Intel Haswell Overclocking Clubhouse.

[Kursah]

I use Intel XTU and then if it passes that from 1-3 hours, I move onto AIDA64. I don't plan on using AVX so I'm not terribly worried about it...but maybe I should be?

Haven't had a whole lotta time to do much with my haswell build, but so far 4.5 @ 1.25v, cache at 8x-40x @ 1.18v, memory manually set to 2133 @ 1.58v (reads 1.60v) at 11-11-11-28 cr1, is so far AIDA and XTU stable! I am just happy as can be! I'm on air, and the highest temps I saw in XTU was 75C, and AIDA was 87C. Which on air and very warm ambient temp I can totally live with! Even my most stressful games won't match either test. Still want to fine tune...but my CPU input voltage is at auto and is reading 1.712v, if it stays stable I'm gonna leave it. All other voltages are auto as well. I did disable SVID. But beyond that I haven't done much beyond the CMOS clear and loading defaults and adjusting as necessary. This chip is AMAZING compared to my previous 4770k!

 

[Anusha]

But it seems a couple of games do use AVX. Crysis 3 and GRID2. I see spikes here and there when I play Crysis 3 with Adaptive Vcore. Don't have GRID2 to check it.

Plus, H.264 encoding uses AVX as well.

So you cannot say you won't use AVX. But definitely not as heavily as Prime95.

 

[Womper]

Where is this BIOS you speak of? I see 1205 which has been out since July 5th but nada on 1405. Maybe it's not for my PRO?

http://support.asus.com/Download.as...e=Z87-PRO-ASUS-1405.zip#Z87-PRO-ASUS-1405.zip

Anyone using Intel XTU for stress testing? It doesn't spike up the Vcore when using Adaptive Vcore. Does that mean ti is not using AVX instructions in its stress test?

Correct, Intel XTU does not use AVX instructions.

Then there is another issue. Yesterday I ran a long H.264 encode queue and the system rebooted (no BSOD, no crash dump, no bugcheck entry in eventlog nothing! This is not the first time either.) after 10hrs.

That happened to me sometimes also. You probably aren't 24/7 stable for AVX instructions.

I use Intel XTU and then if it passes that from 1-3 hours, I move onto AIDA64. I don't plan on using AVX so I'm not terribly worried about it...but maybe I should be?

Haven't had a whole lotta time to do much with my haswell build, but so far 4.5 @ 1.25v, cache at 8x-40x @ 1.18v, memory manually set to 2133 @ 1.58v (reads 1.60v) at 11-11-11-28 cr1, is so far AIDA and XTU stable! I am just happy as can be! I'm on air, and the highest temps I saw in XTU was 75C, and AIDA was 87C. Which on air and very warm ambient temp I can totally live with! Even my most stressful games won't match either test. Still want to fine tune...but my CPU input voltage is at auto and is reading 1.712v, if it stays stable I'm gonna leave it. All other voltages are auto as well. I did disable SVID. But beyond that I haven't done much beyond the CMOS clear and loading defaults and adjusting as necessary. This chip is AMAZING compared to my previous 4770k!

Awesome dude!

But it seems a couple of games do use AVX. Crysis 3 and GRID2. I see spikes here and there when I play Crysis 3 with Adaptive Vcore. Don't have GRID2 to check it.

Plus, H.264 encoding uses AVX as well.

So you cannot say you won't use AVX. But definitely not as heavily as Prime95.

Even a static voltage setting can handle a few AVX instructions sprinkled in a game that only uses a couple cores. The same reason that Per-Core overclocking works- if only 1 or 2 cores are getting used, you can run at higher MHz. When you start running a benchmark that hits all cores, then you need all that extra voltage.

 

[Big Texas]

With this stock voltage
http://valid.canardpc.com/2895895

shouldn't I be able to get 4.5 ghz under 1.25v?

i'm having weird 124 bsod's sometimes and can't tell what it is.

running 4.0ghz/3.9ghz cache 2400mhz memory @ 1.050v vCore stable right now with no tuning.

 

[cadaveca]

With this stock voltage
http://valid.canardpc.com/2895895

shouldn't I be able to get 4.5 ghz under 1.25v?

i'm having weird 124 bsod's sometimes and can't tell what it is.

running 4.0ghz/3.9ghz cache 2400mhz memory @ 1.050v vCore stable right now with no tuning.

Look at higher vCACHE, VCCSA, VCCAIO, VCCDIO. What are they at?

Under 1.25 V? Maybe. Might take a bit more, or a bit less, I think. Let's try.

How much mem, what model, and what are timings?

 

[Big Texas]

here's what my bios is giving me

1.040 vcore
1.032 vring
0.848 vSA
1.008 vIO Digital

it's G.Skill Ripjaws X 2400 mhz 2x4gb @ 1.65v 10-12-12-31 in xmp

 

[cadaveca]

here's what my bios is giving me

1.040 vcore
1.032 vring
0.848 vSA
1.008 vIO Digital

it's G.Skill Ripjaws X 2400 mhz 2x4gb @ 1.65v 10-12-12-31 in xmp

try 1.050V VCCSA and IO's Set Vring to 1.15 V for 39. try 1.265V and 45 multi on all cores.

 

[Anusha]

Dave, is playing with VCCSA, VCCIO only applicable if you have fast (i.e. 1866 or faster) RAM? Or can it still make things more stable if you are running them at 1333? What about having 4 sticks of 1600 CL9 1.5V ones like the Vengeance sticks?

 

[TheHunter]

but my CPU input voltage is at auto and is reading 1.712v, if it stays stable I'm gonna leave it. All other voltages are auto as well. I did disable SVID.

SVID is input voltage, if you disable it will stick at ~ 1.70V. Leave it at auto (usually settles at 1.79v). I end up at auto and its always at 1.79v.


And some of my OC experiences ,
Im now at 4.6Ghz for over 2 weeks and its still stable in everything, the only bsod at end was to low cache voltage.


At first
cpu x46 @ adaptive 1.239v (LLC 5)
cache x35-41 @ adaptive 1.115v
but it wasnt completely it..

now I have
cpu x46 @ adaptive 1.234v (LLC6)
cache adaptive x35-42 @ adaptive 1.122v (can go up to 1.148v in avx)
*in both cases ram XMP @ 2133mhz 1.65v 4x4 CL9

and yeah its great

Max temps per core 75-70-69-65C

 

[Womper]

Z87-PLUS/PRO/A/C BIOS 1405
-Improve system stability.

Nom nom nom. Time to revisit my little 4.8GHz project. And see if this has new microcode too.

Ran through some configs last night, and so far I haven't noticed any difference. 4.9 and 5.0 GHz still require the same voltages to boot, so no magic there. I was not able to drop voltage on my 24/7 either. I was able to run 4.8GHz @ 1.34v adaptive, but it is unclear if that's due to the new BIOS or the higher 1.89v Vin I used. On the previous BIOS, I had used only 1.8v Vin at this setting, and it wasn't stable even though 1.34v manual was.

No new microcode with this BIOS; my CPU is still on 9. Other motherboard brands introduced microcode 12 in their latest BIOS's.

Also, the 1405 BIOS chooses a TFAW of 20 for my 2133 memory, even though 1207 chose 32. I set this back to 40.

SVID is input voltage, if you disable it will stick at ~ 1.70V. Leave it at auto (usually settles at 1.79v). I end up at auto and its always at 1.79v.


And some of my OC experiences ,
Im now at 4.6Ghz for over 2 weeks and its still stable in everything, the only bsod at end was to low cache voltage.


At first
cpu x46 @ adaptive 1.239v (LLC 5)
cache x35-41 @ adaptive 1.115v
but it wasnt completely it..

now I have
cpu x46 @ adaptive 1.234v (LLC6)
cache adaptive x35-42 @ adaptive 1.122v (can go up to 1.148v in avx)
*in both cases ram XMP @ 2133mhz 1.65v 4x4 CL9

and yeah its great

Max temps per core 75-70-69-65C

Thanks for the info, I didn't realize adaptive cache voltage saw a little +AVX voltage like that.

 

[cadaveca]

Dave, is playing with VCCSA, VCCIO only applicable if you have fast (i.e. 1866 or faster) RAM? Or can it still make things more stable if you are running them at 1333? What about having 4 sticks of 1600 CL9 1.5V ones like the Vengeance sticks?

It depends on the CPU. @ 1600 MHz, I would be looking at timings as a fault rather than voltages, however. At the same time, I do think that all voltages in the chip kept within specific ratios of each other can be beneficial as well. For example, some CPUs have stock cache voltages higher than CPU voltage, and vice versa, VCCIO's and VCCSA are different on different chips, too, as is the FIVR voltage. SO there's no real general rule for these things, we all should have slightly different, although similar, experiences.

Thanks for the info, I didn't realize adaptive cache voltage saw a little AVX voltage like that.

Seems to depends on board and BIOS, the best I can tell.

 

[TheHunter]

Yeah at first it was always ~1.045v - 1.128v (Aida64 monitoring), when I tried to encode or test x264 benchmark I noticed few spikes to 1.148v (by both x41 or x42 cache).

Adaptive 1.110-1.118v wasn't enough and bsod 0x124 or random reboot.

I guess its similar to normal cpu adaptive but with very small boost 0.02v max.
Or like Dave said maybe its a DIGI+ thing

 

[cadaveca]

Or like Dave said maybe its a DIGI+ thing

Yeah, some boards offer just offsets. some offer manual settings. The differing options and what voltage mode is used is going to affect it, I think. If the multi is static, and so is the voltage, no boost, like CPU?

 

[TheHunter]

I think, I didnt test it like so yet.


I switched from fixed to adaptive and always thought it doesnt go over like by fixed voltage.. But then again 0.02v is not much



(x35 cache) web browsing idle voltages and finally end of summer cool temps

 

[Mull1s]

There is a good guide by sin0822 at

I recommend trying 45 before 46 and gradually increase vcore
I would suggest clicking CPU and moving memory lower until CPU stable.
For me, I lower uncore to 4.0 at those ranges

Increase cache voltage, try 1.25v adaptive or add more offset.
Increase VCCin to 1.85-1.89.

And remember, a 1.332v adaptive voltage means your chip will be at 1.45v during AVX workloads. If you plan on encountering AVX stuff, then you will probably need to increase VCCin to the 1.9v range, maybe even 2.0v.

For 24/7 I decided to stay under 1.4v even after the +AVX voltage. Also, I have only had success with adaptive voltage in the 1.2x range; when I go after higher frequencies that require static voltages in the 1.3x range, I cannot get a stable adaptive setting, even if it's much higher than the static voltage.

Thx for the good info, i am now running a stable 4.5Ghz Clock but i would really like to reach 4.6.
What i did was raising Vccin to 1.88 and changed my adaptive volt to 1.275v manual.

Im thinking maybe some of Asus Digi+ VRM settings could help me provide some stability?
They are untouched at the moment!

Picture below shows the options and this isnt what i got it set!

Could i change something here to reach 4.6?
Thanks again!

 

[vega22]

just wanted to post and share my first hw oc with you

http://valid.canardpc.com/2897426

nothing to right home about but i wasnt playing for long with the oc, had other issues i needed to work out xD

 

[Anusha]

if i set the memory to 1600, i would get rounding errors (not BSODs) in in-place large FFT of Prime95 :-/ (i run large FFT with 3 minute each test as it seems to catch instability fastest with this chip.)
1333 seems to work fine.

tried the following
1. increase Vmem to 1.65
2. increase VCCIO by +0.2 offset
3. increase VCCSA by +0.05 offset
4. increase VRING from 1.08V to 1.1V
5. increase Vcore from 1.25 to 1.26V (usually 1.24V is stable at least for many hrs)

any ideas? should i have increased VRING more?

I have two sets of these RAM.

maybe this board/CPU doesn't like these RAM. the RAM only supports XMP v1.2 while the current version is v1.3. Maybe it sets some timings incorrectly?

 

[vega22]

i would give the ram a bit more juice before i looked at other things. just like the vcore the vdimm will have vdrop and droop.

i would go to 1.68 and still consider it "stock" volts tbh.

 

[Womper]

Thx for the good info, i am now running a stable 4.5Ghz Clock but i would really like to reach 4.6.
What i did was raising Vccin to 1.88 and changed my adaptive volt to 1.275v manual.

Im thinking maybe some of Asus Digi+ VRM settings could help me provide some stability?
They are untouched at the moment!

Picture below shows the options and this isnt what i got it set!

Could i change something here to reach 4.6?
Thanks again!

Personally, I haven't noticed any impact from tweaking the Digi settings.

if i set the memory to 1600, i would get rounding errors (not BSODs) in in-place large FFT of Prime95 :-/ (i run large FFT with 3 minute each test as it seems to catch instability fastest with this chip.)
1333 seems to work fine.

tried the following
1. increase Vmem to 1.65
2. increase VCCIO by +0.2 offset
3. increase VCCSA by +0.05 offset
4. increase VRING from 1.08V to 1.1V
5. increase Vcore from 1.25 to 1.26V (usually 1.24V is stable at least for many hrs)

any ideas? should i have increased VRING more?

I have two sets of these RAM.

maybe this board/CPU doesn't like these RAM. the RAM only supports XMP v1.2 while the current version is v1.3. Maybe it sets some timings incorrectly?

You might need to go over your memory's timings in the BIOS and set them to what the RAM is actually rated for. Maybe you can toss the memory into a different computer and run Memtest to make sure they haven't gone bad?

 

[TheHunter]

I tweaked DIGI+ a bit and this works ok at 4.4 - 4.6Ghz.

lowered LLC to 6 (auto is 8)
Set both cpu & ram phases to Optimized and kept VRM spread enabled
Raised Cpu Current to 120%

rest at auto or default

 

[Anusha]

You might need to go over your memory's timings in the BIOS and set them to what the RAM is actually rated for. Maybe you can toss the memory into a different computer and run Memtest to make sure they haven't gone bad?

primary timings are set properly. 9-9-9-24-2T no big deal. i'll remove two sticks and see.

 

[cadaveca]

primary timings are set properly. 9-9-9-24-2T no big deal. i'll remove two sticks and see.

it's the other timings that matter on haswell, since some have changed compared to previous paltforms.

 

[vega22]

it's the other timings that matter on haswell, since some have changed compared to previous paltforms.

dave i have that many sub timings now i dont know where to start

clock for clock 1150 seems to be worse than 1155 for me atm

update

 

[Anusha]

it's the other timings that matter on haswell, since some have changed compared to previous paltforms.

how do i know which ones to change to which ones?
they are not usually listed in RAM's specs, right?

edit:
wait a second. i'm not sure if i had the RAM at AUTO timings + 1600MHz or 1600MHz XMP. I've just set it to 1600MHz XMP and now checking if i get rounding errors. hold on.

 

[Big Texas]

curious how you guys are running your VCCIN voltage...

i'm at 1.7v for 4.5ghz core@1.25v/4ghz cache@auto volts with ram @ 2400 mhz 1.65v.