Choosing the best i7 4770k

smsmasters · 12 Dec 2013 at 20:31

Right, so I have two 4770Ks.

Please suggest the best methods to find out which chip is the best for overclocking. I'm going to return the lesser chip.

Nickolp1974 · 12 Dec 2013 at 20:57

Leave everything at stock but set ram to xmp and keep upping the multi and boot into windows, whichever chip gets the highest multi keep that one. It only needs to just get into windows.

smsmasters · 16 Dec 2013 at 22:52

Right,

My first 4770K has a low VID of ~1.040V but runs very hot at 4.5GHz @ 1.3V. max temps around 97C, cannot really get it stable...

My second 4770k has a higher VID of ~1.106V but runs cooler at 4.5GHz @ 1.35V, max temps around 80C. Currently passing 4 hours of P95 with the latest AVX extensions.

Which one would be better to keep?

setter · 16 Dec 2013 at 22:55

Would you ever be tempted to delid? If so id keep the 1.3v one. 1.35v is a bit on the high side for theese. But tbh neither will hit temps that high in normal use.

smsmasters · 16 Dec 2013 at 22:59

setter said:
Would you ever be tempted to delid? If so id keep the 1.3v one. 1.35v is a bit on the high side for theese. But tbh neither will hit temps that high in normal use.

Probably once I get watercooling.

I'm currently using the Phanteks PH-TC14PE CPU Cooler.

I've got a feeling the first CPU has poor IHS to die contact.

setter · 16 Dec 2013 at 23:01

Mine is a very hot runner too, 1.265v but hits 94c in p95 avx. My k2 cooler is pretty comparable to the phanteks. I may delid in future as well.

smsmasters · 16 Dec 2013 at 23:02

setter said:
Mine is a very hot runner too, 1.265v but hits 94c in p95 avx. My k2 cooler is pretty comparable to the phanteks. I may delid in future as well.

Do you know what your stock voltage was? That's a great voltage for 4.5GHz!

wazza300 · 16 Dec 2013 at 23:04

You want the lowest voltage chip imo,even if it runs hotter that's down to intels crappy paste

setter · 16 Dec 2013 at 23:09

smsmasters said:
Do you know what your stock voltage was? That's a great voltage for 4.5GHz!

Screenshot of it running p95 at stock.

smsmasters · 17 Dec 2013 at 01:45

setter said:
Screenshot of it running p95 at stock.

Nice.

I'm keeping the second CPU, it runs cooler even at higher voltage.

setter · 17 Dec 2013 at 09:00

Cheers. Tbh if i were you id still keep the lower voltage chip despite it being hotter. Your normal use temps wont be as high as stress testing. And if you do water cool you could lower them more. 1.35v is a bit high for vcore for a 24/7 oc.

smsmasters · 17 Dec 2013 at 09:55

I did some more testing earlier and the lower voltage chip BSOD even at 1.35V @ 4.5GHz. It appears that it needs 1.4V just to pass P95 w/ AVX.

Havana_UK · 17 Dec 2013 at 10:18

My i7 is very similar to Setter's. Was 1.044 VID at stock, and I have it stable at 4.5GHz @ 1.27V almost exactly the same as Setter! I couldn't get it prime stable at 4.6GHz without loads more voltage so I figured 4.5GHz is plenty anyway.

Stress testing temps go into the low-to-mid 90s, and an AIDA64 run saw one core hit 99C and throttle back, but in day-to-day use (e.g. BF3) load temps don't exceed 68C. This is all with a Megahalems (rev B) with a single Akasa Apache fan on it.

I'm sure you're right about the one with the lower VID having poor IHS/die contact because of crappy paste, but that is possible to rectify with a delid (if you're brave enough!), however there's nothing you can do to reduce the voltage that a chip requires.

Personally, I'd keep the one with the lower VID and settle for a lower clock speed. After all, if it's stable at 4.4GHz and <1.3V do you really need the extra 100MHz? I presume you're not competing in benchmarking championships?

Doug2507 · 17 Dec 2013 at 12:21

Keep the 1st one. Voltage will kill a chip a lot quicker than temps. For running the cpu 24/7 you're not going to be running stress tests all of the time so temps should be fine (especially 28.1!). It'll also give you an opportunity to de-lid when/if needed and you'll squeeze a higher OC out of it than the second one.

The second one isn't the best imo, 1.35v for x45 is high and i'd put money on it not being able to get past x46, even then it'll need over 1.4v to even attempt to get x46 stable.

With regards to using Prime for stability testing, don't, unless you are using an application day to day that requires that level of stability (i.e, work). There's absolutely no point stability testing with it for regular use. Try something like XTU bench, x264, ROG Realbench etc for a quick check then run your normal applications/games (apart from BF4!). Prime requires more voltage than is needed for gaming etc. If you absolutely need to use prime run small FFT only to test core.

Also make sure you're adjusting vrin/vccin/input voltage with vcore. If it's too low you can screw the vcore up as high as you want and it'll still BSOD. Haswell can be quite picky with voltage settings and when you reach the upper limits of the chip you need to juggle a few voltages to get it running smooth across the board. (i.e, core/uncore/ram). It's highly likely that 1.35vcore will need somewhere around 2.0vrin for stability, give or take.