i7 4770k new rig. Give me some feedback. :)

Hi guys, I am making a new rig for my son.

CPU: i7 4770k Vcore 1.36V @ 4.5Ghz
Cache Ratio x44
VRing: 1.36V
VCCIN: 1.88V

Using Gskill 2400Ghz 16Gb Ram
2x GTX Titan

Now, thing I want to discuss is the Haswell CPU.
I see some people getting over 4.7Ghz with my voltage.

I however run into heat problems, which I know haswell burn quite hot without delidding (I hadn't delid, might later). I also occasionally get a BSOD 0000124 if I go to 4.7Ghz, which tells me I am likely short on Vcore there...

Now, I know a lot of people using AIDA64 instead of P95.

is P95 still the way to go to stress test a CPU?
or has it been slowly replaced like how it was with IBT?

Also, is Realbench from ROG anygood for stress testing?

Just an old man trying to keep up with the times, opinions and discussions are welcomed.

Check this out.

http://www.overclockers.com/forums/showthread.php?t=732646

It depends on the chip. I haven't spent a lot of time OC'ng my 4770k. But I've had no success getting it to run at 4.5Ghz stable for Folding. So I leave it at 4.1 24/7.

Every CPU is different. That does seem like a below average voltage for that clockspeed though.

At that clock, you should not have to adjust ring or especially VccIN...that should all work fine at stock voltage.

EarthDog said:

Every CPU is different. That does seem like a below average voltage for that clockspeed though.

At that clock, you should not have to adjust ring or especially VccIN...that should all work fine at stock voltage.

Click to expand...

I thought so too. I guess my chip aint that great.

but without Vccin and Vring voltage boost, it keeps crashing out on P95. (albeit fine on daily operations as far as normal applications go.)

hence my second question, about whether P95 is still considered a good stability test or not....

Great or not, the VccIN really shouldn't need touched...

The newest version of P95 has the latest instruction sets, yes, so it will be fine.

I use AID64 personally.

well, just tried I put VCCIN back on auto.

and 5mins into P95... bsod 00000124

so I guess for now I need it. and yes, I am not sure why either.
I read somewhere that having VCCIN >0.4 of Vcore will stabilize the chip's overclock
and sure enough.. appears to be the case for now.

I guess I will start to use AIDA64 myself too. never tried.

I've been experimenting with different stess test programs lately. To be fair, passing two hours of P95 blend is still the gold standard for demonstrating stability in everyday computing. However, IntelBurntest when run on the high setting drives up temps to the max much more quickly than P95. They both give about the same max temps but it takes IBT only about 5 minutes to get there. Can be handy when testing more for max temps than for establishing stability. For whatever reason, I can pass a long IBT and still fail a long Prime95 blend so that is to be kept in mind.

I have also found that if I can complete the latest Futuremark 3D demo/bench run the machine is pretty stable for everyday computing, even though it may not be quite Prime95 blend long test stable. Actually, I have found Futuremark 3D to be perhaps the most sensitive instrument for revealing a CPU overclock that is not really stable, believe it or not. It doesn't drive core temps up very high but the bench will abort, usually during one of the physics tests if the CPU is not pretty stable.

My parting thought is that it may be wisest to stress test an overclock with a variety of tools instead of just one.

P95 Blend tests memory hard. I don't think IBT does? I havent used it in years because to me, its overkill (akin to a car hitting the redline with P95, but the fuel cutoff/rev limiter with IBT/Orthos/Linx, etc). Usually those applications not only heat up faster, but are several degrees higher than P95. It has been a while since I really paid any attention though.

@BZ5

Does your son's system require a 4.5GHz overclock? I wonder how much lower the voltages would be with your combo to run at 44x multi and ~41x cache?

Yes Prime95 does run cooler than IBT-Orthos-Linx, one thing I like about Prime95 it's free.

PolRoger said:

@BZ5

Does your son's system require a 4.5GHz overclock? I wonder how much lower the voltages would be with your combo to run at 44x multi and ~41x cache?

Click to expand...

haha, of course he doesn't need 4.5Ghz, he will probably be fine with 4.2Ghz at 1.15V. But then, my overclocker blood goes wild when I can bench 3Dmark with 2 Titan Blacks. So I am trying to 'optimize' his system while I am at it.

for my existing system on X79, I have always used p95 as benchmark, and sure enough, if it pass p95 for 4 hours, nothing will crash it. I am just wondering is this application still applicable after 2 generations of chips. And I m glad to see wingman and trents still recommending it, I trust these men. (and Woomack too.)

I was on ROG forum, they are recommending ROG realbench v2.2, which is a 15min test, and they recommend 3 runs will test stability. One thing I worry.. is the temp didn't get that high on that test. (talking p95 brings me to 88'C, RB brigns me to 77'C) so I am not sure if the stress is really enough... though sure enough, for the time being, it hasn't passed a system that I crash later ... yet... my gut feeling tells me exception about to come when I stress test 4.7Ghz tonight.

I guess I stick with p95 for now then. It just feels like an old trusted friend.

btw, anyone else has comment on 4770K OC?

how important is cache ratio? any difference in benchmark that is worth optimization to 1:1 with CPU freq?

EarthDog said:

P95 Blend tests memory hard. I don't think IBT does? I havent used it in years because to me, its overkill (akin to a car hitting the redline with P95, but the fuel cutoff/rev limiter with IBT/Orthos/Linx, etc). Usually those applications not only heat up faster, but are several degrees higher than P95. It has been a while since I really paid any attention though.

Click to expand...

IBT has several levels of stress intensity and they vary according to the amount of ram tested along with the cores. The "standard" level is less intense than Prime95 blend. The "high" level seems to be essentially the same as Prime95 blend, testing 2 gb of RAM. IBT also has a "very high" and an "extreme" level of stressing as well. I like two things about it. One is it heats things up very quickly and the other is you can input the length of the test, which you cannot do with Prime95.