AMD Ryzen For Simracing?

[Durge Driven]

3900x bedding in nicely, low workload 25C
That is almost as good as my stock 8700

 

[Alex Townsend]

3900x bedding in nicely, low workload 25C
That is almost as good as my stock 8700

Lol
Our screens are practically the same, well I've only got a Ryzen 5 3600x, but Ryzen master and uTorrent downloading F1 from Hungary!

What cooler do you have?
I got my digital room thermometer through the other day and my ambient temp is 30 degrees...
I haven't been able to install my Corsair H100i v2 as it didn't come with the AM4 bracket so I have one on order which should arrive sometime this week.

Also what is that app on the right monitoring all your devices?

 

[Durge Driven]

Recommend you have a look at Kraken before Corsair as fan design is better
P.S. I you fitting better aftermarket fans it is irrelevant

I have run all corsair from H60 v1 to H110i

I have Kraken X62 ( 280mm ) overkill really
Kraken X52 would be enough

https://www.nzxt.com/products/kraken-x52

one of tweaktowns highest score
https://www.tweaktown.com/reviews/8106/nzxt-kraken-x52-liquid-cpu-cooler-review/index8.html

 

[Alex Townsend]

Sorry, I'm just waiting on the bracket for AM4, I already have the H100i, I was using it with my i7 3770k.
I would look at the kraken but that would mean selling the H100i and it's a few years old so not sure on the resale value...

One thing I'm not sure about is what is the average usage lifespan of an AIO Liquid cooler?
Like I said above, I've had mine a while now and just wondering how much life it has in it?

 

[Durge Driven]

I know some who have my old ones like H60 v1 and v2 still running after 5 years
I usually buy new AIO every CPU/mobo as saves you swapping anything

 

[Durge Driven]

Sounded dumb when I read that back

What i meant was I almost always sell case, cooling, cpu and mobo together...if it works don't play with it

P.S.
I sell my old Corsair 750D, H110I, 8700, B360 Aus Swift and 2400MHz together don't upset great temps and enable Intel GPU in bios .add old SSD , OEM Home key, fresh OS in 10 minutes and kick it out the door for $800AU
So only paid for 3900X ($789AU ) which is fine by me, have no use for 2 towers, parts owed me another $500 or $10 per week the last year

 

[Durge Driven]

Basic stress testing started but I not going to push to max
4.4GHz @ 1.35 volt 60c max in basic master stress test
That adds 500+ to cpu-z and 30+ single core

P.S. Probably do a 2nd profile for sims 4.6Ghz couple cores ..not sure it would be worth much ?
last 3 builds were Non K Intels , but don't want to ramp volts or temps crazy all for few more fps either
I testing voltage backwards high to low 1.325v next
Once they fix bios voltages / readings I will run stock bios but this is lower V and C @ full load for moment ............maybe run 2nd lot of Cores 100MHz lower ?

 

[Durge Driven]

Mustn't be enough 3900X validations
Beat a stock 9900K single core by 1 mark <<<< >>>>

 

[Dan Costa]

How did it go?

I changed all at once (new WB, delid) so I dont know how much each contributed to the end result. Temps dropped a lot, which allowed to pump +100mv. Unfortunately it was only enough for +100mhz. So, not really worth it. A dog CPU will always be a dog CPU. But for simracing, I was able disable 2 cores + HT, and reach 5.2ghz stable with the remaining active 6 cores.
I could not reach 5.3

 

[Terry Rock]

Mustn't be enough 3900X validations
Beat a stock 9900K single core by 1 mark <<<< >>>>

I take it you're one happy 'camper' again?
They do run nice once setup...don't they.
I don't now...and never really cared about all-out single thread total numbers.
I'm much more interested in the deviation between minimum to maximum framerate.
I actually ran a few test last week to verify how many cores come to the 'party' in a few of the sims.
For a long time we were being told most sims use a single core.
That's not what I saw in real-world testing.
Unfortunately I do not have a tripod to capture the in game core use.
(you have to use an external recording source and start the game with it's own executable, so as not to influence core count).

 

[RobertR1]

I changed all at once (new WB, delid) so I dont know how much each contributed to the end result. Temps dropped a lot, which allowed to pump +100mv. Unfortunately it was only enough for +100mhz. So, not really worth it. A dog CPU will always be a dog CPU. But for simracing, I was able disable 2 cores + HT, and reach 5.2ghz stable with the remaining active 6 cores.
I could not reach 5.3

What vcore and llc you're running for that?

Get on your RAM tuning and start tightening up your secondaries and tertiaries. I'm getting a massive bump in low fps. Easily worth more than than a couple of hundred mhz.

 

[Dan Costa]

What vcore and llc you're running for that?

Get on your RAM tuning and start tightening up your secondaries and tertiaries. I'm getting a massive bump in low fps. Easily worth more than than a couple of hundred mhz.

1.390v Turbo LLC (doesn't vdroop) for 6 cores @ 5.2ghz. Yes also got the CL14 B-die ram last week. Didn't had time though. The old wc loop was a Frankenstein from wc parts I bought between 2007 and 2014. Spent time on that instead. I will make a lot of benches comparing enabled cores and ram speeds and show you guys the results

 

[RobertR1]

Turbo LLC means a Gigabyte board? Which board?

With 2 DIMM, my Z390 hits a wall at 3600. Then I can only tune timings.

 

[Dan Costa]

yes, the aorus master. T-topology so I got 4x8gb of B-die. How low did you go? I'd rather have 3600 CL14 than 4000 cl17. My kit is "guaranteed" to reach 4533mhz though. Team Group Dark Pro 8pack cl14. Hopefully the imc allows for that.

 

[RobertR1]

You'll kick ass with a 4x8 kit.

From my pain, my biggest tip; on these boards, set everything manually in the bios and you'll train better even if the auto recommendation is the same as what you're gonna put in. Learned that lesson when failed training so many times.

I got the 8pack roids edition kit also. Suppose to be a 4000mhz but no chance on my board thus I went down the path of hand tuning.

My timings. Only thing diff is command rate of 2. 1 doesn't like training after reboots sometimes. Annoying. DRAM = 1.5v Training =1.51v VTT (termination) = half of DRAM voltage. Vccio = 1.2 Vccsa = 1.23

https://cdn.discordapp.com/attachments/328891236918493184/607463439262285854/2019-08-03.png

No chance I can get 14 stable. No RTL offset in bios so can't tune those but they should auto set properly but it's a GB bios so who the f knows...

 

[Dan Costa]

cheers! thanks for the timings. Can't you get higher clocks by lowering voltage? I have no experience with B-dies but with hynix temperatures start to degrade performance really soon to the point where I got the best performance with almost stock voltages (1.28v). Bumping it to 1.3v would fail hci memtest and with 1.5v it wouldn't boot at all.

 

[RobertR1]

B-Die scales with voltage so treat it like a cpu. Just start off with 1.5v, get your max stable and see if you have headroom to work down. 1.5v is fine for 24x7 on b-die. Just make sure you got air flow over the heatsink. Annoying we don't get temp sensors on our kits which would be nice to have.

If you're benching for small periods, you can even run with 1.9/2.0v without long term degradation. Just put a fan on it.

B-Die can be temp sensitive esp if your trefi is at max (65000) but otherwise, voltage is your friend.

 

[Case_]

For a long time we were being told most sims use a single core.
That's not what I saw in real-world testing.

I wish people stopped interchanging the terms "threads" and "cores" as if they're the same thing. It's not the same at all, even though it is related.

Yes, most sims generally use between one and three threads. No, not even a single-threaded application is likely to be exclusively using a single core. (It will, in certain situations, but it depends on many other factors.)

It is virtually useless to monitor CPU usage on individual cores, much less draw any conclusions from it. It won't tell you anything about the app you're running. You have to look at the actual thread CPU usage for the given app to be able to see what's actually going on.

 

[RasmusP]

I wish people stopped interchanging the terms "threads" and "cores" as if they're the same thing. It's not the same at all, even though it is related.

Yes, most sims generally use between one and three threads. No, not even a single-threaded application is likely to be exclusively using a single core. (It will, in certain situations, but it depends on many other factors.)

It is virtually useless to monitor CPU usage on individual cores, much less draw any conclusions from it. It won't tell you anything about the app you're running. You have to look at the actual thread CPU usage for the given app to be able to see what's actually going on.

And even then you can't really know what's happening.
Rkxket league pins one thread at 11.5% for me but if I unleash the fps on low resolutions it will keep the 11.5% from 50 to 500 fps.
At least we know:
When the gpu is at 100% it's the limiting factor and if not it's either an fps limit, the engine, a bug or the cpu

 

[Durge Driven]

I take it you're one happy 'camper' again?
They do run nice once setup...don't they..

twas a joke ...1 point higher then Intel.............I don't care about bench tests or gaining another 0.5%
Close enough is good enough I certainly won't pay double for ram that don't offer real world value

My board and ram cost 1/2 the CPU and it runs off it's head

X570 Master is triple the price of my mobo and double the price for top ram what does that get me
If it got me 50% more performance I would have it but not for small increase