Soldato
- Joined
- 16 May 2007
- Posts
- 3,221
That's how I look at it not wanting to take any risks with melting my CPU, but with 8 cores have to be a little careful as any gain is x8. So 0.2Ghz x 8 = 1.6Ghz potentially.
-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
Latest Ryzen overclocks? Are the clocks improving yet?
- Thread starter jigger
- Start date
That's how I look at it not wanting to take any risks with melting my CPU, but with 8 cores have to be a little careful as any gain is x8. So 0.2Ghz x 8 = 1.6Ghz potentially.
You're making me think of those horrible ebay listings which add up the frequency of all cores then list it as one number.
15.2GHz QUAD CORE!
Soldato
- Joined
- 16 May 2007
- Posts
- 3,221
You're making me think of those horrible ebay listings which add up the frequency of all cores then list it as one number.
15.2GHz QUAD CORE!
In that case I have a 29.4Ghz multicore PC
I might go for a 1900X depending of the board prices and features.
That's what I'd be buying if I was building a new system now. No Threadripper rejects for me!
That's what I'd be buying if I was building a new system now. No Threadripper rejects for me!
I've sort of been assuming that the 8c TR parts are two dies with 4 failed cores on each... but maybe they know whether or not the chips are 100% good before they glue them together and the 1900x is just 8 good cores and no bad ones, in which case it does sound like a good way to get a pre-binned one. If you're prepared to pay the HEDT tax
That's what I'd be buying if I was building a new system now. No Threadripper rejects for me!
I'd like everything on AM4 so any upgrades can get filtered down. Will have to wait and see.
Sorry in cinebenchR15 not passmark
This, generally 1800x is an easy 4ghz for stable use. Past that with some messing it will be fine for some use but less straight forward to get things stable. Mine worked straight away for most benchmarks at 4.1-4.15 but mostly with faster memory just set at 4.0 and works really well.
Price is good now so worth a go until the next gen, in theory you'll be able to swap in zen 2 so get a decent board.
Soldato
- Joined
- 19 Oct 2008
- Posts
- 6,042
But it is, it's equivalent to a single core processor running at 15.2GHz.You're making me think of those horrible ebay listings which add up the frequency of all cores then list it as one number.
15.2GHz QUAD CORE!
Soldato
- Joined
- 27 Dec 2009
- Posts
- 2,727
- Location
- Gillingham, Kent
You're making me think of those horrible ebay listings which add up the frequency of all cores then list it as one number.
15.2GHz QUAD CORE!
Wow!! Genuinely thought the craze for doing that died out years ago, but there's someone on ebay still doing it!
http://www.ebay.co.uk/itm/WINDOWS-1...027252?hash=item2f060dbfb4:g:UjUAAOSwf-VWVZvN
Big numbers + Buzz words + Caps Lock = Awesomeness!
It wont die. Its just bad marketing :> that listing is so cluttered that a normal person can't understand itWow!! Genuinely thought the craze for doing that died out years ago, but there's someone on ebay still doing it!
http://www.ebay.co.uk/itm/WINDOWS-1...027252?hash=item2f060dbfb4:g:UjUAAOSwf-VWVZvN
Big numbers + Buzz words + Caps Lock = Awesomeness!
For that spreadsheet all you need is a CPU-Z validation. The CPU I used there is at edge of stability and with high voltage, most of the results in there are as such.
Had 3x R7 1700, from week 05 to 13 production, all reach 3.8GHz high stability using ~1.35V +/-0.02V. 3.9GHz none reached with high stability, even things like RB stress mode/custom x264 had WHEA errors, voltage was ~1.45V.
The 1800X I have, is using ~1.425V for 3.9GHz, week 10 production. TBH the 1800X IMO is best to run at stock as I didn't breach the XFR 4.0-4.1GHz clocks. At stock depending on temps/loading will do 3.9-4.0GHz on 2 cores. The stock 3.7GHz all cores clock is pretty good performance with high efficiency IMO, as it uses ~1.25V.
RAM on all I have gained 3333MHz tight, 3466MHz is about where you start to see which is better. The 1800X has no issue with 3466MHz C15 1T, but same RAM/Mobo, etc used with R7 1700 will not do same clocks/timings. I wouldn't say the R7 1700 is any lesser in IMC quality as I have noted some OCN members having better RAM setup on an R7 1700 than my samples.
Did you get any memory training issues going for 3466? I have 1000% HCI stability on my 1700 at 3466 but it rarely boots without an F9 fail.
PS yes I am stalking you
Did you get any memory training issues going for 3466? I have 1000% HCI stability on my 1700 at 3466 but it rarely boots without an F9 fail.
PS yes I am stalking you
LOL, yeah noted you crossed interweb borders to OCN
.On C6H UEFI 9920 has extra training at post, this solves any rare intermittent Q-Code: F9 I would have on another UEFI when pushing RAM.
AGESA 1.0.0.7 has been confirmed by Elmor to be released in next round of UEFIs. Besides CPU Microcode updates it can include IMC FW updates, so perhaps AGESA 1.0.0.7 will improve this aspect and mobo will not need extra training.
While 4Ghz to 4.2ghz is just 5%, or even 4Ghz to 5Ghz is just 25%. The difference between 2133 Ram and 3466C14 (tight timings), is 40% on pure gaming
Nah, 4Ghz the 1700 or 4.1 at best with 1800X.
And that might seem low, but 4Ghz to 4.2ghz is just 5%, or 4Ghz to 5Ghz is just 25%. However RAM plays huge factor with Ryzen.
Difference between 2133 Ram and 3466C14 (tight timings), is 40% on pure gaming FPS. The difference between 3200 with loose timings to 3466C14 is more than 10% more FPS.
On contrary 2666 to 4133 on intel platforms, barely gives more than couple of percent.
AMD 19xx series, support 3600Mhz ram.
Apparently according to everyone, Zen+ is designed for 4.5Ghz at same envelop, coming some time early 2018 and is compatible with current boards.
true. You can gain more going from 3200 to 3466C14 ram.
Nah, 4Ghz the 1700 or 4.1 at best with 1800X.
And that might seem low, but 4Ghz to 4.2ghz is just 5%, or 4Ghz to 5Ghz is just 25%. However RAM plays huge factor with Ryzen.
Difference between 2133 Ram and 3466C14 (tight timings), is 40% on pure gaming FPS. The difference between 3200 with loose timings to 3466C14 is more than 10% more FPS.
On contrary 2666 to 4133 on intel platforms, barely gives more than couple of percent.
AMD 19xx series, support 3600Mhz ram.
Apparently according to everyone, Zen+ is designed for 4.5Ghz at same envelop, coming some time early 2018 and is compatible with current boards.
true. You can gain more going from 3200 to 3466C14 ram.
Permabanned
- Joined
- 23 Apr 2014
- Posts
- 23,552
- Location
- Hertfordshire
3925 and 3200 ram is best I have at the moment. Not too shabby.
LOL, yeah noted you crossed interweb borders to OCN
On C6H UEFI 9920 has extra training at post, this solves any rare intermittent Q-Code: F9 I would have on another UEFI when pushing RAM.
AGESA 1.0.0.7 has been confirmed by Elmor to be released in next round of UEFIs. Besides CPU Microcode updates it can include IMC FW updates, so perhaps AGESA 1.0.0.7 will improve this aspect and mobo will not need extra training.
Haha yes. Some dedicated people over there. Been through all the results on the 24/7 DDR4 thread and copied various ideas but can't get a good 3466 result without F9 hell. I think my RAM is a relatively poor bin although I'm quite happy with 3333 having 3466 stable when it boots is too tantalising to give up!
Might a magic CLDO_VDDP value be in there somewhere that solves it? Is there an efficient way of finding out or do people literally try 1mV increments?
Here's hoping 1.0.0.7 has some smart training fixes!
As I was convinced my Q-Code: F9 was down to UEFI and not setup I went through trying 1mV increments. It does seem as if it works this way. For example 903mV was pretty 'optimal' on my R7 1700, in that it lasted the longest without a Q-Code: F9. Where as 902mV or 904mV wasn't as lasting. My R7 1700 has a memory hole for 3333MHz RAM. Using the same set up on UEFI with extra training resulted in 0 Q-Code: F9 on over 11 days use, then I went another CPU. If I didn't do CLDO_VDDP tweak on that UEFI 3333MHz was still a hole, so the extra training was not at all masking the issue.
What I noted was, lowering CLDO_VDDP moved the hole down the frequency range, increasing it moved it up. It may work differently for other CPUs, but another on OCN also experienced/posted the same, Gadfly.
I could go to any RAM divider upto 3200MHz on stock CLDO_VDDP, 3466MHz would post as well, but 3333MHz wouldn't. I change CLDO_VDDP to 903mV and 3200MHz gets knocked out, but any other divider works fine from 2133MHz to 3466MHz.
Too low a CLDO_VDDP can also lead to instability. For example I set something crazy low like 650mV, I decided to use 3200MHz, as soon as I ran HCI Memtest system just froze.
If the memory MHz is sitting at the 'cusp' of memory hole then also you can have instability in OS. This is harder to diagnose IMO, you only sorta know if you have have done a lot of tweaking/testing and know your HW well on what it can and can't do.
Recently when testing 3466MHz on my 1800X I found ProcODT of 60ohms lead to issues in stability testing where CPU was loaded, if I used 68 ohms it failed quicker, 53.3 ohms was perfect. I find passing RAM tests easier than when tests where CPU is loaded more and RAM is under load. A member on OCN suggested to me to try his method of using P95 with 1440K/1440K/12000MB as he found that productive and it worked well for me. That was how I tuned ProcODT for that profile without increasing VCORE/SOC/VDIMM, caveat still apply that they must be appropriate.
Anything above 3200MHz IMO really becomes a 'tuning' exercise of settings.
What I noted was, lowering CLDO_VDDP moved the hole down the frequency range, increasing it moved it up. It may work differently for other CPUs, but another on OCN also experienced/posted the same, Gadfly.
I could go to any RAM divider upto 3200MHz on stock CLDO_VDDP, 3466MHz would post as well, but 3333MHz wouldn't. I change CLDO_VDDP to 903mV and 3200MHz gets knocked out, but any other divider works fine from 2133MHz to 3466MHz.
Too low a CLDO_VDDP can also lead to instability. For example I set something crazy low like 650mV, I decided to use 3200MHz, as soon as I ran HCI Memtest system just froze.
If the memory MHz is sitting at the 'cusp' of memory hole then also you can have instability in OS. This is harder to diagnose IMO, you only sorta know if you have have done a lot of tweaking/testing and know your HW well on what it can and can't do.
Recently when testing 3466MHz on my 1800X I found ProcODT of 60ohms lead to issues in stability testing where CPU was loaded, if I used 68 ohms it failed quicker, 53.3 ohms was perfect. I find passing RAM tests easier than when tests where CPU is loaded more and RAM is under load. A member on OCN suggested to me to try his method of using P95 with 1440K/1440K/12000MB as he found that productive and it worked well for me. That was how I tuned ProcODT for that profile without increasing VCORE/SOC/VDIMM, caveat still apply that they must be appropriate.
Anything above 3200MHz IMO really becomes a 'tuning' exercise of settings.