AMD Ryzen For Simracing?

[RasmusP]

Am I right in thinking that the 'threaded performance' referred to is across all the cores and not something I am likely to see the full benefit of with the likes of AC?

I'm struggling to source the i5 for less than the R5 and added to that the R5 price also includes the cooler and I won't have to consider upgrading my PSU either.

Pretty busy right now but have a look here. The site is perfectly usable with Google auto translate, the benchmarks are English anyway:

Test procesora Intel Core i5-12600K vs AMD Ryzen 5 5600X - Więcej rdzeni, więcej wydajności. Porównanie pamięci DDR4 i DDR5 | PurePC.pl

Test procesora Intel Core i5-12600K vs AMD Ryzen 5 5600X - Więcej rdzeni, więcej wydajności. Porównanie pamięci DDR4 i DDR5 (strona 27) Test procesora Intel Core i5-12600K Alder Lake vs AMD Ryzen 5 5600X - Więcej rdzeni, więcej wydajności. Porównanie wydajności Core i5-12600K na pamięciach DDR4...

www.purepc.pl

 

[GeekyDeaks]

Thanks for the links @blekenbleu / @RasmusP

Comparing the benchmarks of the two with common games seems to be well aligned and puts the i5 well ahead (like 25-30% for Cyberpunk), but it's then interesting to note that for ACC in the Polish article, the results are not quite as spectacular (~8%). I think I might wait to see if AMD release something new or update their prices in Jan.

 

[Hired Goon]

Out of the loop for years re. sim racing but want to get back in. Thinking either 12,600k or 5600x with a 3060Ti. Would that get me there for 1440 gaming with the current games out there? I'll probably be going with ACC.

 

[Novopaine]

For standard single 1440p screen you should be fine with the 3060ti. For triple screens you'll probably need to lower the quality settings a bit to stay above 60fps. Either of those CPUs should get the job done, although I have no experience with 12th gen Intel.

 

[Bob Gnarley]

First review is here for Ryzen 5800X3D for anyone thinking about a final AM4 upgrade.

AMD Ryzen 7 5800X3D Review - The Magic of 3D V-Cache

The AMD Ryzen 7 5800X3D is the company's new flagship gaming processor. It introduces 3D V-Cache, a dedicated piece of silicon with additional L3 capacity. In our review, we're testing how much the larger cache can help intensive gaming workloads and applications and compare it to the Intel Core...

www.techpowerup.com

 

[Trebormoore84]

5800x3D is a weird CPU, slightly worse than the 5800x at day to day and 100% load workstation tasks because the frequencies are slower, id assume much slower when you overclock a 5800x, and on the whole no real performance benefit to most games which I'd assume already fit in the 32mb cashe on the 5800x.
But some select games notably Borderlands 3, Far Cry 5 it gives a significant boost over the 12900k by like 30% assuming cache was the bottleneck. Even at 1440p it seems using a 3080 FTW3.

Question is what simracing titles are limited by L3 cashe if any? Would be interesting to see more reviews to build a bigger picture of what other titles it excels at. 30% is significant if it works in the game you actually play.

Does anyone know if there is a way of knowing how much cashe a particular game is using?

But yeah a 12700k still makes far more sense. or a 5800x or the new 5700x now much better value. Its a one trick pony product. Interesting but flawed.

 

[Peppelino]

Upgrade ? Sure.. It seems to be a
Fire-and-Forget in your old mobo.

 

[Bob Gnarley]

Knowing AMD, this stacked cache will have been made for server (Milan X) but didn't make the grade for whatever reason reason so it is being strapped to a desktop part and sold to gamerz.

As you say, there are a couple of impressive results, but generally it is what you would expect from a slightly lower clocked 5800X. Maybe it would be good for large offline grids but that's just guessing. Other than that i don't see value for sim racing alone.

But, that's ok, that's why we wait for the numbers... At least AMD has finally fleshed out the range with more options. I'm now looking at either 5700X or 5900X which is currently listed at £359 and seems to beat X3D in most situations. Overkill for sims but a great gaming chip...

 

[RasmusP]

Does anyone know if there is a way of knowing how much cashe a particular game is using?

I doubt the developer even know this
You code, you optimize, it runs well or not, done.
Afaik there's absolutely no way to read out how much cache is used. the only possible way I can think of would be to run the code in your IDE and monitor what's happening

 

[blekenbleu]

no way to read out how much cache is used

.. but external DRAM access (AKA cache misses) is actually important,
which can be measured and compared.
Edit: Oops, I underestimated how unobvious seeing this has become on Windows;
at one time they were displayed in Task Manager....
The first tool I find that addresses these statistics is Processor Counter Monitor (PCM)
on GitHub, where L3MISS and L2MISS are relevant.

Here is a recent build for Visual Studio 2019

Edit (oops) again, here is AMD's uProf for Windows Ryzen cache analyses

 

[RasmusP]

.. but external DRAM access is actually important, which can be measured and compared.

Maybe you could reverse engineer the cache usage then?

 

[blekenbleu]

reverse engineer the cache usage

Depending on what is meant by cache usage

• I guess that caches are always full, based on LRU
• if miss rate is 10%, then hit rate is presumably 90%..?

 

[blekenbleu]

Having no experience with F1 2021, I don't know how its performance correlates to e.g. AC,
but 5800X3D results seem encouraging for those already owning a compatible motherboard:

With DDR4-3800 still yielding 3% more than DDR-3200, 98MB cache is seemingly not "too much".

In the olden days (a dozen years ago), when tuning our memory-hungry iOS app,
one trick was to build a dummy object file compiled with optimization to unroll loops,
which would yield roughly 10x code size when unrolling an outer loop of 10 iterations,
then calling it repeatedly from a main routine. By altering the number of iterations,
one could observe at what points this competing dummy app would significantly impact
performance of the app of interest. FWIW, the native iOS app most likely to impact ours
was Apple's App Store when other apps were queued for updating,
but that has all been long since refactored.

 

[pattikins]

Here are some results for ACC...

Please, somebody hold me back. Don't let me buy it.

 

[blekenbleu]

hold me back

If you already have an appropriate motherboard and GPU is not at 100% in ACC,
then why would you hold back?

 

[RasmusP]

Damn that looks very promising!
My 10600k is alright for everything but I'd struggle in vr and and have some drops here and there.
I'm currently waiting for the next amd generation but if it's not that much better or too expensive, this new cpu with a midrange b550 board might be a great option!

 

[Bob Gnarley]

I've been saving for a while, but after that initial review i'd settled on the idea of a 5700X and save some cash but...hmmm...

I hope availability is good!

I Would be tempted to wait for AM5 but i have another AM4 machine that i can pass parts down to.

 

[blekenbleu]

struggle in vr

I recall unhappy Reverb G2 (and some other USB device) owners with AMD motherboards;
hopefully that unhappiness is now all resolved.

 

[2stains]

Damn that looks very promising!
My 10600k is alright for everything but I'd struggle in vr and and have some drops here and there.
I'm currently waiting for the next amd generation but if it's not that much better or too expensive, this new cpu with a midrange b550 board might be a great option!

I just tried heavy rain at Spa with my current settings and it maintained 70+ fps the whole time and by the top of Eau Rouge was back to 90fps. While 100 min would be great my 3080 is maxed at my current settings anyway.
Pretty well all Epic with res scale at 115% at 5760x1080.
3080/10600k/ 16GB 3800 c16.

 

[robt100]

Thought I'd drop this in here as it seemed appropriate, anyone using ACC who is wondering if the new 5800X3D is worh the upgrade, the answer is a definite yes!