to 4.2-4.5ghz without needing to change the voltage. it is also what makes DVID offset not so straight forward b/c you can't control your stock vcore and you can't disable SVID. Because of SVID your VID at 3.3ghz and your VID at 4ghz or 4.5ghz are going to differ.