RVE10: OC Question regarding Cache Voltage

[dialsc]

Hi,

I've got a question regarding something strange I'm facing. I'm in the process of tweaking my new RVE10 based build, acctualy performing OC. One aspect is the cache ratio and therefore I'm using manual voltage settings for the cache voltage. But there is one thing I don't get. If I put in a value of 1.3 volt, once in Windows, using a software like CPUID HWMonitor, I see that the voltage applied is 1.331.

So, do I forgett something I don't know of, yet? Is this normal? My understanding is that if I configure a value of 1.3 V it should run with 1.3 V, shouldn't it?

Cheers,

dialsc

[Chino]

Are you sure you're not looking at the Vcore value?

[dialsc]

Yes, absolutely. Well, as long as HWMonitor collects the information form the right "sensor"... 😉 Just kiding. Yes, it is the right value I'm looking at. I see the increases/decreases being applied to this value when I make a change to the Cache Voltage.

[Menthol]

I would assume there will be some tolerances both in applied voltage, sensor reading, and APP reading, if you want exactly 1.3 set 1.28 or 1.29 and see if that ends up being 1.3,
I wouldn't make a big deal out of small tolerances, nothing is exact
my question is why would you want to use a manual voltage for cache, why not offset so the value would decrease at idle and increase to 1.3 during load

[dialsc]

Well, whether or not this is due to some tolerances is something I cannot say. I'm just not experienced enough. This is the second time I try to do overclocking and the first time I do it using something like Excel to do smallest increments in voltages asf. just to track and get an idea about what I'm doing... 😉

Would you consider this to be tolerance related?
CPU Cache Voltage:

- Configured: 1.315
- Value shown (even) in BIOS: 1.331

If this is really a negligible tolerance, I most properly making things much more complicated than I have to... ;). For example for the CPU Voltage I do a rough dial in starting with a dedicated guess and afterwards try to find the voltage which works e.g. 2 or 3 hours in Prime95. I do this by raising the voltage 0.001 every time until it works and then add another 0.01 V to be on the safe side.

I decided to go with manual voltages in order to figure out what I really need and once I got it running nice and stable, change to offset or adaptive which is still to be decided. I thought this might be a reasonable approach to follow. Wouldn't you agree on that?

Well, this whole thing is rather extensive and time consuming but because I do not have the in deep knowledge in this topic at least I have it black on white what worked and what didn't... 😉

But back to the topic. If I got you right you suggest to put in a voltage value which will then read out in the BIOS as the target voltage I'm after? Sounds somehow strange to me, to be honest. Because e.g. the CPU Voltage reads in the BIOS exactly the value I configure.

[Menthol]

Your going about things in a good logical way recording settings like you are, I am not an electrical engineer but in my mind a couple hundredths of a volt is marginal, but that's just me. I would recommend a different stress test but you can use whatever you like, be careful using Prime95 it is not recommended these days, RealBench is a better indicator for stability
Just my opinion, better wait for an expert to reply to your question, there are several around here from time to time

[dialsc]

Maybe this is a bit on the exotic side but I'm mainly doing CAD/CAM stuff and therefore it is absolutely essential that the machine is running 100% rock solid/stable. This is why I thought first to test with Prime95 using Custom settings -> Min/Max FTT 1344/1344 and having Run FFTs in-place enabled. As long as this test is not running without any error, I do not consider the machine to be stable. Once I achieved that (hopefully), I'll go ahead to run different stress test, RealBench is one of them.

Anyways, thank you for your help!

[Menthol]

For your intended use I can see you need stability above all else, memtest or GSAT is recommended for testing memory stability and AIDA64 cache stress test
But your question is on voltage set to voltage read, a dvm is the best way to get a true reading,
you need Raja or Praz to tell you if the variance you listed is normal or not

[dialsc]

I will go with GSAT for now and see what's the results. Beside that let's see if either one of these two will give us an idea about an answer to my question which would be cool... 😉