![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
I would be more interested in the top two core max OCs using PB2 for increasing IPC (instructions per core) and not all the cores. For single threaded calculations that is what matters. When you are doing all cores then you are limited to the slowest core overall. So for the slowest core, average is around 4.3ghz on water - what about the top two cores? 4.5ghz? 4.6ghz+?
With BCLK OCing it does. AMD got rid of the OC mode restriction on BCLK changes or allowed some wiggle room there. Not sure if one can take it up to 130hz and not be in OC mode.
Yikes... That's a whole lotta volts... Sounds like either Duron knows more about what he's doing that they do, or he ended up with a golden chip when comparing to their results >_> Especially considering they aren't using any air coolers.
DuronBurgerMan, Original Post: "Okay so for the lulz I bought a 2700X and dropped it into my Asus Prime X370 Pro board today. <snip>. Note, I am running a Noctua U12 air cooler, so I imagine folks on water will do better."
DuronBurgerMan, Second Post: "Second attempt: 42x multiplier, Vcore @ 1.365. 4.2GHz total.
Booted to Windows. Appears fully stable at this vcore. Ran Cinebench successfully (score: 1883). Will conduct burn test later to verify stability."
Seems like their SOP is: Buy Ryzen. Insert in motherboard. Boot with voltage at 1.425V. Find max stable overclock at said voltage. Re-test in next motherboard until no boards left. Done. CONSIDER IT BINNED!
No bother of finding if it'd handle a lower voltage?
I just see it as if they are going to provide this service to begin with, then they ought to be putting in that effort. Pretty much any yahoo can chuck a Ryzen into a board and get what they are lol What they've done hasn't really binned anything. Showing that it can do a speed at a reasonable voltage, well now you're starting to determine something!
I don't know if it's my age that has imparted this upon me, or if it's wisdom, or... what... lol However, personally, CPU overclocking has not been as fun as it had been back with the S939 A64 and X2's. So I kinda welcome that PB2 can do it all and save me from having to bother. I find, these days, that DRAM is more enjoyable to tinker with, and at least with Ryzen it's a rewarding endeavor with it having a real impact on the performance.
Sad day...]
I'm not certain. That'd be the first time I've heard of that beyond something similar sounding that ASRock did on... Taichi and/or Fatal1ty Ryzen boards. I just recall some feature being talked up that was intended to make like better in that regard. Whether or not all that is, is a stand-alone BClk gen (as giga mentioned) or something actually more in ASRock's case, I have no clue... Alas, my Titanium lacks an independent clock gen completely, so I can't raise or lower. (Now, I did find in the BIOS a functional option that does allow BClk on the Titanium, but it is limited to 103MHz, and I speculate that all this does is tap into Ryzen's internal clock-gen and is really nothing more than what mobo makers use in order to fine-tune their board to running as close to 100MHz for BClk. 103MHz wasn't stable in my initial testing, and 102MHz was mostly-stable. The increase in performance was too negligible to bother, in the end.)
I just see it as if they are going to provide this service to begin with, then they ought to be putting in that effort. Pretty much any yahoo can chuck a Ryzen into a board and get what they are lol What they've done hasn't really binned anything. Showing that it can do a speed at a reasonable voltage, well now you're starting to determine something!
I dunno, I just feel like it's a bit of a ripoff to buy a Ryzen from them. And who knows, this could just be due to them having experience with Intel, meaning they'll hone their process in time *shrug*
I'm not sure I fully understand what you are asking?
I dunno man... Superset/Subset/Successor, it all, to me, means "heavily based on" for whatever it is the superset/subset/successor to. Considering the fact that the Wiki page for Infinity Fabric redirects to the HyperTransport page says a lot in my opinion. HyperTransport was, among other things, a high speed CPU-to-CPU interconnect. The Athlon X2 can be looked at as two CPU's (cores) on a chip, and thus the HT ferries data between them and the IMC (also on chip). Ryzen can be seen as two 4c CPUs on a chip, and the Infinity Fabric ferries data between them and the memory.not really
I'm not sure I fully understand what you are asking?
For starters, when you say "Ryzen2" are you talking about the 2000-series chips such as the 2700X? Or are you talking about what is, on the AMD roadmaps, called Ryzen 2?
This is an important distinction since the 2000-series chips are based on "Zen+" (thus, Ryzen+), and on an architectural level really is just some minor improvements overall (to improve latency, updates the core boost functionality, allows for higher memory clocks, etc).
The second clarification I'd need is: when you say "process", I presume you mean the manufacturing process (node) at which the chips are made on?
If so, then it's been a year since Ryzen came out, not 6 months, and it was made on the Global Foundries 14nm node (LPP, or Low-Power Plus, licensed from Samsung, and is the 'higher performance' sibling to LPE). Ryzen+ is made on a 12nm node, so lower power, less heat, potentially higher clocks --at least, provided that the ~4.1GHz wall we saw with Ryzen wasn't an architecture limit-- and so that naturally needs to be all tested again.BUT in addition to that small process shrink, Ryzen+ is now manufactured by..... ok nevermind! I just double checked and apparently wherever I originally read what I was about to say, was incorrect. So in case anyone else thought this: Ryzen+ is not made on TSMC's 12nm, it's still on GloFo's, their 12nm LP (Leading Process) which is again a process licensed from Samsung. And according to some digging, 14nm and 12nm are qualifiably "Half Nodes" since they aren't one of the official node sizes (16nm and 10nm being the closest).
And while there's arguably not a huge difference between the 14nm process and 12nm process, it still does merit overclocking tests (to spare further rambling, the nodes aren't technically processed entirely at this size, there are aspects that are, but other aspects that are sized more in line with those of the 16nm node). I mean there's always been room to increase clocks, which is sort of the entire point of overclocking. There are limitations of a design or node, but then there are what the CPUs get binned as being capable of... which is just an average of what their samples have been capable of achieving and thus can be sold as "stable at this frequency". Beyond that is where Boost comes in, as it's not a guaranteed speed, at least for any specific period of time.
Last clarification part for me: in regards to "AMD gets some experience making Ryzen2", are you asking about CPU Steppings? I don't believe we saw any with the original Ryzen. The closest thing we saw was the difference in manufacturing location between some of the 1700 (diffused in China I believe) and 1700X/1800X (diffused in Malaysia I believe), where the 1700 was being found to not quite able to achieve the same memory speeds as the X variants. This could all come down to the plant having sourced their chemicals from a different supplier. But that goes to show as well that if 12nm required different chemicals than 14nm, it is another reason for taking a look at overclocking. I seem to recall that being a case with the AthlonXP... but that's a long time ago, so details are sketchy haha Either way, new steppings still generally require retesting. The Athlon64 X2's saw improvements in this way, granted it was marginal, but it all adds up
All in all I have admittedly no doubt read too deep into your question, so hopefully something from my rambling answers your question
I dunno man... Superset/Subset/Successor, it all, to me, means "heavily based on" for whatever it is the superset/subset/successor to. Considering the fact that the Wiki page for Infinity Fabric redirects to the HyperTransport page says a lot in my opinion. HyperTransport was, among other things, a high speed CPU-to-CPU interconnect. The Athlon X2 can be looked at as two CPU's (cores) on a chip, and thus the HT ferries data between them and the IMC (also on chip). Ryzen can be seen as two 4c CPUs on a chip, and the Infinity Fabric ferries data between them and the memory.
Granted, the HT in its original function and how it executed things was eventually replaced by the PCIe bus, but it was still called HT the entire time as far as I am aware. Which as far as I know, the Infinity Fabric has been said to be heavily interfaced with the PCIe bus as well.
It definitely seems like a matter of semantics, but for all intents and purposes, they seem more similar than dissimilar and so saying that it is "pretty much" --which implies for the most part, but not entirely-- a renamed HT isn't accurate to you? It's like the saying "If it walks like a duck and talks like a duck, it must be a duck" in my eyes lol
Not trying to say I'm right, just... more trying to explain how I reached this. If consensus is that no, it isn't, I'll happily edit my post!
I'm not sure I fully understand what you are asking?
For starters, when you say "Ryzen2" are you talking about the 2000-series chips such as the 2700X? Or are you talking about what is, on the AMD roadmaps, called Ryzen 2?
This is an important distinction since the 2000-series chips are based on "Zen+" (thus, Ryzen+), and on an architectural level really is just some minor improvements overall (to improve latency, updates the core boost functionality, allows for higher memory clocks, etc).
The second clarification I'd need is: when you say "process", I presume you mean the manufacturing process (node) at which the chips are made on?
If so, then it's been a year since Ryzen came out, not 6 months, and it was made on the Global Foundries 14nm node (LPP, or Low-Power Plus, licensed from Samsung, and is the 'higher performance' sibling to LPE). Ryzen+ is made on a 12nm node, so lower power, less heat, potentially higher clocks --at least, provided that the ~4.1GHz wall we saw with Ryzen wasn't an architecture limit-- and so that naturally needs to be all tested again.BUT in addition to that small process shrink, Ryzen+ is now manufactured by..... ok nevermind! I just double checked and apparently wherever I originally read what I was about to say, was incorrect. So in case anyone else thought this: Ryzen+ is not made on TSMC's 12nm, it's still on GloFo's, their 12nm LP (Leading Process) which is again a process licensed from Samsung. And according to some digging, 14nm and 12nm are qualifiably "Half Nodes" since they aren't one of the official node sizes (16nm and 10nm being the closest).
And while there's arguably not a huge difference between the 14nm process and 12nm process, it still does merit overclocking tests (to spare further rambling, the nodes aren't technically processed entirely at this size, there are aspects that are, but other aspects that are sized more in line with those of the 16nm node). I mean there's always been room to increase clocks, which is sort of the entire point of overclocking. There are limitations of a design or node, but then there are what the CPUs get binned as being capable of... which is just an average of what their samples have been capable of achieving and thus can be sold as "stable at this frequency". Beyond that is where Boost comes in, as it's not a guaranteed speed, at least for any specific period of time.
Last clarification part for me: in regards to "AMD gets some experience making Ryzen2", are you asking about CPU Steppings? I don't believe we saw any with the original Ryzen. The closest thing we saw was the difference in manufacturing location between some of the 1700 (diffused in China I believe) and 1700X/1800X (diffused in Malaysia I believe), where the 1700 was being found to not quite able to achieve the same memory speeds as the X variants. This could all come down to the plant having sourced their chemicals from a different supplier. But that goes to show as well that if 12nm required different chemicals than 14nm, it is another reason for taking a look at overclocking. I seem to recall that being a case with the AthlonXP... but that's a long time ago, so details are sketchy haha Either way, new steppings still generally require retesting. The Athlon64 X2's saw improvements in this way, granted it was marginal, but it all adds up
All in all I have admittedly no doubt read too deep into your question, so hopefully something from my rambling answers your question
I dunno man... Superset/Subset/Successor, it all, to me, means "heavily based on" for whatever it is the superset/subset/successor to. Considering the fact that the Wiki page for Infinity Fabric redirects to the HyperTransport page says a lot in my opinion. HyperTransport was, among other things, a high speed CPU-to-CPU interconnect. The Athlon X2 can be looked at as two CPU's (cores) on a chip, and thus the HT ferries data between them and the IMC (also on chip). Ryzen can be seen as two 4c CPUs on a chip, and the Infinity Fabric ferries data between them and the memory.
Granted, the HT in its original function and how it executed things was eventually replaced by the PCIe bus, but it was still called HT the entire time as far as I am aware. Which as far as I know, the Infinity Fabric has been said to be heavily interfaced with the PCIe bus as well.
It definitely seems like a matter of semantics, but for all intents and purposes, they seem more similar than dissimilar and so saying that it is "pretty much" --which implies for the most part, but not entirely-- a renamed HT isn't accurate to you? It's like the saying "If it walks like a duck and talks like a duck, it must be a duck" in my eyes lol
Not trying to say I'm right, just... more trying to explain how I reached this. If consensus is that no, it isn't, I'll happily edit my post!
Aaaalright lol Fair enoughSuperset means it contains the thing it is a superset of meaning, HT is only a part of, one of the parts of, what makes up IF.
IT's like saying PCI-E is just PCI renamed.
Aaaalright lol Fair enough
Instead of "is just a renamed HT", would "is an evolution of HT" be acceptable?
I wish I knew more about squeezing out performance on RAM to be honest
Some important factors though would be to let us know what model kit you have and then if they're Single or Dual Rank sticks (CPU-Z tells you in the memory tab). If you want to try mine, have at it: 312 (or 195ns) Daunting... but rewarding!
I'm not sure how I missed this originally, but I only noticed it last night while making the above reply, sorry about that!
Given we seem to have the exact same kit, here are my 200% MemTest stable timings for comparison (1.36V just to keep them >1.35V), as I have a few subtimings that are tighter if you want to give them a go (and I could probably drop my tRTP and tRDWR a tick or two each as well):
Well you have a Ryzen+, I don't, so I can't really partake in the benefits of higher RAM speeds
Which I'm pretty sure is a limitation of either my old BIOS (my setup has ran so beautifully I've not been bothered to update since last June!) with older AGESA, or my IMC. I bought this kit specifically because reviews on Intel platforms had been able to overclock them QUITE easily, and to very high speeds. Nevertheless, my adventures stop at 3333. I've gotten 3466 to POST only twice. First time was the absolute first time I tried the speed, and the second time was after a metric crapload of futzing about. That's why I stuck to dropping my timings instead heh I poked and prodded Chew* for a few months, reading his thread over at XtremeSystems, but after July of last year (annual month long vacation) I never went back to catch up, and there'd be pages and pages to catch up on by now Given the new AGESA and BIOS for my Titanium, with a further wealth of features added, I may give it a shot again. Still sorta reluctant on account of being in a stable spot right now lol Chew never tried my modded ASUS one, so I don't know if it'd work. They are rather locked-down to prevent flashing of modded BIOSes, and the only plausible trick relies on the person flashing the BIOS they want modded, dumping it using AMI's flashing utility, and modding that (which if I do it, requires uploading it for me to download). Apparently that keeps whatever unique code for that system only in the BIOS and will let you re-flash it. As for ASRock or Gigabyte boards, I have no clue. My last Gigabyte board was an FM1 system and they signed their BIOSes, so I couldn't flash my modded ones to it I know that's pretty common on their modern Intel variants, so I'd assume just the same on AMD's.I'm not sure how I missed this originally, but I only noticed it last night while making the above reply, sorry about that!
I presume you tried disabling the GearDownMode as well? Though it no doubt is needed for that much RAM.
Also, how high did you try on the RAM Voltage? And what's your SoC (CPU_SOC or CPU_NB) voltages that you had tried?
Beyond that, a few of those sub-timing suggestions I made to Rauelius would apply to you, though as I"m sure most of what you have in the screenshot are already dialed in by you, I'd personally probably only be attempting to lower: tRC, tFAW, tWR and tRTP. Mine are: 54, 25, 11, and 9, respectively.
Sad to hear, ain't it?
Hopefully those subtimings help a tiny bit then
Hopefully those subtimings help a tiny bit then
I was also very irked initially when Ryzen released with the exact same "No Odd CLs", which was ultimately addressed. So I'm perplexed that it's back again!
Though... You know what... A couple synapses just randomly fired in my brain and I suddenly seem to recall that GearDownMode is what allows for using Odd CLs! With it Enabled it prevents that, but Disabled allows it...
I'll research and edit or post back what I can dig up!
