AMD Ryzen For Simracing?

[Pedro Azevedo88]

Depends on the CPU. For this one it's normal, specially if you take into account that they overclock automatically taking into account their temperature, which increases clock boost, power draw and eventually temperature.
You have Intel CPUs that under load have higher temperatures. If you don't pass the safe temperature it's ok.
Of course that with an higher ambient temperature you'll have an increase of your components temperature.

 

[Alex Townsend]

Question: Isnt a CPU temp mid 70s a bit high?
I know its a (much) older generation but my Xeon x5670 does run at absolute max 59-61 when gaming for 1-2H.

Im no expert but I think the chip designs from both companies are different meaning they would have different thermal limits?
I'm also starting off with ambient temp of 29°C... not much I can do about it unless I have the air con on the whole time which would seriously increase my electricity bill...
These AMD 3xxx chips have TDP of 95°C so 75ish is way below and that is whilst benchmarking.
I read on Guru3d that they stress tested a 3800x using multiple 1024m Prime tests using the stock cooler and it never went over 73°C
I bet they had a lower ambient temp than me though....
When doing Cinebench r20 I'm seeing strange results though. I'm running the tests again in case I forgot to disable a program from running but I appeared to get better scores using the Safe memory settings than the Fast ones...
Using PBO is definitely slower with these RAM timings too.
Will report back with my findings.

 

[Durge Driven]

My second B450 and 3200G got back from RMA with new CPU ..I dunno

So put together on table setup, new SSD ( so peeps can test parts ) selling it

Do CPU-Z and is like 1300 MT should be 1700+ stock lol

After mucking around I went in and did bench on my 3200G and comes back the same even O/C ed lol

Went into tasks and the latest Radeon driver has some Radeon host tool ? using 33% of system lol

End task and benches are okay

 

[Durge Driven]

Well I could not fix it all threads talk about earlier versions with different menu there is no turn off ReLive
Installed Radeon Dec 12th and everything back to normal

P.S. New chipset out

 

[Durge Driven]

Fix is weird but works

Just create a folder called "Radeon ReLive" in your video folder

 

[Trebormoore84]

Im no expert but I think the chip designs from both companies are different meaning they would have different thermal limits?
I'm also starting off with ambient temp of 29°C... not much I can do about it unless I have the air con on the whole time which would seriously increase my electricity bill...
These AMD 3xxx chips have TDP of 95°C so 75ish is way below and that is whilst benchmarking.
I read on Guru3d that they stress tested a 3800x using multiple 1024m Prime tests using the stock cooler and it never went over 73°C
I bet they had a lower ambient temp than me though....
When doing Cinebench r20 I'm seeing strange results though. I'm running the tests again in case I forgot to disable a program from running but I appeared to get better scores using the Safe memory settings than the Fast ones...
Using PBO is definitely slower with these RAM timings too.
Will report back with my findings

That's odd, which memory are you using?

Don't know if this has already been discussed on this thread but have you tried out Manual per CCX overclocking? This was the best way of getting performance boost for me over PBO, it also means you find out which CCX performs the best for games. I also managed to get a score over 10,000 in R20 with this method with my 3950x. Which is mental...

On how to do it check out this video.

I've found about 50mhz difference on the same chiplet, and around 200mhz between separate chiplets (for 3900x and 3950x users). So if you are using a 3900x you might get a nice surprise in performance when you start to test depending on if you got lucky or not.
If you do this you will however need to start setting CPU affinity to the best performing cores for games, as well as other apps to the others, The new Windows 10 scheduler sort of works but its not foolproof yet.

If you are hunting for max FPS (like me making the mistake of buying oversized high FPS monitors), making sure the cores that are running say RF2 has nothing else running in it to max out headroom for the GPU, especially making sure Teamspeak and Discord are set to other cores. Easier said than done, in the end I got Process Lasso to remember which cores to use for each app. https://bitsum.com/

I am rocking a 3950x which is overkill for gaming I know, but the bin is awesome and have one of the CCX's manually set to 4.5ghz (it got 4.6ghz R20 stable!) which along with setting CPU affinity getting amazing results. RF2 on triple 1080p ultrawides at a range of 120-160hz at high settings with a 2080ti. Absolute dream considering this sim is a pig in terms of performance on triples especially on the startline and early laps.

 

[Alex Townsend]

That's odd, which memory are you using?

Don't know if this has already been discussed on this thread but have you tried out Manual per CCX overclocking? This was the best way of getting performance boost for me over PBO, it also means you find out which CCX performs the best for games. I also managed to get a score over 10,000 in R20 with this method with my 3950x. Which is mental...

On how to do it check out this video.

I've found about 50mhz difference on the same chiplet, and around 200mhz between separate chiplets (for 3900x and 3950x users). So if you are using a 3900x you might get a nice surprise in performance when you start to test depending on if you got lucky or not.
If you do this you will however need to start setting CPU affinity to the best performing cores for games, as well as other apps to the others, The new Windows 10 scheduler sort of works but its not foolproof yet.

If you are hunting for max FPS (like me making the mistake of buying oversized high FPS monitors), making sure the cores that are running say RF2 has nothing else running in it to max out headroom for the GPU, especially making sure Teamspeak and Discord are set to other cores. Easier said than done, in the end I got Process Lasso to remember which cores to use for each app. https://bitsum.com/

I am rocking a 3950x which is overkill for gaming I know, but the bin is awesome and have one of the CCX's manually set to 4.5ghz (it got 4.6ghz R20 stable!) which along with setting CPU affinity getting amazing results. RF2 on triple 1080p ultrawides at a range of 120-160hz at high settings with a 2080ti. Absolute dream considering this sim is a pig in terms of performance on triples especially on the startline and early laps.

Thanks for the video, I'll take a look at lunchtime.
I'm doing quite extensive testing now.
Optimised defaults
Profile 2 at 3000Mhz
Profile 1 at 3200Mhz
Safe memory at 3200Mhz
Fast memory at 3200Mhz

My board is a B450 Gaming pro carbon and I think it can only support up to 3200Mhz, not sure on that though
Memory is Kingston Predator 16GB KHX3200C16D4/8GX which is Hynix CJR

 

[Trebormoore84]

My board is a B450 Gaming pro carbon and I think it can only support up to 3200Mhz, not sure on that though
Memory is Kingston Predator 16GB KHX3200C16D4/8GX which is Hynix CJR

Sorry that vid might not help, but it is a way of testing what max clocks you can do on a chip. You'll want to get your memory sorted first though before overclocking the CPU, so i'd keep the CPU on auto while messing with timings first.

Assume you are running HWinfo while doing Cinebench runs. Seeing how the clocks, voltages and temps are performing while testing might uncover whats going on like throttling or something overheating. Also compare single core vs multi core performance. If you get better performance single core running 3200 than 3000 maybe something else is going on for example.

Going back to basics however have you tried the RAMs XMP profile instead of custom timings?

If you haven't got XMP to work, one thing I did on my work system (a Threadripper 2nd gen system) which uses similar Hynix memory is just run XMP profile timing settings with 1.225v on the VDDRSOC, this can only be set in Ryzen Master on some MB bios however. Max is 1.25v, maybe try 1.2v first.

MSI B450 Gaming Pro Carbon supports upto 3466mhz OC on the memory so the memory clock shouldn't be an issue here.

Also B450 boards usually have weak VRMs by comparison to X470, but your one actually looks decent with a 9 phase so its probably not that. Was thinking maybe its bottlenecking at the higher frequency and throttling back during the all core test.

If you do decide to get better 3200 memory get G Skill Trident Neo with C14 timings if you can which performs just as well as 3600 C16, the ones with Neos have Ryzen specific XMP profiles. You just put it in set the XMP and it works perfectly. Messing with timings and voltages is a massive PITA just to get a handful of FPS.

G skill is hard to find the right model sometimes, and its expensive but it just works and you get what you pay for IMO.

F4-3200C14D-16GTZN-G.SKILL International Enterprise Co., Ltd.

G.SKILL

www.gskill.com

https://www.alternate.co.uk/G-Skill/Trident-Z-F4-3200C14D-16GTZN-memory-module-16-GB-DDR4-3200-MHz-RAM/html/product/1565360

Another good source for G Skill Trident Neo memory is http://www.lambda-tek.com/

 

[Alex Townsend]

True about expensive...
Double the price of the sticks I got and the only cl14 ram listed but unfortunately they are not in stock.
There are 3200 cl16 trident royal or 3200 cl16 neo

But will give XMP profiles a go first and come back with results shortly

 

[Alex Townsend]

Would you suggest trying XMP profile 2 at 3000Mhz first or diving straight into Profile 1 at 3200Mhz?
I believe I have had it on either of these settings previously

 

[Alex Townsend]

Cinebench R20 I actually dropped score vs optimised defaults using XMP profile 3200Mhz
Had Multi at 4868 and SC 510
When I went XMP 3200
Multi went down to 4857 and SC 505
Went as high as 85 degrees

Cinebench r15 is 2096 and 203 respectively
This was more sober at around 57 degrees.

Game benchmarks have gotten better. Albeit not amazingly different but different nonetheless.

Will try with the ram calculator and compare later on tomorrow. Getting late now...

 

[Trebormoore84]

Would you suggest trying XMP profile 2 at 3000Mhz first or diving straight into Profile 1 at 3200Mhz?
I believe I have had it on either of these settings previously

I'd focus on getting 3200mhz to work. You want the highest frequency you can for the infinity fabric clock to be as high as possible on the CPU. Set the XMP profile and plug in the specified DRAM voltage manually instead of leaving at auto if you have trouble posting. I'm also assuming you have installed the latest MB bios BTW.

This is your board right? https://www.msi.com/Motherboard/B450-GAMING-PRO-CARBON-AC/Specification

Hmm ok, i know that is the most popular ram kit for Ryzen 3000, Maybe another way to go is get RAM with 3466mhz XMP like this:

G.SKILL F4-3466C16D-16GTZR Trident Z RGB Series 16 GB (8 GB x 2) DDR4 3466 MHz PC4-27700 CL16 Dual Channel Memory Kit - Black with full length RGB LED light bar: Amazon.co.uk: Computers & Accessories

Buy

www.amazon.co.uk

your MB has a RAM limit of 3466mhz according to the specs, so it should work but you will likely need to set the DRAM voltage in the bios to 1.35v instead of Auto to get it to post with the XMP, as its not a Ryzen designed XMP profile.

Trident Royals are just Trident but with blingy covers. Wouldn't bother unless you've got a visual theme going on.

Saying that i'm not sure its worth spending the money on a sideways step, i.e. buying more ram with the same spec. Once you get that Kingston Ram kit working its problem solved TBH. The differences in performance are marginal like within 10fps until you start going to 3800mhz ram with C16 timings, right on the limit of the infinity fabric.

 

[Trebormoore84]

Cinebench R20 I actually dropped score vs optimised defaults using XMP profile 3200Mhz
Had Multi at 4868 and SC 510
When I went XMP 3200
Multi went down to 4857 and SC 505
Went as high as 85 degrees

Cinebench r15 is 2096 and 203 respectively
This was more sober at around 57 degrees.

Game benchmarks have gotten better. Albeit not amazingly different but different nonetheless.

Will try with the ram calculator and compare later on tomorrow. Getting late now...

Fair enough. I'd keep the optimised defaults.

Hmmm, maybe check mosfet / VRAM temps on multicore testing.

Your R20 SC scores are very good, so gaming / simracing FPS should be absolutely fine. Multi is lower than average for that CPU but then how much rendering are you doing?

R15 doesn't stress the CPU as much as R20 as its too quick a test these days.

 

[Alex Townsend]

I'm going to give the fast and safe ram settings another go. I'm sure I had r20 at 513 on SC once. Unless it was a dream... lol will come back on that tomorrow with some game benchmarks too

 

[Terry Rock]

The clocked per CCX gains are nice...and very consistent.
My 3600x for instance, will do as high as 505 in single core test and 3710 in all cores test while split.
It is a great way to gain added performance at no extra cost.

 

[Durge Driven]

I'm also starting off with ambient temp of 29°C... not much I can do about it unless I have the air con on the whole time which would seriously increase my electricity bill...

Now add another 17C to base temp lol

Same here, what I save using 3200G for the majority of day will pay for itself

Yesterday was 41 inside my flat and concrete render over bricks heats up a little more every day ...hottest day 46C The only way to bear it is blacken every window and drink 2-4 litres water a hour


My 3200G laughs .... I drown in sweat while browsing, burning, downloading, encoding and watching that F2 documentary @ same time and lucky to hit 50C ( 3900X cooler )


My 3900X screams even using Kraken 280mm still basically unusable

 

[Alex Townsend]

Wow, so we're experiencing similar issues with cooling then.
I was in the man cave this morning and it was 31 degrees. Always hot around CNY here.
Under normal conditions I'm around 46 degrees with the Corsair H100i v2. Under CB it goes up to around 75. I need to measure temps whilst racing.
Got some interesting benchmarks coming once I've carried out a bit more testing.

 

[Durge Driven]

I relented and brought first air cond. cheap Arlec one room job 7,500BTU
Had 46C 43 and 44C 3 days running
Default it was 22C so I set to 28C and worked great

At the moment we have welcome rain and low 20's so Air con. has become a giant brick I got to push past to get to towers but glad I got it for next heatwave

Now it's 20C windows open
3200G@4.1GHz - 32C idle CPU-Z stress 52C ( 3900X cooler)
3900X@4.35 GHz 33C idle CPU-Z stress 64C ( Kraken X62 AIO )
I find neither tower gets any higher after 10 minutes
Typically sims are 10C cooler I think or around that

Problem with Kraken one fan above 60C is noisy .....have to check that one day lol

 

[Durge Driven]

Also thanks for help with 3200G - Neo 3600MHz C18-22-22
Won't run SPD or manual speed so I set 3200HMz C16 - 18 - 18 for now seems decent enough

How about my 3900X ram ( Neo as well) have not touched it runs 3600 C14 SPD
Try 4000 C16 -17-17 ? ...or ............ lol

I just hate the bios resets testing , enables on board sound and windows updates it before I remember
Guess I disconnect wifi first in future , have settings saved in bios at least

You suggest doing in bios or ryzen master

P.S.
Downloaded this... https://www.guru3d.com/files-details/download-ryzen-dram-calculator.html
Can't make heads or tails of it

 

[robt100]

Right, I'm hoping to pick a few minds on this one. Just built a new system (with old GPU).
R5 3600
DDR4 3600 Cl18 (vengance LPX)
B450 MSI Tomahawk MAX

Now, XMP works to get 3600CL18, but any changes from the specs given on the Ryzen DRAM calculator causes no-boot or stuck with the CPU debug light on. Got a feeling I need to update the BIOS (though MSI's live update says its on the latest, its actually on the original!) as 3 updates ago was a changelog for 'better memory compatability).

BUT...it seems I have lucked out on the Silicon Lottery, 4.35Ghz all core boost at 1.24v...and still testing. Max CPU temp is 56c so far (240mm AIO). I dont really want to loose the clockability if its down to the BIOS, but at the same time, is a better timed memory going to make more difference? (Cas latency maybe dropped to 17 or 16 from 18).

Your opinions please!