4670k overclocking

[braveheart0707]

So after reading couple guides i started dirty overclocking my 4670k (fast test to see if i have a good chip or not).

Vcore 1.2V, 4.6GHZ, Cache Ratio 38, Cache Voltage Auto(1.035)

To my surprise ( in the past i've always had bad chips), not only do i pass the post and gets into windows, i managed to run Intel ETU for 10 minutes, with temp not exceeding 75 degree.

Does this mean i have a good chip?

 

[Burpo]

I would say so..

 

[braveheart0707]

Say i pass more stable testing at this setting,
how important is raising the Cache Ratio to be higher? I know the optimal setting is 1:1.
Is there alot of performance hit with cache ratio as low as 38?

 

[Jacky60]

Sounds a good one, I think Intel may have improved the process since launch keep going and good luck getting it stable!

 

[bgt]

Try it with Intels XTU. It shows how hot it can get when OCing it and doing a stress test also. My 4770K started throtteling when I put the Multi on 40+turbo steps at 41,42,43.
The Vcore was then at 1.25V. So my 4770K is not really a nice chip for OCing. Temp delta between cores is 8C so this looks like bad TIM placement. I now have it on multiplier 36+turbo 37,38,39.
It then just stays under 100C when using stability mode testing. No WHEA errors Delidding comes to my mind................
Picture shows running prime95:

 

[BallaTheFeared]

Say i pass more stable testing at this setting,
how important is raising the Cache Ratio to be higher? I know the optimal setting is 1:1.
Is there alot of performance hit with cache ratio as low as 38?

No, focus core.

I think I read some place that you need 800MHz of cache to match 100Mhz of core.

Core is king..

Set your cache to 36 or 38, then find your max core clock with voltage you find acceptable then go back to cache until it causes your core to be unstable.

Then move on to ram if you want to OC it, best to start at 1600 even if you have a 3000+ kit.

 

[bgt]

Best setting so it does not throttle. Its loaded with prime95 FPU test.

 

[bgt]

Here is my i3-4130 system with stock cooler. As you can see it throttles at default. Load is Prime95 FPU test

 

[bgt]

4770K on 3.5Ghz and Furmark running. IGP only. Also Prime95 FPU mode running.
It throttles like crazy now. Very bad thermal design. Can anybody verify if this is normal?
Or do I have a bad chip? All this after 1,5 minute after start.

 

[BallaTheFeared]

I'd say unrealistically stupid testing parameters are at play.

Why are you running two power viruses and expecting different results?


What "real world" load are you trying to represent running that?

 

[bgt]

Just to see if I have a bad chip. I will test my 3770K now to see if this is normal under stress.
I wanted to buy an i7-4820 but was stopped by the old chipset. Pity.

 

[BallaTheFeared]

Prime95 supports AVX2/FMA3, Ivy Bridge does not.

You'll be comparing apples to oranges.

 

[bgt]

You're right. Is there a way to switch off AVX2/FMA3 support?
If the 4770K is sold with AVX2/FMA3 support shouldn't it be build to handle it?
It means that whenever AVX2 is used it will hit high temps.

 

[bgt]

3770K(with stock cooler) with Prime95 FPU and furmark. What a difference. Haswell looks like a design failure concerning heat(AVX2).

 

[ShintaiDK]

3770K(with stock cooler) with Prime95 FPU and furmark. What a difference. Haswell looks like a design failure concerning heat(AVX2).

What do you base that on? The only time I see your CPU throttle is when it runs above stock speed. or did I miss something? And remember turbodrop is not throttle.

 

[BallaTheFeared]

3770K(with stock cooler) with Prime95 FPU and furmark. What a difference. Haswell looks like a design failure concerning heat(AVX2).

You're putting a lot more load on the cpu, the throughput is likely 1.7 to 2x the throughput of Ivy Bridge.

 

[bgt]

What do you base that on? The only time I see your CPU throttle is when it runs above stock speed. or did I miss something? And remember turbodrop is not throttle.

It throttled also at 3.5Ghz, look at pic. But if its all due to AVX2 taxing the CPU.....so be it.

 

[ShintaiDK]

You're putting a lot more load on the cpu, the throughput is likely 1.7 to 2x the throughput of Ivy Bridge.

Yep, 4670 for example is roughly 75% faster than 3570 in these loads.

 

[ShintaiDK]

It throttled also at 3.5Ghz, look at pic. But if its all due to AVX2 taxing the CPU.....so be it.

In the one where your cache frequency runs 1:1?

 

[bgt]

You're putting a lot more load on the cpu, the throughput is likely 1.7 to 2x the throughput of Ivy Bridge.

Those are nice figures. Did you test it fully loaded with prime 95 FPU and furmark?

 

[bgt]

In the one where your cache frequency runs 1:1?

New measurement with cache 1: 1 and turbo at the same freq.

 

[ShintaiDK]

New measurement with cache 1: 1 and turbo at the same freq.

I think you misunderstood. The cache speed is decoupled with Haswell. So your 4770 result got overclocked cache speed.

 

[ShintaiDK]

On a funny note. Be careful using the XTU utility on non Intel boards as a way of measuring. This is from my PC, and its not photoshopped.

It swings between anything essentially. I am sure I could make a CPU-Z validated too, if hit the botton at the right time. WR anyone?

 

[bgt]

I think you misunderstood. The cache speed is decoupled with Haswell. So your 4770 result got overclocked cache speed.

How can this be? All settings are on 3.5Ghz. As you see cache ratio is also 35x.

 

[ShintaiDK]

How can this be? All settings are on 3.5Ghz. As you see cache ratio is also 35x.

Either wrong information or yet another factory OC setting.
http://www.anandtech.com/show/6355/intels-haswell-architecture/10