AMD Ryzen For Simracing?

[RasmusP]

Some of the 'leaks' indicate quite an uplift in IPC and if you factor in potential clock gains if the achieve half of what some are speculating it should be pretty competitive, especially given Intel's pains in getting reliable 10nm fab running and out of the door.

Read similar things. Can't wait!

 

[Richard Dastardly]

Thirded, fourthed? R7 3700X is looking like my new cpu if it turns out anything approaching how it's looking now. 12 cores ( we'll see ), 5Ghz turbo out of the box and the extra IPC on top, that is looking really really nice.

Now if only AMD hadn't shot themselves in the GPU market foot.

 

[Terry Rock]

I'd fully expect it to be really great if it continues to build on what they have now.
I must say, I'm extremely pleased with the performance of what is currently here.
I'm running rFactor2 'glass' smooth with every detail turned to max...including things like rain and shadows.
I've never, ever done that before.
Its funny to watch Assetto Corsa running at over 200+ fps....also 'glass' smooth.

 

[PA_RacerX]

I'm new to the forum and a bit late to this discussion. I am just now at the stage of picking my wish list of things I need to get started in sim driving/racing, and also to replace my current system. I currently have an I7-6850K with 64GB RAM , GTX 1060, and running an M.2 PCI SSD. I use the system for some gaming but also for video encoding and machine virtualization to test new pre-releases of Windows 10. Right now I think my replacement system will be a Ryzen Threadripper 2920, and most likely a Radeon Vega 64. I may holdout for the Radeon VII as those are soon to be hitting stores and are supposed to be a nice notch above the Vega 64. I chose the Threadripper 2920 because of the cores/threads as the encodes can really chew up a CPU. I have the budget for the hardware fairly well in place and plan to start building in March. My question is if the 2920 would be a disappointment in SIM driving. I have to think it cannot do worse then the 6850K. I also think holding out for Threadripper 2950 wouldn't yield that much benefit. Any thoughts?

 

[RasmusP]

I'm new to the forum and a bit late to this discussion. I am just now at the stage of picking my wish list of things I need to get started in sim driving/racing, and also to replace my current system. I currently have an I7-6850K with 64GB RAM , GTX 1060, and running an M.2 PCI SSD. I use the system for some gaming but also for video encoding and machine virtualization to test new pre-releases of Windows 10. Right now I think my replacement system will be a Ryzen Threadripper 2920, and most likely a Radeon Vega 64. I may holdout for the Radeon VII as those are soon to be hitting stores and are supposed to be a nice notch above the Vega 64. I chose the Threadripper 2920 because of the cores/threads as the encodes can really chew up a CPU. I have the budget for the hardware fairly well in place and plan to start building in March. My question is if the 2920 would be a disappointment in SIM driving. I have to think it cannot do worse then the 6850K. I also think holding out for Threadripper 2950 wouldn't yield that much benefit. Any thoughts?

Look here for single thread and multi thread performance scores nicely shown for comparison:
https://www.cpubenchmark.net/compar...AMD-Ryzen-Threadripper-2920X/2800vs3335vs3354

The overall "cpu mark score" seems to take all scores (there are more but not shown in the comparisons) into consideration so the threadripper are scoring worse than they should.
Single thread points are comparable though and important for simracing!

I picked the i7 9700k for comparison. For simracing it's 90% Single thread, 10% multi thread. The threadripper will give you probably a very slight boost for simracing but not more than 5 fps from my judgment.
The 9700k scores almost 1/3 higher though. But I read it like you really need the amount of cores...
The new ryzen CPUs seem to finally catch Intel in single thread performance, no clue if the new threadripper will be such a jump forward too or not.
If yes, you should wait for them...

A final thought: the single thread performance of the 2920 is actually the highest of all 2xxx threadrippers. Due to the tdp you can probably overclock that one the most too.
When you pick a 2970 for example, you might end up with lower fps than with your current Intel.

I9 9900k would be the best combination of both but it seems you don't want to feed Intel or nvidia any money anymore.

 

[Sim Strata]

Sorry to butt in, but I was looking at the Ryzen 2600X today. I have been looking to replace my old shagged out o/c'd i7 3770K. But when I do a Passmark analysis, it looks like I would have to spend silly money to get anything noticeably better. It still performs better than all the mid range Ryzens at single threaded stuff. Even my 1600mhz DDR3 is within 5% of DDR4 installations! A big "Hmmmmmm"

 

[RasmusP]

Sorry to butt in, but I was looking at the Ryzen 2600X today. I have been looking to replace my old shagged out o/c'd i7 3770K. But when I do a Passmark analysis, it looks like I would have to spend silly money to get anything noticeably better. It still performs better than all the mid range Ryzens at single threaded stuff. Even my 1600mhz DDR3 is within 5% of DDR4 installations! A big "Hmmmmmm"

Yep, for simracing only it would be a waste of money. You might gain some fps since more cores still boost performance but not by much.
Start a game like battlefield or assassin's creed and the ryzen will crush your old Intel though.
Just wait for the next gen Ryzen. Looks very promising regarding single thread performance!

 

[Sim Strata]

Getting 100fps in Verdun (my fave shooter) maxed out. Insurgency Sandstorm is another story...

 

[Case_]

Well, I just got Ryzen 2600 yesterday as I, after a lot of consideration, decided to go with an unexpected chance to finally bring my PC at least partially into the reality of 2018-2019 instead of 2011...and as someone who got quite interested lately (I blame Rasmus) in the performance aspects of simracing/gaming on a multicore/hyperthreaded CPU, it's been quite an eye-opening experience to say the least. I plan on doing a video comparison as soon as I learn more about overclocking the Ryzen, since I want to see how overclocking changes things (it often responds fairly well to OC in benchmarks I've seen), but then again some may be familiar with how much I enjoy making videos and how much time it can take me... (but this one I *really* want to put out, unlike the previous one, so fingers crossed).

So in the meantime, let me just put it this way - even with a relatively weak GPU as my GTX 970 is nowadays (and that's staying with me for a while, could not upgrade everything, sadly) and running at medium to high detail in 1920x1200, Ryzen 2600 at stock speeds coupled with a DDR4 3000 memory, from the games I tested, rFactor 2, Raceroom, AC and AMS are still bottlenecked by the CPU and *not* by GPU. It's a fascinating thing to behold, because it goes against a lot of what I personally would expect otherwise.

All four games will do 60 fps without breaking a sweat, obviously. AMS will be doing 200-300 fps, the remaining games will get you close to 120 fps quite often (and sometimes even over), but they will still be CPU limited, not GPU limited. Only PCars 2 and ACC get GPU limited.

 

[John-Eric Saxén]

So in the meantime, let me just put it this way - even with a relatively weak GPU as my GTX 970 is nowadays (and that's staying with me for a while, could not upgrade everything, sadly) and running at medium to high detail in 1920x1200, Ryzen 2600 at stock speeds coupled with a DDR4 3000 memory, from the games I tested, rFactor 2, Raceroom, AC and AMS are still bottlenecked by the CPU and *not* by GPU. It's a fascinating thing to behold, because it goes against a lot of what I personally would expect otherwise.

AMS and rF1 based games, perhaps Raceroom as well, rely on single threaded performance for most things, so it doesn't surprise me that they are CPU limited. rF2 shouldn't be CPU limited apart from in VR, at least not after last year's performance update.

 

[Case_]

@John-Eric Saxén I know these games rely on single thread performance, but it was pretty surprising to me that even with a CPU like this, it is still not enough to max out even a GTX 970. I just wouldn't expect that. You hear so much of "you should upgrade GPU, not CPU if you want better performance" in gaming that you kinda start assuming that's always the case.

As for rF2, it is absolutely relying on single thread performance. It does use one or two more threads, but it uses them pretty lightly, so they don't really matter. I'm assuming one of the threads is for controls + FFB, because I have set that option on in the JSON.

 

[RasmusP]

Rf2 is a difficult one for me either way though. Jump online with 30 drivers and my cpu limits the fps at around 65 when passing the pitlane etc. (I already set the garage detail lower)

But then the transition to night happens while I'm passing through a dark and shadowed part of the track with 5 other cars close to me and my 1070 just dies and my fps drop to sub 50 on some tracks.

I spent around 4 hours testing different settings, combinations, ai, replay and online situations and if I won't limit my fps they would fluctuate between 50-150 depending on track + car + situation.
I could live with it if the devs would supply a "worst case benchmark" but there is none.
So you set this sim up, look forward to your next race and after 40 minutes the weather or daytime changes and your fps will drop heavily. You can't change settings without leaving the server so you're simply screwed then

Anyway, I think 25 ai at Longford with both gt3 packs, sunset to night + rain is the worst case I've found yet. There's a dark and shadowed part of the track with a slow and right chicane so the field gets bunched up.
Nothing else dropped my fps that much yet

 

[Terry Rock]

@Martin... Do not manually OC that Ryzen5 2600. You can actually lose performance by doing that.
Put the best cooling solution within reason...'price-wise' on it, manually set the DRAM speed and voltage to whatever the memory manufacturer states.
Set the CPU voltage to whatever is stable at or below 1.40 volts....and let it do it's 'thing'.
I have the 2600X on a 470 board with Ballistix DDR4-3000. ACC and everything else I run is 'glass' smooth...even in VR with no manual OC.
I've got just about every setting at epic with anisotropic at 2x and antialiasing at 4X in NVCP at override.
Anything above those two settings is proven to be a waste of GPU resources since the benefit is minimal at best.

 

[Case_]

But then the transition to night happens while I'm passing through a dark and shadowed part of the track with 5 other cars close to me and my 1070 just dies and my fps drop to sub 50 on some tracks.

Yeah, I'm currently running with shadows on low, but the difference between low and medium is pretty significant when it comes to GPU power. Medium (which I normally use) would likely tip those scales in favor of GPU bottleneck a bit more, at least under certain conditions.

@Terry Rock I am aware of the various "schools of thought" when it comes to overclocking the Ryzens. I'll see what works best for me. I've seen people mentioning a lot that the boost features seem to work better on 2600X, though. Which I would've like more, to be honest, but just couldn't justify the 60 euro difference in price it currently has around here.
But cheers for the tips.

 

[Redvaliant]

I'm thinking of getting a R2600x and 470 board bundle for about £350. I saw a review that mentions that Ryzens like fast ram, the faster the better; the reviewer recommended 3400hz minimum! But holy **** are ram prices stupidly high - 180 quid for 16gb is the cheapest 3200 I saw on Amazon. Still, looks like a cheaper, better option than the relevant Intel options.

 

[Case_]

But holy **** are ram prices stupidly high - 180 quid for 16gb is the cheapest 3200 I saw on Amazon.

Really? Holy crap. The cheapest 3200 2x8 GB kit I can get around here (Kingston HyperX) is around 115 GBP (converted, obviously). I did get a 3000 G.Skill 2x8 kit for 100 GBP myself.

And as to the cheaper, better option than Intel...I guess it depends. I don't regret getting the Ryzen at all, but knowing what I know now, I might've stayed with Intel, as I was very undecided which way to go. But I guess in a way it's great to have options.

 

[Terry Rock]

I'm thinking of getting a R2600x and 470 board bundle for about £350. I saw a review that mentions that Ryzens like fast ram, the faster the better; the reviewer recommended 3400hz minimum! But holy **** are ram prices stupidly high - 180 quid for 16gb is the cheapest 3200 I saw on Amazon. Still, looks like a cheaper, better option than the relevant Intel options.

They're lots of folks using the DDR3000 and DDR3200 who've reported no difference in performance between the two.
They have reported instability above those numbers.
I personally use the 3000 with the recommended voltage of 1.35 as called out by Micron.
It offers me a tremendous amount of stability.
On my MSI gaming Plus, I noticed the CPU voltage would occasionally go as high as 1.438volt ...even set to auto, so I capped it at 1.40 and it is rock stable there.
My chip will auto overclock to 4.4 GHz all cores since I have adequate cooling.
It has been a very pleasant experience thus far. I can run every game I own currently on high to max settings.
My last AMD chip was an Opteron 170, so you can see how many years I owned Intel chips.

 

[Terry Rock]

Yeah, I'm currently running with shadows on low, but the difference between low and medium is pretty significant when it comes to GPU power. Medium (which I normally use) would likely tip those scales in favor of GPU bottleneck a bit more, at least under certain conditions.

@Terry Rock I am aware of the various "schools of thought" when it comes to overclocking the Ryzens. I'll see what works best for me. I've seen people mentioning a lot that the boost features seem to work better on 2600X, though. Which I would've like more, to be honest, but just couldn't justify the 60 euro difference in price it currently has around here.
But cheers for the tips.

There is very little difference compute-wise between the 2600 and the 2600X where their operation is concerned.
As long as you can supply adequate cooling, both will self overclock to the exact range...provided you have a good chip.
The only reason I have an 'X' is because I got it on Black Friday for the cost of the non-X chip.

 

[Sim Strata]

They're lots of folks using the DDR3000 and DDR3200 who've reported no difference in performance between the two.

According to passmark, I am not surprised. my DDR3 1600 is just a couple of hundred points slower than most DDR4 2400, 3000 and 3200 installations for buffered and unbuffered reads with a bigger difference on writes, but not much. Indeed, my RAM is often faster than many DDR4 setups. The greatest exception seems to be fast Gskill RAM.

This seems to be backed up by bench-marking articles for LGA1151 configs looking at different RAM speeds too, which found not noticeable difference for gaming.

 

[Case_]

RAM speeds didn't matter much with DDR3 for gaming and performance in general. It's a bit different with DDR4 on current CPUs where RAM speed can at times have a fairly noticeable impact on performance. Though the difference between 3000 and 3200 is obviously not that large.