Cat's mini(ITX) Ryzen 5 review!

[CAT-THE-FIFTH]

The prologue!

I have been on socket 1155 since launch,so its served me well as has my Ivy Bridge based Xeon E3 1230 V2 mini-ITX system. However,as time progressed the motherboard has started to have various niggles,like on and off cold boot issues,a broken PCI-E card latch,audio bugs,etc and OFC the Intel security bugs,which caused issues in a game or two I played. However,more tellingly in some non-game workloads the system was also starting to struggle.

I had been hoping to hold off an upgrade until 2019,but ultimately throwing more money at the current system(let alone finding a new motherboard for not silly money) was probably an exercise in futility,so ultimately it was time for the first platform upgrade in 7 years.However,I can put up with the game issues,so ultimately for me any improvements are just incidental.

The story begins

I have not had anything other than a Shuttle form factor or mini-ITX system,since 2005,and going mATX didn't seem cost effective(I would need a new case),so again the system is a mini-ITX one. After looking at the relatively narrow range of AM4 mini-ITX motherboards which were available,I ended up with the Asus ROG STRIX B450-F GAMING being my choice,after a combination of the cheaper ones not being available,having gone up in price,or having design defiencies. Hence,this meant I waited a few weeks,to get a deal on one! Luckily at the same time I managed to bag a Ryzen 5 2600 for well under £140,and got a relatively good deal on some 3200MHZ RAM which I hoped was single ranked Samsung B-die stuff. No,I didn't use the Potato 2400MHZ set I had as that is being ditched.

So in the end this is what I settled on:
1.)Ryzen 5 2600
2.)Asus ROG STRIX B450-F GAMING
3.)16GB of T-FORCE VULCAN TUF Gaming Alliance 3200MHZ RAM.

The reason why I was aiming for single ranked Samsung B-die RAM is Ryzen is more finicky about RAM than Intel CPUs are,and also tends to perform better with faster RAM due to the memory-CCX latency issues. Modern BIOSes due seem more mature though as Hynix single ranked RAM seems OK too,but dual ranked is still more problematic.

Some Unbox Therapy?

Now for a picture of some boxes.



The motherboard gubbins and some interesting bumpf.





I made the Choice of Champions!! Yay! Go me! I didn't have the Breakfast of Champions today though!

Now some board pictures.



The colour scheme is not too OTT,although it does not escape having RGBness(more on that later).

A picture of the underside.



Unique amongst AM4 mini-ITX boards,this one has dual M.2 slots,the top one is PCI-E/SATA and the bottom is SATA only.



The heatsink looks actually useful - its probably needed for the 6 phase VRM!



If you are wondering,the audio ports have LEDs in them.

The backplate gets a special mention - its padded and silver!!



Its so Disco!! Probably fits in with the RGBness of the board!!

The CPU unboxed. The Ryzen 5 2600 comes with the slimline Wraith Spire cooler.





The fan shroud is taller than the actual heatsink!!



The RAM is branded "TUF" and to show it is military camo style. Don't mess with this RAM!!

Also I had to do a double take looking at the board connector - its the first RAM I have seen which has a curved one.





There is not much clearance between the RAM and the cooler shroud.


New cooler installed 10/10/2018

I managed to get hold of a Wraith Spire from the OcUK store to replace the Wraith Stealth as I was not too happy about the temperatures under heavy load(see the later tests).



I removed the Wraith Stealth to check if the cooler was making enough contact and yes it was.



The Wraith Spire is a much more substantial cooler overall.





Yay I got the copper core version of the Wraith Spire - AFAIK there is also a one with a different fan and just a normal aluminium base like the Wraith Spire.



There are definitely no clearance issues with the Wraith Spire!

 

[CAT-THE-FIFTH]

Lets get ready to rumble - benchmarks are INCOMING!!

The test setup.

The main focus of this upgrade was for reliability purposes,and also an upgrade for non-gaming purposes. Hopefully games will improve too!! TBH,for the most part outside 2 games,the old CPU was good enough,but lets find out.

Now inb4 benchmark time,lets get some clear.

The old system had the following specifications:
1.)Xeon E3 1230 V2/Core i7 3770 with a Corsair H40 cooler
2.)ASRock B75 mini-ITX motherboard
3.)16GB 1600MHZ DDR3
4.)Corsair SF 450 SF-X 450W PSU
5.)6 year old Kingston Hyper 3K 120GB as a boot drive
6.)Crucial MX300 275GB as a games,cough,Fallout 4 drive
7.)A 2TB 7200RPM drive
8.)Aluminium Cubitek Mini Cube case
9.)GTX1080FE 8GB

Now I was hoping to get some decent chips in my RAM,and it seems the chips delivered!!







Samsung B-die!! Yay! I applied the D.O.C.P. settings in the BIOS(equivalent to XMP) and it booted fine at 3200MHZ C16-18-18-38. The max officially supported RAM is actually 2933MHZ with Ryzen 2 son this is actually a bit better overall.

Now appparently I can tweak stuff more to gain improved performance,but I decided that would be worth investigating at a later date,as this is more about out of the box easily attainable performance.

The new parts are as detailed before. I will be using the stock Wraith Spire cooler as this is what comes with the CPU and I wanted to see how it performed.

The CPU was run at stock and Asus auto overclocking was switch off.

Unless otherwise stated Windows was set in Windows Balanced power mode for both systems as this is what AMD now recommends,and this is despite the AMD drivers still installing the Ryzen Balanced plan for Ryzen MK!

Nvidia 399.24 drivers were used for the graphics card in both cases.

The Windows version was Windows 10 Pro Edition 1803.

 

[CAT-THE-FIFTH]

Right ho! Benchmarks are ready to go.

In all the benchmarks I will be listing peak power consumption,as this I feel as an important metric when it comes to a SFF rig.

Mr Silvercrest,power meter extraordinaire,take a bow!!

Looking at idle power consumption it varied between 50W~54W for the old setup and between 49W~53W for the new one. So at least when I argue on the internet,Ryzen won't cost me more money. Phew,it had me worried for a moment!

So Cine-matic!

The first benchmark is the perennial benchmark of gamers everywhere.....Cinebench!!



Peak power consumption was 117W during the OpenGL test,68W during the single threaded test and 112W during the multi-threaded test,which had all cores taxed.

Now lets see how Ryzen does.



It ws 50% faster in the OpenGL test,20% faster in the single threaded test,and almost double the speed(97.44% faster) during the multi-threaded test.

This is despite the Xeon E3 having a better cooler - clockspeeds were around 3.8GHZ for the single threaded test and 3.6GHZ for the multi-threaded test.

Power consumption was higher though at 133W for the OpenGL test,73W for the single threaded test and 130W for the multi-threaded test.

Now,moving to the next benchmark,its the creatively named Userbenchmark.

Much User Benchmark...wow??

So here is the score for the old setup:

http://www.userbenchmark.com/UserRun/11204325



Peak power consumption was 109W during CPU only sections of the benchmark and 262W when the GPU ws used.

Now lets see what Ryzen can muster:

http://www.userbenchmark.com/UserRun/11287655



Power consumption was 125W and 265W respectively.

The CPU benchmarks show a good improvement overall - 17% in single core,28% in quad core and 86% in all core performance.

But wait,what has happened to the GPU performance - its dropped?? How,why,wut??



It seems latency was way off - so I tried a few other power plans.

It seems with Ryzen Balanced and High Performance the latency bug isn't there.

http://www.userbenchmark.com/UserRun/11287232





Peak power consumption was at 128W and 266W respectively.

PS:Yes,the storage is in a mess,but I will be changing over to newer drives overtime,and clearing up the old ones a bit.

 

[CAT-THE-FIFTH]

Just Zip(and Unzip) it

Now,onto something totally different. 7-zip!

I used the internal benchmark and stopped it at the 11th run.

First,the Xeon E3 1230 V2.



Peak power consumption was 108W.

Now,the Ryzen 5 2600.



The max power was 125W.

Significant uplifts in performance overall.

Now,picture this!

Now,enough of the synthetic benchmarks! How about something closer to what I want to use the rig for?? How about some RAW conversions - yep that sounds about right.

I use a piece of software called DxO Photolab for managing and processing my pictures - one of the things it does very well is RAW noise reduction via DxO Prime. It can even leverage OpenCL on your GPU. That sounds all fine and dandy,so what is the catch?? Emm,it kind of hammers every thread on your CPU. Oh,OK!

This was tested to great effect during the summer,when the poor Xeon E3 had to put up with processing 100s of RAWs from a D600,and the Corsair H40 was not really being a fan of me!

So what I did to test performance,was to take the first 50 pictures of that group of pictures I processed,apply the same individual adjustments,but made sure DxO Prime was active on all the selected pictures.

I selected OpenCL and 2 concurrent encodes at any one time(you can do more).

Firstly,lets see how the Xeon E3 fares.



Time to get the egg timer out,get a good book and put your feet up,yeah you have a wait.

Power consumption was between 118W to 150W. The average clockspeed was around 3.5GHZ or thereabouts. The system was unusable for anything else.

Now,lets apply Ryzen speed!!



Wowsers,its nearly double the speed!!

Clockspeed was around 3.55GHZ and the system was responsive.

It does it nearly half the time- I just might be able to read the prologue. Ryzen,stop affecting my book completion targets!!

Its not all jam tarts and doughnuts though. Power consumption was between 135W to 184W,and the poor little Wraith Stealth was between 70C to 90C with an average of nearly 80C. The max temperature of Ryzen is 95C BTW.

The cooler however was not that noisy,and generally even at full pelt unobtrusive. The only thing is AMD has obviously engineered acoustics over cooling performance. Some reviews have hinted it is a good low profile cooler,and even though for most of the benchmarks it kept the CPU under 70C,I think for any sustained CPU loads,the larger Wraith Spire cooler is a better choice or decentish aftermarket one.

From what I gather once Ryzen moves above 60C~70C it does start to affect the maximum boost - the CPU won't throttle,but just won't boost as much.

Personally,I think AMD should have just stuck with the Wraith Spire with the Ryzen 5 2600 just like they did with the Ryzen 5 1600.

Video world

The lost results have been found!! I decided to try a quick video encoding test,and settled on the x265 HD Benchmark:

http://x265.ru/en/x265-hd-benchmark/

HEVC is an increasingly used codec for video encoding of files due to its effiency but it does need proper hardware support to decode.

The test does 4 runs in total.



The Xeon E3 averaged 98.82 seconds and the Ryzen 5 51.66 seconds.

 

[CAT-THE-FIFTH]

Game On!!

Marking time!

Now time to start testing the gaming mettle of Ryzen.....with another synthetic test I cannot play. Step forward,3DMark!!

Lets see how the old Xeon E3 does.

Firstly in Firestrike Extreme.

https://www.3dmark.com/3dm/29013375?

Now Ryzen in its High Performance,Windows Balanced and Ryzen Balanced power plans:

https://www.3dmark.com/3dm/29139349?
https://www.3dmark.com/3dm/29139936?
https://www.3dmark.com/3dm/29140234?

Now lets compare the results.




Its interesting how each of the Windows power plans slightly shifts the bias of the score,but either way overall the scores does seem to improve,although there is a slight decrease in the GPU only score(!).

Lets see how the temporal agent does in Timespy,which looks at DX12 performance.

The score for the Xeon E3 1230 V2:

https://www.3dmark.com/3dm/29012999?

Now the scores for the Ryzen 5 with three different Windows power plans(High Performance,Windows Balanced and Ryzen Balanced):

https://www.3dmark.com/3dm/29135209?
https://www.3dmark.com/3dm/29138570?
https://www.3dmark.com/3dm/29136471?




Again we see the differences in performance bias with each plan,but again the Xeon E3 1230 V2 yields a better graphics score even if the overall score is worse. Weird,as you will see this is not true of the game I tested!

Futures made of Virtual....Reality now??

Time to get those VR goggles on,lets wheel out VRMark now!!

The benchmark consists of three tests - the Orange Room,the Cyan Room and the Blue room,each progressively getting more and more taxing on the graphics. I decided to test VR,since it is very important to have enough stable CPU performance,to keep framerates stable.

Now lets look at the good old Xeon E3 1230 V2:

https://www.3dmark.com/vrm/29013452?
https://www.3dmark.com/vrm/29013501?
https://www.3dmark.com/vrm/29013558?

Lets see how it compares to the Ryzen 5 2600:

https://www.3dmark.com/vrm/29185929?
https://www.3dmark.com/vrm/29186057?
https://www.3dmark.com/vrm/29186258?

Some weird results again.




Both systems could not pass the Blue Room test as it requires a faster GPU than a GTX1080(!) to pass.

In the Orange Room which is more CPU limited we see the Ryzen 5 2600 fly past the Xeon E3 1230 V2 by 34% but in the more GPU taxing tests the scores are ever so slightly worse for Ryzen. The tests were run in Windows Balanced plans for both systems.

I then decided to run Blue Room with the Ryzen system on the Windows High Performance power plan:

https://www.3dmark.com/vrm/29186428?

The score went up and the power consumption went up from 254W to 262W.

Then I looked at the CPU clockspeeds between the two runs.

The first is High Performance and the second is Windows Balanced.





Even though GPU utilisation was the same,it seems that High Performance had much higher CPU clockspeeds.

 

[CAT-THE-FIFTH]

Time for the Fallout!

Ok,now for some realworld games testing. TBH,most of my games run well enough on the Xeon E3 1230 V2,outside two specific instances - Planetside 2 and Fallout 4. Planetside 2 being online is hard to bench,but when I reinstall it I will probably report back,but Fallout 4 being single player is easier to benchmark.

The game in its normal form is not too hard to run,but becomes a system hog if you do a few things:
1.)Build settlements
2.)Build settlements
3.)Build settlements
4.)Increase NPC numbers.....in settlements
5.)Add 4K textures for everything

Guess,what I did?? TBH,in my last playthrough the old Xeon E3 was OK - I did have largish settlements,with a fair number of NPCs,and the game was kind of playable. However,something happened this year which changed that - Spectre and Meltdown,or more the mitigations for them. These affected disk I/O and whereas many games were not affected,one which was,is the open world RPG,The Witcher 3 which showed a 10% drop in framerates on a newish Intel CPU. Openworld games tend to stream assets off the install drive,so its not really rocket science to see why this might have happened. Fallout 4 in its vanilla form is somewhat I/O bound and can have stutters even in this case.

Did it affect Fallout 4 - yep. With the largish settlements and improved textures of in my last playthrough(not 4K ones for the most part),areas suddenly become much more jerky,as even before disk I/O is taxed in the game,and I remember trying the modded game off an HDD and it was unplayable,and it started having stuttering in a similar way.

In the end I decided it was best to avoid the downtown areas of the map,and since I was not getting the best FPS anyway(and the GPU was underutilised),I decided to use the spare GPU power for MOAR graphics. So I increased the texture quality to 4K custom ones,and used an ENB to give the game a cinematic look. As long as I avoided the city for the most part the game was playable.

However,this comes at a cost - draw calls can get upto 20000 in many parts of the game and in my settlements,and the downtown areas(which I avoided due to the issues) are now at 40000(!!). So the game is a system wrecker.

This is also compounded by the fact that Bethesda has not released any patches for any newer CPUs like Ryzen or even Skylake X,and the game likes lower memory latency access. Hence,Intel consumer CPUs are still the best option for the game,providing the mitigations don't affect I/O. Fallout 4 is one of the worst games for Ryzen in terms of performance.

Now to test both CPUs,I tested in two areas - one in my custom settlement in Abernathy Farm and the second in Diamond City. I also decided to use a multiple follower mod,and to push a higher NPC load at each location,and recruited nearly 20 followers. Hence there is at least 30 to 40 NPCs at each of the locations.

Draw calls in each location are upto 20000. In a vanilla game,or one lightly modded with more performance friendly textures,you shouldn't in theory see more that number even with a decent sized settlement at least IMHO.

This current save file uses around 200 mods,but here are the ones of importance(4K textures used where possible):
Vivid Fallout - All in One
Stromberg Retexture Project
Antique Aura Texture Overhaul
Water Enhanced
True Grass
Moon Retexture - 2K (by Sakura9
NMC's Sanctuary Bridge
OpTicTrukZ's Plants and Herbs Revamped
Prydwen 2K Textures
Better Settlements and Camps 2
Haul'd Out 4K
4K Castle Retexture
Enhanced Flickering Firelight v2
HQ Bathroom Stalls
OJO BUENO MILK VENDING MACHINE - 4K 2K 1K
Better Nuka-Cola Machine
OJO BUENO PORT-A-DINER MACHINE - 4K 2K 1K
OJO BUENO TEXTURE PACK for Fallout 4
Radio Re-Done 4K
Interiors Enhanced - Darker Ambient Light and Fog
Enhanced Lights and FX
Duffle Bag Retexure
Vehicle Overhaul
DECENT ENB - Natural and Performance Friendly
FogOut
True Storms - Wasteland Edition (Thunder-Rain-Weather Redone)
Protectron HD - 2K and 4K


Some other mods which are of revelance:
More Realistic Cats
FO4 War Tags
Crows And Creatures
Sim Settlements
Sim Settlements - Industrial Revolution
Sim Settlements - Rise of the Commonwealth
Multiple Followers Overhaul - Tee Party Companions
Unique NPCs - An Overhaul of the Commonwealth
Diamond City Expansion

The settings used in-game:




The game was capped at 60FPS as the Creation Engine links physics with framerates,so anything much over 70FPS leads to big issues. FRAPs was used to make framerate and frametime measurements,so there will be an additional CPU load due to this.

A farming we go!

The first benchmark was in Abernathy Farm,which has the highest build height of any area in the game - in a previous playthrough I managed to construct a 23 story tower. The scene used was not as extensive,but would count mostly as a medium to large size settlement. Including settlers and followers,there were 30 to 40 NPCs in the area. Including settlers and followers,there were 30 to 40 NPCs in the area. There are also transitions between interior and exterior areas of the settlement.

A roughly 50 to 60 second time run was used,however due to the way NPCs spawn in the game,there is a degree of randomness on the positions and movements of a number of them.Five to six runs were made for each test.





The Xeon E3 1230 V2 at the end produced an average of 43.84FPS and the Ryzen 5 2600 an average of 49.2FPS,ie,it was 12.47% faster overall. However,what is quite evidence that in some areas the percentage is much higher and the Xeon E3 1230 V2 dipped significanly to nearly 30FPS for a lot of the runs. Remember this is at 2560X1440 and not the 720p resolution many reviews use to "show" a CPU bottleneck. This is a proper CPU bottleneck!!

However,what was not quite evident is that the Ryzen 5 2600 was significantly smoother overall,as any sudden movements would lead to stuttering on the Xeon E3,which was from assets being loaded.

This is evident if you look at the frametime plots.

First if we look at the Xeon E3 1230 V2,you can see whether looking at individual runs(top) or even an average(bottom) there was frequent latency spikes.





Now moving over to the Ryzen 5 2600,there was some occasional stutter,but for the most part stuttering was far less noticeable when running the game,and it felt much more smoother,and this is reflected in the individual runs(top) or an average(bottom)





Time to move to the big city!

Now the next test scenario,is a one minute test run through Diamond City. This area is generally known to be quite CPU intensive due to its many NPCs.



Again we can see the Ryzen 5 leading the Xeon E3 and staying above 40FPS for far longer. The Xeon E3 1230 V2 produced an average of 43.89FPS and the Ryzen 5 2600 an average of 50.42FPS and was 15.27% faster overall.



However,in parts it could be upto 30% faster!

Also as with the Abernathy Farm test run,the framerate was far more stable and less susceptible to sudden spikes if turning quickly,etc.

This is bourne out if you look at the frametimes for the Xeon E3 both individually(top) and as an average(bottom).





There was a lot of spikes even in Diamond City. However,if you look at the Ryzen 5 2600,the spikes were generally not as severe either individually(top) or as an average(bottom).






Windows power plan testing

I tested each area under three different Windows Power Plans - Windows Balanced,High Performance and Ryzen Balanced.

Lets start with Abernathy Farm.



So the results were 49.35FPS,48.66FPS and 48.95FPS respectively - the Windows Balanced plan was the best plan in this scenario.

Now lets move over to Diamond City.



The results were 50.77FPS,50.18FPS and 50.31FPS - the Windows Balanced plan was also the best in this scenario.


12/10/2018 Update!!

Testing with the Wraith Spire:

https://forums.overclockers.co.uk/posts/32193644/

13/10/2018 Update!!

RAM testing with improved timings:

https://forums.overclockers.co.uk/posts/32194611/

It's the Final Countdown??

Perhaps for the time-being,once I CBA to tweak the RAM some more!!




The biggest improvements have been in the minimums which has been very noticeable!

Summary

Am I happy with the upgrade so far - yes I am. Especially for the non-gaming workloads I do it seems a very big upgrade overall. For gaming,even the most unoptimised Ryzen unfriendly game like Fallout 4 sees a decent jump in playability,so for more modern titles which are better optimised for Ryzen and scale better with more threads I suspect the gap will increase over the next few years.

OFC,if you can wait,something like 7NM Ryzen might be even a bigger jump,but for someone on Ivy Bridge even a Ryzen 5 2600 can be a useful upgrade IMHO.

Future to do list

1.)Add in X265 comparison if I can find the other figures
2.)Once I get Wraith Spire cooler,retest to see if it makes any difference
3.)Tweak RAM settings and see if there is any difference

 

[CAT-THE-FIFTH]

Good comprehensive review. You should definitely tweak the ram timings, I got a good boost doing do.

I will give that a try,but this was just testing out of the box performance.

Good review. What a cute little motherboard that is.

Will you be overclocking it at any point?

I might try the enhanced core boost part once I get a better cooler,but not before the cooler arrives!!

Really interesting read, really want to go am4 or if I can wait am5 in the future.

I was upgrading from an Ivy Bridge CPU - your system will easily last until the 7NM Ryzen CPUs are out. Having said that the secondhand value of the Skylake CPUs is still decent though,although for Fallout 4 your CPU is pretty good!

 

[CAT-THE-FIFTH]

You have the same ram part number as mine. I dialed in the fast timings but I had to leave geardown enabled for stability, just a heads up.

Any chance you can tell me the settings you used please??

 

[CAT-THE-FIFTH]

I use 1.37v for ram.
1.05 soc voltage.
geardown enabled.

Termination block was left on auto apart from RTT_WR that was set to off.
CAD_BUS block was left on auto.

Thanks! When I eventually get the better cooler,I will give those settings a try!!

 

[CAT-THE-FIFTH]

Thats the newest version of the Ryzen Mem Calc right? from what i understand you just run it, Press R-XMP, copy down the settings and slap them into your BIOS right? i might update the BIOS on my CH6 and see if i can push the ram to faster timings as its running the 3200mhz default profile at the moment and i imagine the timings are pretty slack. I have 3400mhz CL16 Ripjasws.

I will try and get the latency down to C14 - wish I had gotten the 3600MHZ set for a few quid more now!

Impressive thread @CAT-THE-FIFTH very nice and entertaining work.

Thanks I try to make my reviews entertaining - you should check out my GTX1080 review!

Great review mate

Thanks.

Looking forward to reading this later. Thanks.

I will be adding some more stuff to the review as time progresses.

A couple of times you mention that power consumption on the ryzen is higher than the previous system but the ryzen is completing much faster.

Does this mean that overall due to the faster time to idle the cpu is drawing less power. Sure it uses more during a given task but if it goes idle in half the time the previous cpu takes then it still comes out on top right?

I quoted peak figures due to it being a mini-ITX as a worse case scenario,but higher instantaneous power does not mean more task energy overall.

 

[CAT-THE-FIFTH]

I have added an X265 test,Fallout 4 testing under different power plans,and pictures of the new Wraith Spire cooler I installed.

Now I am doing some tests with the Wraith Spire instead of the Wraith Stealth.

 

[CAT-THE-FIFTH]

wow, excellent review cat.

Great work, nice to see this on here, make's a change.

Thanks,I will be adding a few benchmarks here and there,with the new cooler and hopefully with the improved RAM timings.

 

[CAT-THE-FIFTH]

Enjoyed reading that. Very comprehensive.

Thanks!

 

[CAT-THE-FIFTH]

Some IN-Spiring testing

So I decided to do some testing with the Wraith Spire plonked on.

The first test was with the same 50 RAW DxO jpeg conversion test.



A whole 8 seconds faster,eh! That gives slightly more time to open the first page!

The core boost didn't seem to change that much - maybe 50MHZ to 100MHZ at most.

However,something did change - the temperature!!

Instead of around a 90C peak,the CPU peaked at just over 70C!!



So was the cooler any noisier - not exactly but there was a bit more noticeable bearing noise IMHO,although YMMV.

Now onto everyone's favourite gamers benchmark....Cinebench.



A 5 point increase in the single threaded score and a 3 point decrease in the multi-threaded score.The OpenGL score slightly increased too.

Now see if there is any "Fallout" from further testing?? Hehe!!

Lets move onto Fallout 4.

The first sequence is Abernathy Farm.



The FPS with the Wraith Stealth was actually slightly higher at 49.2FPS as opposed to the 48.77FPS with the Wraith Spire,but the latter had better minimums.

Now,lets look at frametimes.




Its six of one,and half a dozen of another TBH,but temperatures hovered at worst around the 60C mark,but a slight win for the Wraith Spire though.

Now,lets have a look at the Big City....Diamond City.



Framerates definitely seemed better with the Wraith Spire with regards to minimums but surprisingly enough,but the average was the same in both cases at 50.77FPS!

Lets,look at the frametimes now.




Frametimes,tad worse?? Hard to say overall,but I did have a Windows update before testing the game(still Win10 PRO 1803 though),so that might not be helping. However,Fallout 4 is finicky to test especially due to its sensitivity to storage throughput when modded.

Is the Wraith Spire worth it?? Yep,I think for the better cooling performance it is,and the CPU did seem to hit a slightly higher max clockspeed of around 3.9GHZ instead of the 3.7GHZ~3.8GHZ it did before. However,the Wraith Stealth will suffice but under heavy load you will reach near enough the max temperature the CPU can handle.

 

[CAT-THE-FIFTH]

When are you going to do the ram timings, they look very slack.

I am doing it stages - so during the weekend I will muck around with the timings.

 

[CAT-THE-FIFTH]

I would have probably have sorted the timings out first as they can give big gains across a number of applications, then again you can always re-bench everything and compare.

That is the plan,I wanted to see how much of an effect the improved cooling would have,and pushing the memory controller more will probably mean more cooling is required anyway. I am actually surprised that even though the Wraith Spire was better,performance was still very close with the Wraith Stealth although the temperatures were much worse!

 

[CAT-THE-FIFTH]

Another day.....another Windows update!

It looks like the minions have been hard at work at Microsoft,so another set of Windows updates to get in the way!! Luckily I am still on Win10 PRO 1803.

Now,I decided to try and tweak the RAM. There is a useful utility which helps Ryzen CPU owners with this:

https://www.techpowerup.com/download/ryzen-dram-calculator/

So I selected the SAFE preset to get things started. Now the system booted fine after applying the new RAM timings in the BIOS,but after trying to do a few Fallout 4 test runs the game would rage quit in protest!! So I tried some more tweaking and more of the same - I suspect with some additional tweaking it would have worked and it was some of the more esoteric RAM settings which are the issue here.

However,Gibbo suggested that since the RAM was B-die that 14-14-14-31 settings should work,so I decided to try them out and left everything else on auto,ie,the original D.O.C.P. settings,and so far the RAM seems stable.

Now lets look at Cinebench R15. C14 scores are on the left and C16 scores are on the right.



Useful improvements in the OpenGL and multi-threaded score but a regression in the single threaded score,which seemed repeatable.

Now moving onto DxO,I tried the same test sequence as before.



A further 8 seconds knocked off the total time - I better start reading books quicker!! Time to go to the speed reading classes.

Peak power consumption was at 176W. The max temperature peaked at 74C although it might be slightly warmer today.

Not Fallout 76

Now moving onto some games testing with more Fallout 4.

First lets trouble more settlers at Abernathy Farm!



There is definitely an improvement to the minimums. The average FPS increased from 48.77FPS to 49.44FPS,which is not too bad for a few quick tweaks in the BIOS.

Now lets look at some frametimes.




Possibly slightly better with C14 timings,but hard to say IMHO although YMMV.

Now lets make a trip back to Diamond City.



Again minimums seem better. The average FPS went up from 50.77FPS to 51.51FPS,which is not too bad of an improvement.

Now lets look at the frametimes(again).




I would say with the C14 timings it does seem slightly better IMHO although YMMV.

 

[CAT-THE-FIFTH]

Sub timings are where the real uplift is. I'm surprised you never got the safe timings stable, you need to bump up the ram volts up to 1.4 and the soc to 1.1.

TBH,I will get around it at some point,but I went middle of the road,and everything else apart from Fallout 4 seemed OK. The issue is the game seems to be very RAM dependent too,so any slight instability and the game will have issues,so its going to a real PITA,especially since there is a metric ton of settings to tweak. At the very least I know the fallback settings.

Regarding RAM,I really don't like pushing past the recommended voltage settings for the RAM,as its a mini-ITX system,airflow will be an issue,and the extra heat needs to go somewhere. I will try to see what I can achieve on 1.35V and move on from there.

Also from what I gather the Ryzen 5 2600 is a bit worse off with RAM than a Ryzen 5 2600X - Techspot saw it with their examples.

 

[CAT-THE-FIFTH]

Those are the settings suggested for Ryzen+ by the DRAM Calculator - anything I need to be wary of??

 

[CAT-THE-FIFTH]

I will give it a try today and if needs too much faffing around I will go back to the fallback settings,as tweaking,then benchmarking takes time which I am not doing other stuff instead!