synphul :
According to your post where you quoted figures, it again goes back to 'to each their own'. I won't tell you not to run at 4.7ghz with 1.375 on a haswell/d.c. chip. I can tell you I surely won't run mine that high on the vcore. 100mhz isn't worth the .1v on the vcore to me. There's more to overclocking than thermals and stability. Vcore has to be taken into account as well. Regardless if a system is running at 50c on p95 and never crashes under realbench or anything else I wouldn't be comfortable pushing it so close to 1.4v. We're talking a 100mhz (or by your quoted figures 200mhz) difference, so if by those figures alone p95 were in fact the only thing limiting the oc (vcore ignored) - what is someone actually losing? It's not like it's shorting that particular system of 400-500mhz it would've had otherwise.
I stated that that P95 was useless because it artificially limits your overclock. I have asked you 3 times to answer the question and 3 times you ignored or provided a limited response to side step the question. If you can not provide a answer, then the statement stands w/o a counterpoint.
What is the purpose of using P95 to "thermally" test your CPU to determine if your CPU and cooling system can handle a specific OC when nothing you ever do again in the life of that PC will ever bring it to those levels. ?
If I limit my OC on basis of a) stability and b) temps < 80C .... which serves me better.
1. 4.5 Ghz @ 1.275 volts gets me 71, 73 ,74 ,79 under P95
2. 4.7 Ghz @ 1.375 volts gets me 72, 74, 75, 79 under Rog Real Bench
Nothing I will ever run will approach the load of RoG Real Bench .... ever. So why limit my OC to 4.5 ? What does one gain by using P95 ?
Every chip varies and you said you were limited by voltage not temps..... so you didn't do so well in the silicon lottery but that doesn't cover every situation .... what if your voltage was 1.2 ?
What if you had a 4770k where your temps would be 10C higher ????. You created a situation using your own set of limited criteria to side step the question.
-You're analogies are not on point. If you want use air pressure in a car tire, then the proper analogy would be testing it to 60 psi to see if can hold 36.
-Using 12 different synthetics still fails to test the CPU in a multasking environment and let's also remember you are talking about 26.6 which does not even test the modern instruction sets. All you get out of P95 26.6 is being able to say that you're system is P95 stable. AVX results in a 30% performance increase in P95.... in other words you're testing only 70% of the CPUs potential performance. So why include it amount your suite of synthetics if it tests only part of the CPU and IBT gets it much hotter ?
Mine for instance is voltage limited before it's heat limited. ... My thermals were still under 80c on p95 at 4.6ghz, but to get to 4.7ghz without crashing I had to push my vcore from 1.28v to 1.34. Something I wasn't comfortable running 24/7, personally.
What if a user says he isn't comfortable over 70C , your position doesn't hold. You're not comfortable with 1.34v ???? Your P95 26.6 test does not test AVX nor the other modern instruction sets. When it is, your system will bump VCore by 0.13 or better. So, sorry to inform you but your system is routinely breaking 1.4 every time an AVX instruction is present.
It's been stated on many overclocking forums that p95 is not a conclusive stability test, which is why others like xtu, ibt and realbench are also recommended.
Again let's stay on point..... you are professing 26.6 not 28.5. XTU is P95 Lite....P95 is useless as a stability test because it does not test all the CPUs instruction sets..... If you are using the old 26.6 then obviously you are also using the old IBT which also doesn't contain AVX..... So using 142 programs that don't contain AVX or any of the other instruction sets means your CPU is likely stable except when any of the new instruction sets are present, when the other 30% of the processor's capabilities are used and except when used in a heavy multitasking environment.
I did extensive testing on this matter and repeatedly experienced OCs being, P95, AIDA, Linpack, OCCT, XTU stable ..... and then failing under RoG Real Bench. So what did I accomplish by using P95, AIDA, Linpack, OCCT, XTU ? I wasted a lot of time. I later switched things up and started using RB 1st ... after establishing a stable OC w/ RB, I have never failed under any other stress test. With RB, I top out at 78C, P95 (26.6), I'll see 84C, P95 (28.5) I'll see 86C. Linpack 79C, XTU, 72C. IBT, will go straight to 99C
According to your post where you quoted figures, it again goes back to 'to each their own'. I won't tell you not to run at 4.7ghz with 1.375 on a haswell/d.c. chip. I can tell you I surely won't run mine that high on the vcore. 100mhz isn't worth the .1v on the vcore to me. There's more to overclocking than thermals and stability. Vcore has to be taken into account as well. Regardless if a system is running at 50c on p95 and never crashes under realbench or anything else I wouldn't be comfortable pushing it so close to 1.4v. We're talking a 100mhz (or by your quoted figures 200mhz) difference, so if by those figures alone p95 were in fact the only thing limiting the oc (vcore ignored) - what is someone actually losing? It's not like it's shorting that particular system of 400-500mhz it would've had otherwise.
That's the point .... you already are..... I'm fine with "to each their own", it's your reasoning that is faulty.
1) You "don't want t go near 1.4 volts" but you already are breaking 1.4 volts. 1.28 + 0.13 (or more) = 1.41 ... that's where you are at every time AVX is present
2) You "don't want to have high 24/7 voltages". You don't, neither do I .... as I am typing this, my cores are at 0.704 volts. During gaming they are typically 1.2 something....in image editing (handbrake), I get up to 1.32..... only when I use Open CL does it get any higher....with Open CL present, my 4770k can go as high as 1.5 .... for a millisecond every now and then. An event that causes me no concern whatsoever
3) If you are using 79C under P95 for thermal testing, what's the point ? You said you are also using IBT so that means you're seeing high 90s. But did you build your PC to run synthetics or actual applications ?
That's why the MoBo manufacturer's give you the tools you feel the need to address this concern. I can use the 46/43 profile as the normal boot, then if I have anything that needs Open CL that I need to run extensively on a particular day, I can just boot to the 45/42 profile.
I should also note that a significant part of the required voltage increase is due to keeping the cache multiplier with the recommend 3 of the CPU multiplier. Leaving the cache multiplier at 39 reduces the need for Vcore and Vcache but hurts performance in image editing and certain other programs. Of course if you're just looking to validate your OC rather than get the best performance, you'll leave the cache at 39.
P95 is meant to test the performance of whatever cooling setup you have. It is not meant to be a stability test
It is certainly is in fact "
meant" to be a stability test. It says so right on the mersenne.org website and every website you download it from as well as just about every overclocking guide ever written and you'll find thousands of enthusiasts sites recommending such. You might not think it is enough,
I think it's totally useless for other than setting TIM but that doesn't change history or the perception of the majority of users who have used it over the last 20 years (19.5) for exactly that purpose. Just read wiki
https://en.wikipedia.org/wiki/Prime95
It is my position that it is no longer useful for this purpose since peeps are afraid of using any version that tests all of the CPUs instruction sets. Using P95 ....
1. Does not show whether OC is stable when those instruction sets are present
2. Does not show whether OC is stable while multitasking
3. Is useless as a cooling performance evaluator (unless you built the PC solely to run synthetics and get your name on website leader boards) as it can artificially limit your OC because temps can be higher under synthetics than with any combination of real programs. More then ever before with HW and DC, the amount of voltage bump and amount of temp increase we see is exponentially larger with each 0.1 Ghz increase. The "artificial" limit the P95 can leave you wthl a 4.5 OC when you easily could have had 4.7 and not exceed your self impose temperature limits using any combination of "real applications". P95 became less relevant with SB, and to my eyes, of no value with HW.
4. P95 is very useful for setting your TIM, cycling it up above normal operating temps and back to room temp in a short time....not for determining maximum safe OCs "using real applications".
Facebook
Twitter
Instagram