Dayaks
[H]F Junkie
- Joined
- Feb 22, 2012
- Messages
- 9,774
What did you have the 5960x OC to?
24/7 was 4.4.
Follow along with the video below to see how to install our site as a web app on your home screen.
Note: This feature may not be available in some browsers.
What did you have the 5960x OC to?
Is that really true if you're not using a 144 or 120Hz monitor?
Read what I wrote above your reply (if you haven't had the chance to): games aren't standing still and neither are GPUs, and games tend to err toward single-core performance even as they continue to make more use of extra threaded resources. The 8700k already presents 'extra' threads; this is why the 2600(x) is such a great budget deal, and why stepping up to a 2700x for gaming solely doesn't make as much sense, as it has slower single-core/single thread performance and more 'extra' threads. It's not bad! It's just less balanced toward gaming.
Oh it's still there if you wanna look for it. Just in different places. With Ryzen in particular it's ALL about dat memory. Really gotta go balls deep into them subtimings if you wanna make her purr. Expect hours and hours of bluescreens playing with like 5-6 different voltages, and plenty of good old-fashioned anger & frustration and then finally.... Glorious, oh the Glorious significant gains to be had!!! Totally worth it. (Sold my 1700 to a friend and got a 2700X).Yeah it's kind of an interesting question too, when you consider that Kyle is not taking it lightly to say what he did in his conclusion:
I've been always insistent on the DIY experience, and whatever I upgrade to next, it's likely I still do it manually, as I have been since 1999. But if Precision Boost 2 is as good as advertised - and Ryzen 2000-series is a testament to that, it certainly helps the more layman get the same performance that many of us at [H] have been getting for nearly two decades through manual overclocking. I wonder if it's too early to start getting nostalgic about it, to see effective boosting mechanisms now on *both* CPU and GPU. What is left to manually tweak and change?
Oh it's still there if you wanna look for it. Just in different places. With Ryzen in particular it's ALL about dat memory. Really gotta go balls deep into them subtimings if you wanna make her purr. Expect hours and hours of bluescreens playing with like 5-6 different voltages, and plenty of good old-fashioned anger & frustration and then finally.... Glorious, oh the Glorious significant gains to be had!!! Totally worth it. (Sold my 1700 to a friend and got a 2700X).
Thanks for the advice!
What?
Any thoughts about this?
https://www.newegg.com/Product/Product.aspx?Item=N82E16820232623
Any thoughts on the kit I linked and how well it will work with the 2700X on a ASUS ROG Crosshair VII Hero?
Far more then 2%. Some of the numbers I have seen had me spray milk out of my nose. And I wasn't even drinking milk!I got the TridentZ FlareX 3200Mhz 14 and called it a day lol.
I'll pass on the BSODs and crying kids because they lost all their progress rather than the negligible 2% extra fps
Far more then 2%. Some of the numbers I have seen had me spray milk out of my nose. And I wasn't even drinking milk!
In all seriousness the gains are substantial especially in gaming.
Then you haven't researched much. My right?Link? I haven't seen anythung crazy out of the same stick of ram.
Yeah, RAM tweaking is best done with as few details as possible.This is about the best information I've found but they don't give many details: computerbase.de
I've seen bigger improvements mentioned but they had even less context.
Then you haven't researched much. My right?
Not higher clocks, but low latency and tight timings. The stilt (apparent current expert on this) says he would rather have super tight 3200MHz the CL16 3600 on Ryzen.I've looked at some charts, but it was large jumps. 2133, 2400, 3200, 3466 Ect. Didn't look like enough gain for me to go through the instability of trying to get another 200 Mhz out of my 3200Mhz ram... you seemed to think it was more than a negligible jump. I thought maybe you had a review in mind that made you say this.
Not higher clocks, but low latency and tight timings. The stilt (apparent current expert on this) says he would rather have super tight 3200MHz the CL16 3600 on Ryzen.
Exactly right.It is a combination of both. Clocks > 3000 + CL14 or less = massive jump in Ryzen gaming performance. See attached chart. (Note, LL means the massaged timings. I think he was using CL14, but I will have to go back and check later).
It is a combination of both. Clocks > 3000 + CL14 or less = massive jump in Ryzen gaming performance. See attached chart. (Note, LL means the massaged timings. I think he was using CL14, but I will have to go back and check later).
Exactly right.
That was the chart I was trying to dig up. Those LL scores are drastic.
Your showing nothing to do with LL tight timings. CL16 at various speeds has minimal returns as you have linked. This is not what we are talking about.If you need those FPS go for it (VR or high Hz monitors) I already have 3200 14CL RAM which puts me way above the 90Hz I need for VR and from what I've seen I could hope for 1-2% more fps if I bothered to tweak it... even less in a realistic scenario.
At normal frame rates RAM speed matters much less.
https://www.anandtech.com/show/11857/memory-scaling-on-ryzen-7-with-team-groups-night-hawk-rgb/6
http://www.legitreviews.com/ddr4-me...tform-best-memory-kit-amd-ryzen-cpus_192259/4
If you have the time and will have at it but be realistic in what typical scenarios will see for returns.
The point is it's at diminishing returns with slower ram at real world FPS. Even faster ram is even further along the curve.Your showing nothing to do with LL tight timings. CL16 at various speeds has minimal returns as you have linked. This is not what we are talking about.
Well, then there's never been a better time to get yourselves all nice & reacquainted! Because as long as AMD continues to so tightly couple the Zen's uncore (with the IMC being kind of the chip's Infinity Fabric bus command & control center; including providing their clock signal) & the ever approaching fab process wall, it's gonna be all about memory first and foremost from here to ad nauseum.Cooe I haven't done any serious memory tuning since my DFI days (A64 and Core2).
How does the 2700 compare to the 2700x? 65w vs. 105w ratings -> Does the performance show that much difference? Does the nonX version not have PB2? Anyways if the 2700 is truly a 65w part that maybe rather fantastic for small system builds if the performance is there. Hopefully that comparison is in the works.
How does the 2700 compare to the 2700x? 65w vs. 105w ratings -> Does the performance show that much difference? Does the nonX version not have PB2? Anyways if the 2700 is truly a 65w part that maybe rather fantastic for small system builds if the performance is there. Hopefully that comparison is in the works.
Stilt's Ryzen 1st Gen B-die timings (the 14-13-13-13-26 ones; I have B-die DIMMs but dual-rank [16GBx2] so required some extra juice) is where I started with both my timing tightening quests for both my 1700 & 2700X. If your chip/DIMMs can handle those, that's a fantastic place to start, and has a number of easy improvements you can still make from that starting point (prolly not gonna get anything further outta the primary's, but the subtimings are ripe for further optimization if one's willing to invest the necessary time & effort. Dropping both tRDRDSCL & tWRWRSCL to their minimum possible values of "2" for example [or as low as you can get them] gives decent, easy gains & is prolly the first place I'd start).Not higher clocks, but low latency and tight timings. The stilt (apparent current expert on this) says he would rather have super tight 3200MHz the CL16 3600 on Ryzen.
The testing I did, 3466 memory with Cas 14 (Stilts 3300 safe sub timings) beat out Stilt's 3200 Fast settings (Cas 13 . . .). Only a few programs actually showed any significant advantage, for games it was the minimal FPS speeds that gained the most. Sitting at the computer not knowing if it was 3200 Cas 14 or Stilts Safe 3200 speeds that has Cas 14 timing or Stilts Fast with Cas 13 or 3466mhz with fastest sub timings supported- I would be very surprised if anyone could accurately tell the difference other then certain benchmarks. Any intensive CPU, cache based type programs will see insignificant improvement from 2666mhz to 3600mhz. Hitman was the game that showed the most improvement - other games less to much less improvement with memory speeds.Stilt's Ryzen 1st Gen B-die timings (the 14-13-13-13-26 ones; I have B-die DIMMs but dual-rank [16GBx2] so required some extra juice) is where I started with both my timing tightening quests for both my 1700 & 2700X. If your chip/DIMMs can handle those, that's a fantastic place to start, and has a number of easy improvements you can still make from that starting point (prolly not gonna get anything further outta the primary's, but the subtimings are ripe for further optimization if one's willing to invest the necessary time & effort. Dropping both tRDRDSCL & tWRWRSCL to their minimum possible values of "2" for example [or as low as you can get them] gives decent, easy gains & is prolly the first place I'd start).
And to whomever said there's no noticeable gains to be have from manually tightening your memory timings & maximizing it's clock-speed either has never done it or looked at results from those whom have, or is absolutely freaking delusional. The potential gains are HUUUUGE in comparison to just about any other tweaks you can do with Ryzen, & 2nd Gen in particular. My 3200MHz CL15 kit after applying my current ridiculously tight primary & sub memory timings smokes even vastly pricier 3600MHz CL16 kits. Improving latency at both the memory level (by tightening timings), and cache level (by increasing DDR4 clock & thus Infinity Fabric speed) are absolutely KEY to maxmizing Ryzen's performance, and ESPECIALLY in CPU limited gaming scenarios. That's why AMD focused on that area in particular so intensely for Zen+.
You compared awesome with awesome, so we could expect close results. But Joe blow at 3200MHz CL16 (or far, far worse) being standard Ryzen proud to hit 3200MHz, isn't in the true zone of what these chips can perform.The testing I did, 3466 memory with Cas 14 (Stilts 3300 safe sub timings) beat out Stilt's 3200 Fast settings (Cas 13 . . .). Only a few programs actually showed any significant advantage, for games it was the minimal FPS speeds that gained the most. Sitting at the computer not knowing if it was 3200 Cas 14 or Stilts Safe 3200 speeds that has Cas 14 timing or Stilts Fast with Cas 13 or 3466mhz with fastest sub timings supported- I would be very surprised if anyone could accurately tell the difference other then certain benchmarks. Any intensive CPU, cache based type programs will see insignificant improvement from 2666mhz to 3600mhz. Hitman was the game that showed the most improvement - other games less to much less improvement with memory speeds.
I should redo this, especially if I get a 2700K. Anyways I went to 32gb (4 sticks of TridentZ 3200 cas 14 ram) running it at 3200mhz Stilt's safe preset which is actually much tighter than D.O.C.P memory settings (X.M.P).You compared awesome with awesome, so we could expect close results. But Joe blow at 3200MHz CL16 (or far, far worse) being standard Ryzen proud to hit 3200MHz, isn't in the true zone of what these chips can perform.
Now that is what I call closing the gap!
Faster is faster. Whether your system is capable of the benefits is irrelevant to the focus of the thread.Another note on ram speeds, while some of the numbers may look greatly improve like 99% percentile, for the user with 60hz monitors it usually is meaningless since FPS is already faster than refresh rate or you are at higher resolution making the GPU the restriction. So in essence I am monitor/GPU limited while faster memory timings only just give better benchmark results but no significant user experience improvement due to other restrictions. My thoughts at least on that.
Faster is faster but also may be irrelevant in the end. Like OCing the 2700x, the gains are so pitiful compared to smart XMR2 and PB2 it becomes irrelevant.Faster is faster. Whether your system is capable of the benefits is irrelevant to the focus of the thread.
10-15% in select applications (games) is almost as much as we had from SB to Skylake, certainly not pitiful.
Well, if you checked the video Steve used Flare X here, with the baseline result representing standard 3200 CL14 (5:30 in the video). If you push the same ram to 3400 speed and use timings from The Stilt, you can net 10 - 15%, in three out of four games.
He promised another test with 30 games, looking forward to seeing those results.