Jump to content
HWBOT Community Forums

sin0822

Members
  • Posts

    757
  • Joined

  • Last visited

Posts posted by sin0822

  1. Thanks @sin0822. I saw the guide pop by on Facebook - looks very nice!

     

    Request: can you have a look at the relationship between the AVX instruction set and power management for Skylake? I recall AVX performance increasing the Vcore on Haswell-E CPUs and based on the Skylake non-K unlocking it seems there is a direct relationship.

     

    So I plan to add a section on non-K OC using Supermicro's C7Z170-OCE since it's an OC board and Supermicro did have that BIOS first. I will look at it there.

     

    I know that AVX does seem to have a pretty big impact. A quick glance at the LLC results (each boards' page has LLC tests) I test maximum load with CINEBENCH R11.5 (I don't think it has AVX) and then I also do Intel Burn Test with AVX. Of course the load of IBT itself could just raise or lower voltages more, but obviously there is a large increase in power considering the increases in voltages when utilizing positive LLC and decreases with negative LLC levels.

     

    I will look into a more consistent way to test this out and the impacts on power management.

     

    There might also be an impact depending on the voltage mode used, but i would think that with non-K OC you would use manual mode instead of adaptive or offset since the frequency doesn't drop. Using LLC on manual is there, but can be negated depending on the LLC of the board.

     

    You want to know whats really crazy, all the five boards I used in the article (and including the Supermicro #6) all use the same PWM! lol, you would think their LLC would be quite similar, but its greatly differs between brands (two brands are quite close tho).

     

    BTW thanks!

  2. There's really no need to defend your product or its pricing, Roman. I don't know why you keep bothering haha.

     

    In the end, yours is the only on the market that allows for 100% infinite delids. As said before, the price is less than a third of the cost of a new CPU after a failed delid. In addition, the material used and build quality is solid.

     

    If people want to print on for $10, fine. If they want to buy a printed one for $25, that's fine. If they want yours for 90, that's fine too.

     

    Thanks for giving people who want to buy advanced tools the opportunity. I think we all too quickly forget about the difficulties of getting our hands on low quantity OC tools ...

     

    Sent from my A0001 using Tapatalk

    Exactly.

  3. Hey Alex,

     

    I think you did a great job. Also I don't agree with hiding information to make it challenging, there is literally no guide on tweaking tertiaries for DDR4, plus isn't not like he wrote a guide on "How to Splave" and beat Splave at his own game in the HOT comp. He just wanted to share his knowledge of one of the more intimidating areas of DRAM overclocking, where things differ a lot, and he gave advice on how to overcome odds that give people headaches and make them burn motherboards and set them afloat on lakes. There is a reason that ROG guys release those LN2 benching guides at launch, there is no reason for them to hide the information, as if they make it more public on their platform people will buy it. It's the same thing as DRAm timing profiles they had, I bet a lot of money that boosted their ROG sales a tiny bit as well.

     

    Anyways nice job Alex, people always going to find problems, and there is always room for improvement.

  4. Whatever, what a pain. Why bother loading the CPU with this xtreme OC thing? Isn't the point not to have a load?

     

    Also how to get this feature where it saves validation file every 3sec>? I tried with set to 0 and 1 and neither saves every 3 sec

  5. is it just me or does 05 gives a hard time matching a certain score?...it's soooooo random xD

     

    I had +2 and -2 already but next run --> +/- 100 points....well there is still some time left so re-run, re-run, re-run...

     

    @Massman don't blame the server, according to what I know from servers, dbs and stuff this lag is caused by all the other multiple submissions not just the one sending it (server actually doesn't care where the data comes from, it just reserves IDs and starts to handle them first-in, first-out!)..

    my first result in stage 1 took 38 seconds to be accepted because there have been like 20-40 results in queue from others...I only sended one single result and it lagged like never before...If everybody only submits one result -> less/no lag -> no problem :)

     

    Yea don't blame server, if the server can't handle it(then fix the servers, I think you have some sort of obligation, they are still broke from the last issues I told you about Pieter) let the forums handle it, they have time stamped posts. That is how we handled these types of comps back in the day, forum time stamps.

×
×
  • Create New...