AMD Ryzen For Simracing?

[RasmusP]

To clarify on this and what I've just learnt a few days ago :
With Intel you only get the faster ram speed but with amd ryzen the ram clock controls some internal cpu clocks too.
So the higher the ram frequency, the higher the frequencies of some parts of the ryzen CPUs become!
And that seems to be the reason why amd cpus can become unstable with too high ram clocks as they overclock some cpu parts and not only the memory controller!

 

[Case_]

And while we're on the subject of clarification - after spending most of the night researching overclocking the Ryzens and testing things myself, there seems to definitely be a bit of a difference between the non-X and X Zen+ chips when it comes to overclocking. It mostly seems to boil down to the fact that when left to their own, the X chips under lightly threaded load can boost single cores to a higher level than what can usually be achieved by doing a manual all-core overclock, which is the reason why it might be counterproductive to manually overclock the X chip - you'll get all cores running slightly faster (so you might get a slightly better performance under heavily threaded load, but you'll lose that higher boost in lightly threaded scenarios.

Meanwhile the non-X chips do not boost as high and you are usually able to achieve higher all-core manual overclocks compared to what the XFR can do on its own. So you're not losing anything by doing that, you're only gaining, unlike on the X chips.

Now whether that difference on the non-X chips can actually make a difference outside of synthetic benchmarks is yet to be seen. But it can definitely be seen in benchmarks, that I can confirm myself.

 

[RasmusP]

And while we're on the subject of clarification - after spending most of the night researching overclocking the Ryzens and testing things myself, there seems to definitely be a bit of a difference between the non-X and X Zen+ chips when it comes to overclocking. It mostly seems to boil down to the fact that when left to their own, the X chips under lightly threaded load can boost single cores to a higher level than what can usually be achieved by doing a manual all-core overclock, which is the reason why it might be counterproductive to manually overclock the X chip - you'll get all cores running slightly faster (so you might get a slightly better performance under heavily threaded load, but you'll lose that higher boost in lightly threaded scenarios.

Meanwhile the non-X chips do not boost as high and you are usually able to achieve higher all-core manual overclocks compared to what the XFR can do on its own. So you're not losing anything by doing that, you're only gaining, unlike on the X chips.

Now whether that difference on the non-X chips can actually make a difference outside of synthetic benchmarks is yet to be seen. But it can definitely be seen in benchmarks, that I can confirm myself.

My thoughts: since even single thread applications put some load on all cores (sure, it's not perfectly even but deffo every core is working to some extend), does boosting 2 cores really give more performance than a slightly higher average clock on all cores?
I somehow doubt it and I have the feeling increasing all cores will give a higher fps boost in the end.

So you did the absolute right thing imo and I'm looking forward to your report how much overclocking does regarding fps

 

[RobertR1]

Gamersnexus does a good job if you guys want to dig in.

Ryzen needs good ram but is also more sensitive to ram/board combos. On intel, past 3600mhz doesn’t net you much so you’re better off tightening timings manually (a real pain).

 

[Sim Strata]

I was playing around with the BIOS t'other day. I disabled all but one core and HT off. Assetto ran at 45fps (usual it is 120+). Enabled two cores, a bit faster. AC only runs well with all cores enabled on a 3770K (4 cores). So yes, it is defo the case that there is bleed over to other cores with lightly threaded apps, may O/S optimizations too.

 

[Bob Gnarley]

It's a bit of a long watch but does a decent job of explaining some of the IPC difference between Skylake and Ryzen.

Also for any Ryzen users interested in memory tuning, 1usmus's Dram Calculator is a very useful tool.

https://www.overclock.net/forum/13-...lator-ryzena-1-4-1-overclocking-dram-am4.html

There are a couple of video guides, fairly clear written instructions and a thread filled with every question ever asked on the subject.

Firstly save your current profile in the bios, should you need to return to it!

Note that you will need to download a program called Typhoon Burner.

http://softnology.biz/files.html

Unzip, open and click "Read". Dram Calculator will ask you for some information such as manufacturer, dye, rank etc which you get from Typhoon burner. Get more in depth info by clicking "Report" on or off.

As far as I know Dram Calc does not work for non Ryzen systems.

 

[RobertR1]

I was playing around with the BIOS t'other day. I disabled all but one core and HT off. Assetto ran at 45fps (usual it is 120+). Enabled two cores, a bit faster. AC only runs well with all cores enabled on a 3770K (4 cores). So yes, it is defo the case that there is bleed over to other cores with lightly threaded apps, may O/S optimizations too.

I was running similar tests and turned HT back on (9900k). Between VR and AC, the core/thread usage is all over the place and min FPS seems better with HT on. On a side note, minimum fps in BF V multiplayer is a lot better with HT on also.

 

[Case_]

My thoughts: since even single thread applications put some load on all cores (sure, it's not perfectly even but deffo every core is working to some extend), does boosting 2 cores really give more performance than a slightly higher average clock on all cores?

I'd say yes, because the boost is not static either, it can also move to whatever core needs it. Also, one thing I'm noticing with Ryzen is that it tends to switch threads between cores a lot less than the i7 did. Might have something to do that it has 2 more physical cores and 4 more virtual to work with, or might be due to something else - Ryzen kinda does its own thing regardless of what it's told I'm noticing, for example in regards to power management.

I was playing around with the BIOS t'other day. I disabled all but one core and HT off. Assetto ran at 45fps (usual it is 120+). Enabled two cores, a bit faster. AC only runs well with all cores enabled on a 3770K (4 cores). So yes, it is defo the case that there is bleed over to other cores with lightly threaded apps, may O/S optimizations too.

At least from what I'm seeing, AC uses two threads heavily, plus there's more lightly used threads. And then there's the rest of the threads from the other apps in the system. So it makes sense running it on two cores or less cripples it.

 

[callumjtc]

Time to revisit this thread after the announcement of Ryzen 2 chips. A few months ago I got a Ryzen 5 2600X for a low price, and it has been sitting ever since. Is it time to grab a new X570 board, 16GB of decent RAM and upgrade the sim PC from an i5 4690K and Z97? This will take the system forward 5 years or so and when the cost of the new Ryzen 3000 chips goes down I can move to those? Their claimed single core matches and even slightly betters Intel chips.

 

[Case_]

We need to see some actual benchmarks first. But to be honest, what was shown so far would probably not make me want to update my OC 2600 to a 3xxx series.

 

[RasmusP]

It's not enough for a direct upgrade, I agree.
However the 3700x with 8/16 seems to match the i5/7/9 single thread and therefore also the i9 multi thread wise.
Which seems to be really really great. It's gonna be interesting how well they can be overclocked. Looking at gaming benchmarks the Intels scale 1:1 with oc so bringing the 9700k from 4.6 GHz to 5.0 GHz, which seems to be easily possible with a good cooler, gives 8.5% more performance.
Looking at the current ipc, the new ryzens would need to be overckockable about 250 MHz to still match the performance.

So far, I'm sold! I only need more performance for more or less only single thread applications and games so seeing the ryzen being up there with the Intels, finally, makes me getting the funds ready!

 

[Dan Costa]

I disagree.

If you are still on a 2600K @ 5.1ghz (like I was) you are losing double digits % in cpu-limited scenarios from the single-thread performance and limited ram bandwidth of the 2600k platform. This gap is Irrelevant if using an 60hz monitor, but paramount with VR or 3x1440P 144hz monitors.

Having said that, the R7 3000's are a no-brainer. The price, the single-thread performance. These are 9900K at half the price, with overclock still left to explore. The only reason I wont jump, its because I already have the i9 (@ 5.2ghz and it still boggles down with full tracks and the like).

 

[RasmusP]

I disagree.

If you are still on a 2600K @ 5.1ghz (like I was) you are losing double digits % in cpu-limited scenarios from the single-thread performance and limited ram bandwidth of the 2600k platform. This gap is Irrelevant if using an 60hz monitor, but paramount with VR or 3x1440P 144hz monitors.

Having said that, the R7 3000's are a no-brainer. The price, the single-thread performance. These are 9900K at half the price, with overclock still left to explore. The only reason I wont jump, its because I already have the i9 (@ 5.2ghz and it still boggles down with full tracks and the like).

Yeah I also disagree with the ryzen 3000 not being an upgrade to my old as heck I7 2600k.
I agree with Martin, that it's not worth it to replace his ryzen 2600 yet
Worth for simracing fps maybe, but not really in general

I on the other hand, with my old 2600k (that only runs 4.4 GHz sadly, 4.7 is possible but needs unreasonable high vcore), am getting my funds ready as I said to upgrade this old thing

 

[Dan Costa]

Well if he is on a ryzen 2600 its even more of a reason to upgrade My 2600K @ 5.1ghz was faster than my R7 2700X @ 4.3ghz.

What does he gain? 15% IPC + ~10% clock speed (overclocked ryzen 2600 compared to STOCK Ryzen 3000).

All this assuming he is in VR or 3x1440P@144hz. Anyways, my 2 cents.

 

[callumjtc]

So summary as to what I should do... follow through with what I said?

 

[RasmusP]

So summary as to what I should do... follow through with what I said?

Do you have the i5 overclocked?
I saw quite a few running at 4.7 GHz and higher.
With stock settings (3.9 GHz all cores), you have basically the same, slightly higher single thread performance compared to the 2600x. Multithread the Ryzen is far better of course.
So for simracing you will probably get exactly the same fps, if not less.
In AMS and AC probably a bit less, especially if you overclock the i5, in ACC a bit more.

In the end the best would be to sell the 2600x right now before the 3xxx really hit the market and then buy the 3600.
Mainboard: It's not clear yet if the new chipset really gives anything to normal users. Main improvements seem to be regarding PCI-E 4.0 and nvme SSD RAIDs.
You might get an active cooler, which I always find hugely annoying and pay more.
You might be totally fine with a current motherboard. It's not known if the new chipset will unlock better CPU features/performance.

Memory: yep, 16GB is still the sweetspot I'd say. Only big photoshop projects or movie editing need more.
It might be useful to grab some very high clocked memory together with a x570 board but nobody knows yet. And that memory is ridiculously expensive...

Overall, I would sell the 2600x, overclock the i5 and wait for the new Ryzen systems to be well tested.

 

[Case_]

My 2600K @ 5.1ghz was faster than my R7 2700X @ 4.3ghz.

My 2600k at 4.6 GHz certainly wasn't faster than my 2600 at 4 GHz. And given the diminishing returns of increasing overclock you start seeing even at that point (one reason I didn't even bother making an effort to run it higher), I wouldn't even expect the 5.1 GHz one being faster while having a pretty steep requirements to even run it at that frequency not really worth the effort and money for the gain, but if you say it was, then I have to believe you I guess.

All this assuming he is in VR or 3x1440P@144hz.

That's a very wrong assumption. I'm not even at 1x1440p@144Hz. (And honestly, I would expect anyone having that kind of money available to run something like 3x1440p at 144Hz - because that also assumes a pretty hefty GPU, not just the monitors - to not be running a Ryzen 2600, or any other Ryzen really, nor do I think they should.)

 

[callumjtc]

Do you have the i5 overclocked?
I saw quite a few running at 4.7 GHz and higher.
With stock settings (3.9 GHz all cores), you have basically the same, slightly higher single thread performance compared to the 2600x. Multithread the Ryzen is far better of course.
So for simracing you will probably get exactly the same fps, if not less.
In AMS and AC probably a bit less, especially if you overclock the i5, in ACC a bit more.

In the end the best would be to sell the 2600x right now before the 3xxx really hit the market and then buy the 3600.
Mainboard: It's not clear yet if the new chipset really gives anything to normal users. Main improvements seem to be regarding PCI-E 4.0 and nvme SSD RAIDs.
You might get an active cooler, which I always find hugely annoying and pay more.
You might be totally fine with a current motherboard. It's not known if the new chipset will unlock better CPU features/performance.

Memory: yep, 16GB is still the sweetspot I'd say. Only big photoshop projects or movie editing need more.
It might be useful to grab some very high clocked memory together with a x570 board but nobody knows yet. And that memory is ridiculously expensive...

Overall, I would sell the 2600x, overclock the i5 and wait for the new Ryzen systems to be well tested.

Good shout, mainly ACC is the main sim and an ipgrade would be to chase better FPS in that, the other sims I am well into the 100s so no issue there. I'll need to find a cheap motherboard to test if the Ryzen works for now.

 

[Durge Driven]

RIP Intel at least for my next upgrade

 

[Durge Driven]

On a side note Intel announce 9900KS - 5GHz turbo all cores just so they can kick 9900K owners in the nuts