Overclock.net - An Overclocking Community - Reply to Topic
Thread: (Gigabyte Z390 AORUS Owners Thread) Reply to Thread
Title:
Message:

Register Now

In order to be able to post messages on the Overclock.net - An Overclocking Community forums, you must first register.
Please enter your desired user name, your email address and other required details in the form below.
User Name:
If you do not want to register, fill this field only and the name will be used as user name for your post.
Password
Please enter a password for your user account. Note that passwords are case-sensitive.
Password:
Confirm Password:
Email Address
Please enter a valid email address for yourself.
Email Address:

Log-in


  Additional Options
Miscellaneous Options

  Topic Review (Newest First)
02-16-2020 11:08 AM
reachthesky Just going to post this screenshot of you quoting and arguing with yourself.

02-16-2020 10:35 AM
sygnus21
Quote: Originally Posted by reachthesky View Post
Gigabyte doesn't say anything about 4200+ being extreme overclocking. They just say their boards will do it. Has anyone gotten 4400+ stable on ram? It is perfectly reasonable for a customer to expect the product to work as advertised. Gigabyte says 4400+ is doable on their marketing literature.
Instead of putting words in my mouth to fit your narrative you should actually read what I wrote. And the post you quoted is a clarification on an earlier post...

Quote: Originally Posted by sygnus21 View Post
I'd like to chime in by saying one should consider these things "before" buying their components. If you're going to buy a cheap board and components then you should not expect top end overclocking performance. There are extreme overclocking boards out to include Gigabyte's own Z390 Aorus Xtreme board (of which I have... and is overkill for my needs). The other part of that equation paying for the RAM needed needed to achieve those extreme overclocks... as well as having the adequate cooling for the system.

Sure just about any board and RAM kit can overclock, but how high and stable depends on the quality and design of the components... and those come with higher costs. And yes, not all boards specifically built for extreme overclocking are great "gaming" or multimedia boards. So in the end one needs to decide where they want to go.

My two cents.
Apparently you're here to argue and not read.
02-16-2020 06:46 AM
reachthesky Ya this kit is pretttty good. Best 4 dimm kit i've worked with so far. I'd like to get my hands on and test one of those GSkill [email protected] @ 1.5v vdimm xmp kits that should be coming out soon. I think they are geared for amd/x299 or something like that i think but i'm sure they'd be fine on z390 as i've used a quad channel x299 kit on here before. IC quality must be insanely high at those timings to pass qvl process on x299 because they have to support performing double of what we pull on z390 due to quad channel single dimm per channel config, much higher ic quality than something like 2x8gb cas 17-4000 and maybe a couple steps above my kit. probably won't even need anywhere close to 1.5v on a 2dimm dual channel config like the dark or apex for c15-4000 and I think could more than likely squeeze out somewhere between c15-4133 and c15-4200 @ 1.5v. I'd LOVE to work with those for sure.

EDIT: O lord I just saw for the first time what gskill revealed at one of the tradeshows earlier this year.....a cas17-5000 bdie kit on a 9900K. Nuts. They also showed a c18-5200 kit on a 9700k.
02-16-2020 06:37 AM
iunlock
Quote: Originally Posted by reachthesky View Post
Yeah i'm not so much scared about putting up to 1.6v vdimm into the sticks, its the cpu IMC and board traces i'm concerned about. I actually kind of need somewhere between c15-4200 and c16-4400-4500 performance. Can the dark do c15-4133/4200+ or c16-4400-4500 at 1.5v vdimm or less for daily use? I heard the dark requires less vdimm than usual compared to most other boards.
The Dark can do 4000/c15 for sure as that is what I have it set at, and I'm pretty confident it can do 4133/c15. The limiting factor are the RAM sticks when it comes to the Dark, granted it's within specs of the board. Oh and the 4000/c15 is at 1.50v, which I keep as my max.

Quote: Originally Posted by reachthesky View Post
@Falkentyne I think i know why a lot of people get instability with avx offsets(on auto/adaptive/dvid voltage at least). It's because of the cache. When the cpu turbo ratio multiplier is toned down(when the avx offset kicks in), it goes off the vid table for that multiplier when it comes to requested/received voltage). That multiplier's voltage is factored with a cache ratio 300mhz below it. So lets say you used a typical 5ghz all core, 4.7ghz cache overclock with avx offset of 2. When avx kicks in, cores goto 4.8ghz as expected. The cache is still at 4.7ghz though but the cpu is requesting/receiving only enough voltage to power 4.8ghz with 4.5ghz cache. That's why I think there is instability, not enough voltage for the 4.7 cache.

The solution to this is to turn on ring to core so that the cache fluctuates at 300mhz lower within whatever all core frequency cpu is currently at so that you are staying within the vid table when avx kicks in. This means that in the bios a user should make sure to also set the cache manually in the bios to 300mhz behind the all core frequency, ring to core will handle upclocking downclocking in windows when needed.

With ring to core enabled with the same config, behavior would look like this:

5ghz all core 4.7ghz cache during non-avx loads
4.8ghz all core 4.5ghz cache during avx loads

Figured I'd share this incase anyone may have been banging their head against the wall as to why they were getting instability or if they tried to mitigate the instability by starting with an even lower cache ratio at the get go specifically with an avx offset oc.
100%. This is very true. As a rule of thumb when tuning my CPU I try to always keep the cache within 3x from the core frequency. It's an area that I don't revisit much since I prefer static, but great great point that you bring up. *thumbs up*
02-16-2020 06:14 AM
reachthesky @Falkentyne I think i know why a lot of people get instability with avx offsets(on auto/adaptive/dvid voltage at least). It's because of the cache. When the cpu turbo ratio multiplier is toned down(when the avx offset kicks in), it goes off the vid table for that multiplier when it comes to requested/received voltage). That multiplier's voltage is factored with a cache ratio 300mhz below it. So lets say you used a typical 5ghz all core, 4.7ghz cache overclock with avx offset of 2. When avx kicks in, cores goto 4.8ghz as expected. The cache is still at 4.7ghz though but the cpu is requesting/receiving only enough voltage to power 4.8ghz with 4.5ghz cache. That's why I think there is instability, not enough voltage for the 4.7 cache.

The solution to this is to turn on ring to core so that the cache fluctuates at 300mhz lower within whatever all core frequency cpu is currently at so that you are staying within the vid table when avx kicks in. This means that in the bios a user should make sure to also set the cache manually in the bios to 300mhz behind the all core frequency, ring to core will handle upclocking downclocking in windows when needed.

With ring to core enabled with the same config, behavior would look like this:

5ghz all core 4.7ghz cache during non-avx loads
4.8ghz all core 4.5ghz cache during avx loads

Figured I'd share this incase anyone may have been banging their head against the wall as to why they were getting instability or if they tried to mitigate the instability by starting with an even lower cache ratio at the get go specifically with an avx offset oc.
02-16-2020 06:07 AM
lucasfrance My memory OC that was rock stable for months suddenly whas no that much stable under heavy stress tests for a few weeks...

After trying to recover this by increasing voltages I went back to the stable configuration and just made a cleaning of the memory sticks contacts with Ether and all is back to normal !!!

Even if the weather is not that much wet here in Switzerland (even not cold at all this year but this is an other story...) I suspect a very thin layer of oxidation appeared on the memory stick contacts that made my previously stable config a bit unstable.

Should you have this kind of experience (or also possibly to improve you memory OC!) try this very easy trick.

Note the very same applies to GPU/PCI (on the GPU side) and CPU/MB contacts (on the CPU side of course!).

Give a try and report back.
02-16-2020 05:38 AM
reachthesky
Quote: Originally Posted by iunlock View Post
Ah I see... nice. Ya know?... 1.60v isn't as scary as it seems as I know several fellow overclockers (that compete) to use that as a base to tune. Would I run it there daily? Wellllll honestly no haha, but even if I had to I wouldn't be too nervous. It's not so much the fact that I think the kits can't handle the voltage, rather it's just way overboard for what I realistically need for my usage habits on the gaming rig. I'm going to try and tighten things down at 4266 and see where it hits a wall for real world daily usage stability. It shall be interesting...

Edit: I just saw your edit haha... regarding the timings at 4000, not yet but I sure will later today! Looking forward to it. - You have really good RAM... total winners.

Yeah i'm not so much scared about putting up to 1.6v vdimm into the sticks, its the cpu IMC and board traces i'm concerned about. I actually kind of need somewhere between c15-4200 and c16-4400-4500 performance. Can the dark do c15-4133/4200+ or c16-4400-4500 at 1.5v vdimm or less for daily use? I heard the dark requires less vdimm than usual compared to most other boards.
02-16-2020 05:31 AM
iunlock
Quote: Originally Posted by reachthesky View Post
For cl15/4200, vdimm needs to be 1.57v or 1.58v in bios tehe. The c20/4600 actually benched but I wasn't able to tighten the timings as much as cl19/4533 so cl19/4533 outperformed c20/4600 across the board with something like 9.7k multicore memory performance with hyperthreading disabled in geekbench 3 all without the need to be on LN2.
Ah I see... nice. Ya know?... 1.60v isn't as scary as it seems as I know several fellow overclockers (that compete) to use that as a base to tune. Would I run it there daily? Wellllll honestly no haha, but even if I had to I wouldn't be too nervous. It's not so much the fact that I think the kits can't handle the voltage, rather it's just way overboard for what I realistically need for my usage habits on the gaming rig. I'm going to try and tighten things down at 4266 and see where it hits a wall for real world daily usage stability. It shall be interesting...

Edit: I just saw your edit haha... regarding the timings at 4000, not yet but I sure will later today! Looking forward to it. - You have really good RAM... total winners.
02-16-2020 05:24 AM
reachthesky
Quote: Originally Posted by iunlock View Post
Great kit. Good ol' Samsung B-Die ...



Great break down. Thanks for sharing. Rep+

With your 4200/c15 and above... I have a funny feeling that you're scratching 1.60v'ish on the voltages eh? That's very impressive that you were able to post at the 4600/c20...great milestone.

Gaming and RAM wise, very true... it does make a difference. I'll spend some time with the RAM on the gaming rig later today and see what surprises it may bring. So far the 4266 settings are doing pretty well for gaming with no mem or cache errors... I'm usually pretty optimistic, but I don't think it'd pass any major memtests with a clean sheet, but dang ... I'm almost expecting a BSOD anytime now during games, but it hasn't happened yet. Who knows... only time will tell.
Thank you . Yeah i'm pretty happy with this gskill kit. A++++ to be honest. It's a breeze to work with. Gskill left just enough extra meat on the bone to let the customer feel like they got something extra for free provided they spent the time to obtain it). Out of the 4-5 b-die kits i've worked with, this one takes the cake hands down.

For cl15/4200, vdimm needs to be 1.57v or 1.58v in bios tehe. The c20/4600 actually benched but I wasn't able to tighten the timings as much as cl19/4533 so cl19/4533 outperformed c20/4600 across the board with something like 9.7k multicore memory performance with hyperthreading disabled at 5.3ghz all core 4.8ghz cache in geekbench 3 all without the need to be on LN2. I guess that's general 4 interleaved dimm performance though and nothing specific to the master.

Have you tried [email protected] on your sticks yet on the master?
02-16-2020 05:20 AM
iunlock
Quote: Originally Posted by Grizzly111 View Post
Cheers! They are nothing special just these ones: https://www.gskill.com/specification...-Specification
Great kit. Good ol' Samsung B-Die ...

Quote: Originally Posted by reachthesky View Post
THese are the sa/io requirements for different scenarios for my chip:

43-47 cache generally needs 1.25v sa/io with xmp
48/49 cache needs 1.3v sa/io no matter what
50 cache needs 1.35v sa/io with tuned ram no matter what

ram timings, tight as possible, 12 hour HCImemtest/karhu stable.

c16/4000, c16/ 4100 need 1.25v sa/io
c15/3900, c15/4000 need 1.3v sa/io
c15/4100, c15/4133 need 1.35v sa/io
c17/4000, c17/4100, c17/4133 need 1.25v sa/io
c16/4133 needs 1.3v sa/io
c16/4175 needs 1.3v sa/io
c18/4200 needs 1.3v sa/io

Non-memtest stable:
c15/4200- 1.35v sa/io (benchmark/gaming stable)

Benchmark stable(but won't play games):
c17-18/4500 1.35v sa/io
c19/4533- 1.38v sa/io
c20/4600- 1.4v sa/io

Misc:
cX/4266/4300/4400 - once in a blue moon training on 4 dimms. Stability never found, sometimes benchable, sometimes not. Never usable for gaming
c16/4500+ - can get into windows but no benchmarks
cX/4666/4700- 1.4v sa/io (trains but won't go into windows no matter what)
4800 or higher-No train, motherboard reset.

Ram tuning is indeed rewarding. There is about a 10-12% increase in average FPS and a 30%-45% improvement in 1% lows/minimum fps just from moving from a basic c17/4000 xmp profile to a fully tuned low latency profile. People say it doesn't make a difference in games but I have to disagree. Ram tuning can allow a user to maintain around a 240 average fps in shooters at full ultra settings with a 2080 ti, otherwise you hover around the 200 range at xmp give or take. It also comes in handy when playing starcraft 2 team games like 2v2-4v4 since it is extremely cpu heavy when hundreds of units are on the battlefield at once. The low latency is key there. It's also the one thing that makes intel better than amd for gaming, 20-25% fps difference when comparing fully tuned systems.
Great break down. Thanks for sharing. Rep+

With your 4200/c15 and above... I have a funny feeling that you're scratching 1.60v'ish on the voltages eh? That's very impressive that you were able to post at the 4600/c20...great milestone.

Gaming and RAM wise, very true... it does make a difference. I'll spend some time with the RAM on the gaming rig later today and see what surprises it may bring. So far the 4266 settings are doing pretty well for gaming with no mem or cache errors... I'm usually pretty optimistic, but I don't think it'd pass any major memtests with a clean sheet, but dang ... I'm almost expecting a BSOD anytime now during games, but it hasn't happened yet. Who knows... only time will tell.
This thread has more than 10 replies. Click here to review the whole thread.

Posting Rules  
You may post new threads
You may post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off