Jump to content

Ryzen Build


32 replies to this topic

#1 NARC BAIT

    Member

  • PipPipPipPipPipPipPip
  • Ace Of Spades
  • 518 posts
  • Twitch: Link
  • LocationAustralia

Posted 12 July 2017 - 08:33 AM

so, a few days ago, well, nearly a week at this point ... I went out and brought some new parts ... and while it may not be as pretty as Bill's build, overall I've really got to say that I'm quite happy, although, there are a few niggling things .... firstly though, some parts

CPU ) Ryzen 1600X ( I wanted the X model for the higher TDP ... )
MEM ) 8gb (2x4gb) Corsair Vengeance LPX CL19 @ 4000 MHz ( on special! )
MOBO) ASUS CROSSHAIR VI HERO x370
I got a cheap random thermaltake AIO cooler, 120mm

GPU ) existing 6gb GTX 1060 with an AIO cooler strapped on
HDD ) existing ones, primary is SATA SSD, haven't got that M.2, yet ...
MON ) ACER CB280HK, a bare minimal 4K screen, its ok, but ...

did a fresh windows install, because it would have been impossible to 'clear' all the registry tricks and stuff I've done in the past to get a point that I might have considered the system 'clean' .... some baseline cpu benchmarking revealed an expected single core performance increase over my previous FX cpu, and I was ready to have a few games ... and pretty much straight off the bat, with no real tweaking, I was pretty happy ... so, then of course, onto some tweaking ...

I was unimpressed when I had dug deep enough to find out that I hadn't really 'upgraded' from my FX at all ... because apparently the ryzen consists of multiple FX's on one die ... so much for the much heralded 'redesign' ... sure, its not exactly the same, but in a lot of ways, it actually is, and the key point of difference between the two, comes down quite simply, to the fabrication size, the vishera was 32nm, and the ryzen is 14 ... do the math, you can fit two 14's side by side within the space of the 32, no problem ... and we have added some extra instructions along the way, yeah, probably more useful in the coming years than right now, but ok, and what probably has the biggest performance impact is the generational increase to DDR4 and the electrical pathway shortening that comes from miniaturizing the FX ....

once this becomes apparent, people will start accepting the 'voltage wall' for what it truly is, a hard limit on just how much juice you can squeeze into such a small area, versus how much you can dissipate from that small area .... I didn't win the 'silicone lottery' with my ryzen, so she likes a fair bit of juice to sit on 4ghz, but beyond that, well I haven't got the ability to dissipate the heat required for the voltages that get insane pretty quickly beyond 4.2 ghz, I get the feeling that most people never even manage to get that stable, but maybe its more because they get afraid of the requirements .... anyway

DDR4 is a whole different beast to DDR3 .... and I felt like I had stuff to relearn ... but ... well ... not much really ... but the whole 'training' behaviour is not well explained in any of the videos, most of them just shrug as the pc obviously restarts 5 or 6 times ... with my kit, when I dial it in right, it will boot first time ....

the default auto settings had it set to 2133mhz, clearly not what one would intend with a 4000mhz kit ... I completely expected to be doing a bios update before I purchased the board, but I wanted to try before I did it anyway, and had no real luck with frequencies higher than 2933mhz .... so I grab the bios update for the board, jam that in, and I'm now able to lock in 3200mhz and 3600mhz, even though AGESA 1.0.0.6 is meant to support 4000mhz, I couldn't get it to play, and I couldn't find reference from anyone else that said they had ... although 3600mhz is technically faster than 3200mhz, the latency (and possibly FSB ratio at 3:54, like who does that ...) means that in the real world, its doesn't come out much faster, the difference between 2133mhz and 3200mhz is very noticeable on the other hand ....

this is one area that the motherboard really shone through, it has a boot counter, and if your not in the know, the boot counter displays a 2 digit code to tell you what the system is doing, before it gets around to turning the screen on .... and without that I probably would have just been another dude shrugging his shoulders ....

I've noticed in the board configuration that amonst the myriad of options is the ability to setup the core configuration, and two of the options had me thinking, the options are (3+3) and (4+0), implying that my 6 core, actually has at least 7 .... only time will tell if we end up with manufacturers offering methods to 'unlock cpus' as they have in the past ... its probably bad form to do it to a platform so early on ... who would buy the flagship models if we all just planned to unlock 1-4 more cores on every quad ....

#2 Bill Lumbar

    Member

  • PipPipPipPipPipPipPipPipPip
  • Death Star
  • 2,073 posts

Posted 25 July 2017 - 03:06 AM

Hey Narc....

Congrats on the new build man. I am not sure why you are having issues with your build or believe its not much of an upgrade from your FX. My 1700 replaced my I74790K @ 4.8 ghz and I noticed a improvement over the Intel. I am sitting at 4ghz on the Ryzen, 3200mhz on the ram, and I am loving this new build.

Don't sweat your build not being as pretty as mine, I am a over builder on everything I touch, and I love building custom **** man. Posted Image Posted Image Posted Image

You should be able to sort all your issues out over the next few months. I do not believe Ryzen is an FX core just shrinked man. If you want, please explain

#3 Flapdrol

    Member

  • PipPipPipPipPipPipPipPip
  • The 1 Percent
  • The 1 Percent
  • 1,986 posts

Posted 26 July 2017 - 06:09 AM

View PostNARC BAIT, on 12 July 2017 - 08:33 AM, said:

I was unimpressed when I had dug deep enough to find out that I hadn't really 'upgraded' from my FX at all ... because apparently the ryzen consists of multiple FX's on one die ... so much for the much heralded 'redesign' ... sure, its not exactly the same, but in a lot of ways, it actually is, and the key point of difference between the two, comes down quite simply, to the fabrication size, the vishera was 32nm, and the ryzen is 14 ... do the math, you can fit two 14's side by side within the space of the 32, no problem ... and we have added some extra instructions along the way, yeah, probably more useful in the coming years than right now, but ok, and what probably has the biggest performance impact is the generational increase to DDR4 and the electrical pathway shortening that comes from miniaturizing the FX ....

Ryzen has very little in common with the FX and outperforms it massively, especially in gaming.

#4 NARC BAIT

    Member

  • PipPipPipPipPipPipPip
  • Ace Of Spades
  • 518 posts
  • Twitch: Link
  • LocationAustralia

Posted 27 July 2017 - 06:52 AM

View PostBill Lumbar, on 25 July 2017 - 03:06 AM, said:

Don't sweat your build not being as pretty as mine, I am a over builder on everything I touch, and I love building custom **** man.

I do think about your build ... because I found on mine, that I needed to put bonus cooling fans over the VRM, and over the PCH (southbrige), just be sure that I was going to get the mileage out of it with the extreme conditions I was putting it through ...

View PostFlapdrol, on 26 July 2017 - 06:09 AM, said:

Ryzen has very little in common with the FX and outperforms it massively, especially in gaming.

if you look real hard ... nowhere will you find anyone from AMD say that it was a total redesign ... its the evolution of the product .... the momory controller has gone back to being separated, 'modules' are now just 'cores', and the FPU situation ... well ... that's different ... that's the only place we are really different ... but in the same way, its an echo or an ode to the past that brought us here, each CCX shares one FPU, and that one FPU is capable of servicing 10 tasks, almost at once, they do have to run through a scheduler, which is tied to the infinity fabric speed, so the quicker the memory speed, the quicker FPU tasks return a result .... might be interesting to do some FPU based comparisons on ryzen 3/5/7's to see if the FPU on the 3 is just as good as on the 7 ... it should be ...

excerpt - https://support.amd....els_00h-0Fh.pdf

AMD said:

2 Core Complexes (CCX). The Core represents the x86 ISA core from AMD designed for FX and 7th generation APU offerings.

so, I'm sure there are more similarities, than there are differences between the two chips .... and if you really want to make a ryzen act like an FX, just disable one CCX, tone the memory down to a DDR3 speed, and checkout how similar it is ...

if you ever spent time overclocking a 95xx series FX, its gonna feel like a stroll down memory lane ... and just like before, make sure to have a fire extinguisher handy, or be prepared to be looking at LN2 ...

initially I guess I was just hoping for a wider degree of separation, but what we really have is it just acting more like it should have been in the first place ... like windows vista to windows 7 ... 7 delivered what vista promised ... ryzen delivers what we didn't get for FX ...

and don't get me wrong, not one day have I been 'unhappy' about buying any part of it .... although, I do wonder, if I would be able to lock in 4000 MHz ddr if I had brought the sticks rated at 4266 MHz ...

god only knows how long I will play with the configuration for it ... probably wont stop until I'm over the rated frequency of the RAM ... in terms of other configuration, I can clock the ram up to 3733 MHz ... better than most, but at the moment I'm sitting at 3200mhz and lowering the latencies .... it seems to me, that the XMP profiles are not absolutely the best speeds a stick can do ... and then theres the FSB ratio's ... which get weird beyond 3600 MHz ... if I post at 3600mhz, it uses 3:54 ... but if I overclock there, it uses 3:52 .... none of which really make sense to me, and I'm leaving trying to attain higher speeds for more microcode updates from AMD, nearly everyone else gave up at 3200 MHz, so I guess I cant complain too hard there ... https://rymem.vraith...etailed_ram/984

I've got SMT disabled on my hex core, it seems to me that in synthetic benching, enabling SMT seemed to yield about an extra two cores worth of raw integer speed, assuming that the task was able to thread well ... but at the end of the day, only two factors need consideration, realistic overall CPU usage / ability to dissipate heat .... in terms of usage, while playing MechWarrior, browsing the internet ( x20 windows because I'm lazy ), recording the game at 60 fps, playing music from one sound card to another to mix back in and whatever else, I'm struggling to see more than 50% usage ... maybe I should play something else, for comparison ... but the last game I brought, GTAV, isn't overly appropriate with a 10 year old sitting next to my screen ...

but MWO is 'code bound', which you might hear inappropriately described as 'CPU bound', with the result at the end of the day being pretty much the same, the faster a single thread executes, the more frames, might be rendered, at any given moment ... and when you disable SMT, each core executes faster, not by a lot, but faster is faster ... and MWO is only designed to take advantage of 6 cores anyway, at best ... and my sweet spot for heat dissipation, is right around 4120 MHz on the CPU, during a long heavy load, it will heat up, to around 60c-70c, taking the temperature offset into account, but I mean a really heavy load, like overnight stress testing for hours and hours .... on 'average' it sits at about 40c for 25% usage .... and in terms of synthetic benchmark scores, it comes in between a 6700k and 7600k for single thread performance, and BLOWS them away for multi core, as you would expect with 50% more actual cores ...

and that's about where I come to my last, issue / non issue for the moment, two voltage wall ... when I juice it too much, the processor goes into a failsafe .... ugh ... meaning that I haven't gotten anything past 4.2 ghz 'stable' .... when that load goes to 100%, she hits that failsafe as the current gets around ( I assume over ) the TDP ... I'm not sure if this is a feature of the X models, the motherboard or something that isn't set right within the BIOS ...

maybe tomorrow I will set the temperature offset +50% (or more, that drops it to about -1) and mess around in LN2 mode ....

#5 NARC BAIT

    Member

  • PipPipPipPipPipPipPip
  • Ace Of Spades
  • 518 posts
  • Twitch: Link
  • LocationAustralia

Posted 27 July 2017 - 07:36 AM

so, I've pretty much found my limits ... and my ram seems to be slightly better @ 3600 MHz, using an XMP profile which is only meant to be good for just less than 3200 MHz ... yielding a saving of around 8% on memory latency .... benchmarks suggest slightly higher throughputs, and it reacts, even quicker, yet again .... https://valid.x86.fr/u4yilz .... this time giving a slightly better multi core score, implying to me that the cores were loaded more often ... anyway, its past my bedtime ...

#6 Kill Chain

    Member

  • PipPipPip
  • 71 posts

Posted 27 July 2017 - 05:04 PM

I have a Ryzen 7 1700 that I've barely OC to 3.4ghz and I run everything butter smooth on max settings with my R9 290X (aging well). My only gripe is that I can't get my 3200mhz ram over 2933mhz on my MSI Gaming Pro Carbon MB with the last bios update.

#7 NARC BAIT

    Member

  • PipPipPipPipPipPipPip
  • Ace Of Spades
  • 518 posts
  • Twitch: Link
  • LocationAustralia

Posted 27 July 2017 - 06:17 PM

View PostKill Chain, on 27 July 2017 - 05:04 PM, said:

My only gripe is that I can't get my 3200mhz ram over 2933mhz

yeah this is the thing going around ... a lot of people cant get to the speeds that the ram was rated at ... and whos fault is it ... well, its intel's fault ... ( pause for fanboy flame war ) ... but can you really blame intel , for it being their fault ... no ... not really ... intel has ended up being highly influential with manufacturers for DDR4, resulting in XMP ... AMD decided not to play ball with DDR4 up until now, and have a lot of catching up to do .... some might say this was stuff that should have been more sorted before the launch ... if you look around you might find stuff about AMP profiles ... because AMD cant even use the terms that intel basically got to dictate to the market ... and from what I'm seeing, XMP profiles don't line up optimally on the AM4 platform .... with the genral advice from AMD being, set your timings manually ... but it seems to me, that I can set tighter timings than what the XMP profiles suggest ... and if I leave it to auto training, I get a totally different set of values ... I've swapped back to 3600mhz with tighter latencies than the XMP profile ... but I probably wont stop fiddling with it for a few days ...

when an XMP profile is used with an intel, it raises the FSB, and everything is highly tuned for something other that what we have ... like strapping the ECU from a Honda onto a Chevrolet, it just seems to be spitting out random values ...

it seems to me, that to help DDR4 training on my ASUS x370 CH6, enter the tCL value from the XMP profile, and leave the rest set for auto ... your probably looking at a value of 15 or 16 ... and all the others will *probably* train to be values slightly higher (higher latency equals slower) than suggested by the XMP ... if that comes out stable, drop the value by one, stability test, rinse and repeat ... the tighter you get those timings, the 'snappier' it will feel ...

#8 Bill Lumbar

    Member

  • PipPipPipPipPipPipPipPipPip
  • Death Star
  • 2,073 posts

Posted 30 July 2017 - 05:01 PM

I have not bumped my ram up past 3200 yet.... but I am sitting at 4.1ghz stable, 1.40 volts. Loving it Posted Image Posted Image

#9 Metaccini Alfredo

    Member

  • PipPip
  • 25 posts
  • LocationMaui

Posted 01 August 2017 - 07:09 AM

my trusty old intel 3770k finally failed me (4 or so years at 4.9ghz! it's actually the motherboard that failed, cpu is fine) so I popped to the store and picked up 32gb of g.skill ddr4 3200, a ryzen 7 1700, and an asus crosshair vi motherboard. After assembling everything and settling in at 4.1ghz with the ram running at the expected 3200... the system feels just "ok". MWO performance is down substantially from the 3770k, even at stock clocks, but the FPS "dips" don't occur nearly as often. So, lower average FPS, but fewer "dips". I was a little disappointed, but it IS nice to be able to run all my audio software, web browser, obs, a bunch of other applications with no real slowdown.

#10 Lizardman from Hollywood

    Member

  • PipPipPipPipPip
  • Philanthropist
  • 135 posts

Posted 01 August 2017 - 07:21 AM

As stated above, built a 1700 system ran mwo off an nvme used a gtx1080 16g of 3200 gskill ram on an asus hero 6 mwo performance was barely above my old i7 2600 with a 970 and the game on an ssd. Then I built a i7 6850k with same gpu and 32gb of 3k ram and I get 40% more max and average framerate. Mwo hates ryzen cpus, plain and simple. 0 optimization for them yrt, and I really doubt that they ever will optimize for it. If mwo is your main game i cant stress enough to build an Intel system. Even an old z170 and 6600k just totally trashes even the most expensive ryzen set up.

#11 Bill Lumbar

    Member

  • PipPipPipPipPipPipPipPipPip
  • Death Star
  • 2,073 posts

Posted 02 August 2017 - 02:35 AM

I Just have to laugh at the last two replies..... I am coming from a I74790K at 4.8ghz 24/7 stable and I am not having any issues gaming at 1440P on a 32" BenQ monitor. My Ryzen 1700 at 4.1ghz stable handles this extremely un-optimized game with no real issues other than the ones everyone has.

Sephrus, not trying to call you out man, but you claim you built a 6850K with just a 970 card running and you are seeing 40% greater frame rates then on a Ryzen 1700 system? If that is the case.... something isn't right with your system, you need to figure that out. lol. I am running the 1700 at 4.1ghz, and a seahawk 1070 8gb card.... and even in this wonderful game that is so optimized for Intel systems, I am pushing 60-100 fps with very little drops, and butter smooth game play at 1440P.

I'm sorry, but if you guys are gaming in 1080P you are doing it wrong. If you are basing your claims of Ryzen getting trashed based on that, you are doing it wrong, lol. But, to each there ownPosted Image

Wired wanted me to run test on my system, and throw up some numbers and bench marks, but I haven't had a lot of free time right now. Hardly even playing the game, working on rebuilding my 3.5 in my Altima due to a headgasket going bad. First engine rebuild... taking my time and doing it right.


The things I have noticed with my system:

Ryzen 1700 @ 4.1ghz 1.40 volts EK Water blocked
Asrock Professional X370
16gb G.skill amd ram 3200mhz
msi Seahawk 1070 8gb EK water blocked

Amd Ryzen master OC software kinda is buggy. If I OC with it, it bumps my ram back down to 2400mhz when I reboot the system. I stopped using it, and OC from the bios now. No more issues with Ram speed dropping back down to 2400mhz. Maybe this is your guys issue? Ryzen loves 3200mhz+ ram.

This platform is really new yet, bugs are being worked out, fixed.... I believe it has a lot of room to improve yet.

Edited by Bill Lumbar, 02 August 2017 - 02:48 AM.


#12 NARC BAIT

    Member

  • PipPipPipPipPipPipPip
  • Ace Of Spades
  • 518 posts
  • Twitch: Link
  • LocationAustralia

Posted 02 August 2017 - 07:50 AM

I little while ago I read a ryzen opinion piece, that I'd have to say I agree with, AMD/ryzen is the platform for true enthusiasts, because your going to spend a lot of time tweaking things ... and for me, that's fine .... a few days ago, the wife asked me if I needed to take something back to the shop, because my pc had been crashing all day long ... until I told her that I was making it crash, and it was only the first year ....

I'm certainly glad that I spent the money on the CH6, even though they just announced a 'newer' version ... its pretty much going to have the same core on it and the bios's will be mostly interchangeable ... that being said however, there are some things to watch out for ...

by default, my NVME will auto train in gen 1 mode, sounds right, that's auto ... but in actuality, its, at best, 50% of gen 3 speed ... and to actually get the maximum rated speed out of the Samsung SSD, is apparently beyond reach for now ... but whatever ... apparently pulling up 2000 MB/s uses plenty enough CPU IO, and I really don't pull up that much data, sequentially, ever ... but the auto settings are dismal, compared to what the potential promised by the box is ....

if you really want the fastest performance in MWO, disable SMT ... SMT may provide more 'raw number' power over for multicore situations, but it detracts from the single core performance, and if you've ever wanted more performance from this game, that's where this forum says its all at ... intel killed AMD in the single thread arena, and that's why they dominated the last few years ... but my cpu-z bench ( https://valid.x86.fr/zqcp74 ) looks fairly good to me ... overall, I've lost about 20% multithreaded performance, and gained far more than that in single core performance, which is what matters for MWO .... and generally all that matters for me .... I'm not doing transcodes, or folding, or blah or meh, this is merely a system for use, and if I click that button, it had damned well better show me that info, before my finger releases ....

other than benchmarking, how often is your CPU actually redlining anyway ... mine isn't doing it very often ... and turning off half the cores, lowers the juice that its drinking, and lowers the heat, making the thermal profile much nicer .... at the moment I'm floating between 1.438v and 1.461v for 40c actual temp ... it actually runs that cool, that the AIO seems, excessive ... anyway ...

which brings me to problem number three ... back to RAM ... such fun with RAM, for the enthusiasts right ... so I cant get my promised 4000mhz, and hope that that situation will improve with time, as has to get to current speeds, before I had gotten around to investing, and was still merely conjecturing ... but I have found, that even though I can set speeds up to 3733 MHz, its actually not running faster at my end, well not completely anyway, and identifying the weirdness took me a bit ... and before you say it, no, its not related to latency .... beyond 3433 MHz, I take a 50% hit to memory read throughput ... kind of severe, but still, fast enough, that the problem isn't as noticeable as you might think, because well, 27,000 MB/s isn't exactly slow either ... but clocking in at 3433 MHz, yields me 49,000 MB/s .... latency is a whole other story, AMP profiles would be nice, but that isn't what I've got, or whats available in my neck of the woods ...

in terms of synthetic benchmarks, its my GPU that bottlenecks now ... and that actually makes me happy ... it tells me that I could upgrade my GPU from the 1060 and get an improvement ... but for now, I just drop my resolution slightly from 4K, to be around 3.5 .... but the monitor that I use, doesn't do so well above 60 hz, so I I've ended up conceding to vsync, which generally, I've always found detestable ... after so many years of MWO, being well, MWO and laggy, and buggy, and annoying, I'm just happy enough at 60 FPS ...

in terms of optimization, the current engine will never be properly optimized ... even if you managed to pinpoint the system that causes the problems, you couldn't get them to fix it ... not in any reasonable time frame ... that's my experience ... the least optimized part of the game at the moment, relates to the map, and how often that code is executed, not how often its actually drawn, you can set a delay on how many frames until you re draw it, but the underlying flash actionscripts will execute ALOT more than that, so often in fact that its mind boggling, and was clearly never checked for metrics .... it scaled fine when the game was 8v8 .... and will get better for most players as people die off and stop requiring updates ....

#13 NARC BAIT

    Member

  • PipPipPipPipPipPipPip
  • Ace Of Spades
  • 518 posts
  • Twitch: Link
  • LocationAustralia

Posted 02 August 2017 - 08:23 AM

oh, another note, I've noticed that running MWO and OBS with the realtime priority helps them stay most responsive .... and I still use a custom user.cfg ... I'll post that when it becomes more solid ...

#14 Lizardman from Hollywood

    Member

  • PipPipPipPipPip
  • Philanthropist
  • 135 posts

Posted 02 August 2017 - 09:18 AM

Nope gtx 1080 on the 6850k. I avergage 200 to about 80ish fps mostly butter smooth. I had my 1700 @4.0 on an aio for gaming. Admittedly that was April/May and some bios improvements and win 10 improvements have happened since then that may be helping you out. I plan on revisiting the ryzen build and updating everything and installing a bright shiney new 1070 in it so me and my son can game in the same room.

To be honest the perfoemance of the Ryzen stumped me. It never went above 20% usage in MWO. It ran the same frame rates no matter the graphics setting i set it to and it gave me the same FPS at 1080 on one screen and 1440 spread across 2 monitors flipped vertical. But like i said getting frame rates from 120 dipping to the 40s. The averages were fine but the dips were killing me.

#15 Tarl Cabot

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • Tai-sho
  • Tai-sho
  • 7,642 posts
  • LocationImperial City, Luthien - Draconis Combine

Posted 02 August 2017 - 02:34 PM

View PostSephrus Shanadar, on 02 August 2017 - 09:18 AM, said:

Nope gtx 1080 on the 6850k. I avergage 200 to about 80ish fps mostly butter smooth. I had my 1700 @4.0 on an aio for gaming. Admittedly that was April/May and some bios improvements and win 10 improvements have happened since then that may be helping you out. I plan on revisiting the ryzen build and updating everything and installing a bright shiney new 1070 in it so me and my son can game in the same room.

To be honest the perfoemance of the Ryzen stumped me. It never went above 20% usage in MWO. It ran the same frame rates no matter the graphics setting i set it to and it gave me the same FPS at 1080 on one screen and 1440 spread across 2 monitors flipped vertical. But like i said getting frame rates from 120 dipping to the 40s. The averages were fine but the dips were killing me.

Did you also set Power Options to High Performance?

#16 NARC BAIT

    Member

  • PipPipPipPipPipPipPip
  • Ace Of Spades
  • 518 posts
  • Twitch: Link
  • LocationAustralia

Posted 02 August 2017 - 07:15 PM

View PostSephrus Shanadar, on 02 August 2017 - 09:18 AM, said:

To be honest the perfoemance of the Ryzen stumped me. It never went above 20% usage in MWO. It ran the same frame rates no matter the graphics setting i set it to and it gave me the same FPS at 1080 on one screen and 1440 spread across 2 monitors flipped vertical. But like i said getting frame rates from 120 dipping to the 40s. The averages were fine but the dips were killing me.

if your not really doing brute force number stuff with it, I'd highly recommend disabling SMT, reducing the core count, to the amount of cores that there are, actually ... but the overall average cpu usage is *NOT* a fair indicator of actual usage, during heavy combat, CPU usage stays between 30% and 50% overall, but, when you look at it a bit deeper, the usage per each core is not so evenly spaced or scaled .... one core might have more actual load on it, and I see individual core usage get as high as 80%, for a moment ... if I was using the stock clocks, that 80% would probably have red lined a core, and caused a stall ...

MWO uses cryengine, and its configuration actually makes sense for 6 cores, its the maximum that was available on any machine during its design, which, by the way was an AMD ... its not that cryengine have never supported AMD, this has been a PGI bias ... tasks are not evenly spread across cores ... I can have one core pushing 60% usage, and the one next to it, pushing 5% usage ... and if you've got SMT on, well, its probably only going to be, more ridiculous over 16 cores ...

if I let my gtx1060 do resolutions under 4K, it barely raises a sweat ... ummm how do I express this properly, mmmmm pixel dimensions .... my 4K screen is 8.3 million pixels per frame, the given usage example of 2560x1440 + 1920x1080 ... both screens together work out to be about 5.7 million pixels ... and if only ne of them is doing anything 3d, and your not watching x264/5 videos on the other panel, its not going to increase the load all that much ...

#17 Bill Lumbar

    Member

  • PipPipPipPipPipPipPipPipPip
  • Death Star
  • 2,073 posts

Posted 03 August 2017 - 03:43 PM

View PostTarl Cabot, on 02 August 2017 - 02:34 PM, said:

Did you also set Power Options to High Performance?

I have mine set to AMD's Balanced performance for Ryzen's Cpus..... It seems to be working well. I had it on High Performance. Anyone else have it on this setting? Issues?

I am also running on a Samsung SSD Think its a pro, 512gb, still haven't upgraded to the 1tb M.2 drive yet. I want to, but after building this new rig, and all the tools I needed to buy to fab the new aluminum wall mount case, I have kinda stalled on blowing another $500 on the drive. I mean, how much does one have to throw at parts, and when does it really start to make a real world difference? Posted Image

This is MWO after all..... I have seen guys with $10,000 rigs still seeing drops in FPS with this game.... its funny.... or kinda sad really. Posted Image

I am still on the install of Windows 10 ported from my Intel I74790K build.... I did not do a clean install. I just uninstalled my AMD 7970 drivers, and let Windows do the rest when I set up my account on sign in with Microsoft vs. a local account. So... as far as I can tell, Windows picked up the new hardware with the Ryzen build, and I have had no issues. Normally, I always do a clean install, always. I did get the product key from the Microsoft tech that took remote access of my computer when I couldn't activate my key, so if I need to do a clean install, I have the key.

Edited by Bill Lumbar, 03 August 2017 - 03:56 PM.


#18 Lizardman from Hollywood

    Member

  • PipPipPipPipPip
  • Philanthropist
  • 135 posts

Posted 03 August 2017 - 04:35 PM

I just find it a bit silly that i have to detune my cpu in order to make MWO run better. For trying to ""grow" the game they sure do suprisingly little to it besides adding mech packs. But thats a rant for another time. When i finally get home Im gonna try out your idea narc and use the master tool to disable smt and chop it down to 6 cores.

#19 NARC BAIT

    Member

  • PipPipPipPipPipPipPip
  • Ace Of Spades
  • 518 posts
  • Twitch: Link
  • LocationAustralia

Posted 03 August 2017 - 07:14 PM

View PostSephrus Shanadar, on 03 August 2017 - 04:35 PM, said:

Im gonna try out your idea narc and use the master tool to disable smt and chop it down to 6 cores.

I gave up with ryzen master .... I found that it liked to run more voltage through the cores than 'needed', and in terms of cpu cycles used to get simple monitoring information, it becomes a joke ... not to mention that it limits to total memory frequency range, I tend to like lean software, and it certainly isn't ...

in terms of 'growing' the game, they probably don't seem very proactive on that front because they are co developing two titles at the moment ... and only time will tell wether the multiplayer side gets any benefit from the engine switch that we might never see in the multiplayer ....

the artists make models etc in other software and then import them into the engine ... so most 'assets' should benefit both games .... hopefully when they write the underlying game code into the new engine, they don't just 'adapt' the bad bits, I'm looking at you battlegrid, the pre alpha footage from a while back had nothing like a map in it .... but only time will tell if the two projects get unified down the track ....

but we desperately need an engine update .... the current engine is from 2012, with some updates from 2013, but it was designed in 2010/11 ....

#20 Tarl Cabot

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • Tai-sho
  • Tai-sho
  • 7,642 posts
  • LocationImperial City, Luthien - Draconis Combine

Posted 05 August 2017 - 02:05 AM

View PostBill Lumbar, on 03 August 2017 - 03:43 PM, said:

I have mine set to AMD's Balanced performance for Ryzen's Cpus..... It seems to be working well. I had it on High Performance. Anyone else have it on this setting? Issues?

.....

AMD actually recommends High Performance when the Ryzen came out. Though they said they would provide updates to better utilize the OS balanced plan, I can not say how that is doing.

https://community.am...md-ryzen-system


Quote

Mind Your Power Plan

Make sure the Windows® 10 High Performance power plan is being used (picture). The High Performance plan offers two key benefits:

Core Parking OFF: Idle CPU cores are instantaneously available for thread scheduling. In contrast, the Balanced plan aggressively places idle CPU cores into low power states. This can cause additional latency when un-parking cores to accommodate varying loads.

Fast frequency change: The AMD Ryzen™ processor can alter its voltage and frequency states in the 1ms intervals natively supported by the “Zen” architecture. In contrast, the Balanced plan may take longer for voltage and frequency changes due to software participation in power state changes.

In the near term, we recommend that games and other high-performance applications are complemented by the High Performance plan. By the first week of April, AMD intends to provide an update for AMD Ryzen™ processors that optimizes the power policy parameters of the Balanced plan to favor performance more consistent with the typical usage models of a desktop PC.

Edited by Tarl Cabot, 05 August 2017 - 02:07 AM.






1 user(s) are reading this topic

0 members, 1 guests, 0 anonymous users