Where's your 2500K voltage jump?

[Jman13]

So I've been experimenting with my 2500K overclock, and I've noticed that somewhere between 4.2 and 4.4 GHz (I haven't actually tried 4.3 yet), there is a noticeably jump in voltage for true stability. At 4.1, I could run 1.22V all day with no issues. At 4.2, I run at 1.23V (where I am now). I tried 4.5 a while back, but needed 1.34V for that. I tried 4.4, and while I can do most things at lower voltages, it BSODs in Prime 95 at 1.27V, and at 1.28V-1.32V, there is error correction required for IBT to complete (full burn slows a bit and Linpack speeds are slower....when fully stable, the GFLOPS are much better than when it needs to go back and correct stuff). So I'd need at least 1.33V at 4.4GHz...a full 0.1V higher for 200MHz.

I know each chip is different, and I've heard of people having like 1.3V at 4.5GHz, but they need 1.45 at 4.8 or something.

Where does your chip make a jump in needed voltage for stability?

 

[Blades]

Could also depend on the PSU and mobo.. 5ghz @ 1.384,1.392,1.400 (what I see via CPU-Z, variable).. I use "(max?) additional turbo voltage" paired with a negative offset.. This is at 100% load (7z benchmark, 128MB dictionary size, 8 threads). Haven't tried to go with anything higher.

I know what you mean, I have two tegra2 devices that hit a wall (or point of greatly diminishing returns) at a given frequency.. 1.652ghz @ ~1.45vcore for TF101 (asus), and 1.7ghz @ 1.375(viewsonic gtab). My Phenom II 1090T also seems to have hit a wall at 4.1ghz.. Theres a certain point where more vcore does not equal more clock speed.. and that point is usually very obvious.

 

[Idontcare]

Here's the voltage curve for my 2600K

One thing to understand is that the minimum voltage necessary for stable operation at any given clockspeed is both clockspeed and temperature dependent.

So when you see a "knee" in the voltage curve where it seems like higher clockspeeds are taking a disproportionate increase in voltage you also have to consider that part of the reasoning for that higher voltage bump is due to the operating temperatures increasing from the clockspeed bump itself.

If you could keep the temps in check then the voltage curve for clockspeed would be much shallower.

 

[Jman13]

Well, I can 'pass' IBT at 4.4 GHz and 1.28V. The problem is that on the "Very High" setting, after 4 runs or so, it begins to slow down, presumably due to error checking. It's not throttling speed, just going back and redoing calculations, slowing things down. But, I see your point. Once I get to 1.32-1.33, the heat starts to build. I am at 72C max on IBT, and 67-68C on P95 at 4.2 and 1.23V. At 1.32V and 4.4GHz, I hit 80-82C in IBT and low 70s in P95.

It is interesting to see that the inflection point on your voltage/frequency curve on the 2600K also happens right around 4.3GHz, though.

 

[Absolute0]

4.7 1.25
4.8 1.3
5.0 1.4
5.1 1.5

so my jump comes going past 5

(edit: i don't run prime95 longer than 30 minutes because I've found it pointless)

 

[Jman13]

Wow....golden chip....

 

[Blades]

Well, both 5.0ghz @ 1.4V (Abs0 and I) are using water cooling.. Although, mine is using H80 and sitting on a counter.