[FM20 Performance Benchmarking Thread]

29 minutes ago, Brother Ben said:

I think the major issue this time around was that unbeknownst to me the English transfer window had moved this season so that would definitely have had an effect especially as Watford are used for the test.  Most people on this thread tend to gauge their results against other systems and not other game versions in my experience.

Like I said though its very much a work in progress although I am liking the idea of a bear minimum test as a mirror to the "maximum everything" test i'd like to know if there is a point where there's so little to process that it doesn't really matter what processor you have.  You can still play the game albeit at bare minimum.  The bare minimum in this case would be probably just the premier league with a small database.

There is a How many nations do you usually load? thread but it doesn't really allude to how many leagues people run which is the big processor drain

Food for thought anyway

Oh I see. That makes sense and will affect processing time. Of course, it shouldn't matter for a CPU bench as it's the same save file everyone is going to use. But it will skew the results when comparing to previous versions. But comparing different versions isn't really meaningful, except to satisfy curiosity maybe and to gauge improvements in efficiency over the years. At the end of the day, how many people are going to play a particularly fast old version? I know there are legendary versions that people just love, and data update activity is pretty high. I myself have CM 00/01 installed right now, It was to test some things and also it is clutter free very responsive and you can breeze through it. Most will play the latest or the one before that, just in order to not abandon a save they invested in. 

One tip. Make a standard benchmark that takes place outside transfer windows. Since youth intake also significantly slow things down, especially the big nations, you'd want to avoid that period too. It's not just the slowing down, you just want to minimize variance. So, very early February after the window has surely closed to mid-February, when there won't be any intake from the Big 5 would make a good candidate. And retroactively, the same saves can be created for earlier versions, as long as people are willing to bother with it and players are willing to participate in the benchmarks.

Edit: Not allowing matches to be rescheduled for TV broadcast would also be needed.

[FM20 Performance Benchmarking Thread]

2 hours ago, Brother Ben said:

will do, cheers for the advice.  No doubt it'll be me doing the upgrading though.

I've been thinking about next years benchmarks and I think its gonna be a bit different next year, I kept them the same as previous editions mainly to give comparison but its almost impossible to recreate the same conditions so I may well have a complete rethink.  I'll keep the all leagues all players benchmark as its the bread and butter of this sort of thing and lets the guys with the mighty hardware flex their muscles but i'm going to have a long hard think about the other benchmarks

I keep wondering if maybe SI could help out with this and whether they know what the most common amount of leagues/nations player count etc is

Why do you believe that it is so hard to recreate the conditions? I never really understood this. After all, it's just a certain amount of data and the processing. I mean, added features, like I don't know, social media, press conferences, media etc. might adversely affect processing time, but the database and game setup should work the same way, maybe?

As we talked about earlier, if you're looking for a meaningful comparison of different FM versions, all you need to do is choose the right dates. So it starts in the same period of a season, say CL first week matches, and then the normal over the weekend matches for the leagues. So, the same type and same amount of matches are played and calculated by the same method over a long enough time to reduce standard deviation and give statistically significant results. 

AS for the most common setups, maybe a sticky ongoing poll in the main forum would help. I generally pick 1 active nation, the remaining big 4 top 2 tiers view only for both players and increased transfer activity. Maybe internationals, and top clubs, etc. as well. Don't remember the player count I had in my last saves. Maybe 80K or so.  

[FM20 Performance Benchmarking Thread]

9 minutes ago, Brother Ben said:

Definitely something to think about but his budget is pretty tight and he’s not very tech savvy so he’d be unlikely to appreciate the difference. It’s gonna be like night and day compared to his current second gen i5 laptop

Yeah. Even my i2500K was just fine not really bad at all the last time I was playing FM. Even in a poorly optimized game like SCUM it was just fine. At least give him some upgradability options, ie motherboard. The B450 is the budget chipset to go for. He'll enjoy the option of upgrading to 3rd Gen if he pleases, 2 or 3 years down the line. It'll be less confusing than an Intel build. 

[FM20 Performance Benchmarking Thread]

17 minutes ago, Brother Ben said:

@djinni999

I like the Look of the Ryzen 3 3200G, it’s just £75 and I’ve got a mate on an insanely tight budget who wants a desktop for FM and I’m looking for a reason not to build this for him

https://uk.pcpartpicker.com/list/dmGsNq

he only wants new parts and will probably only run 5 nations maximum

I think it’s a pretty sweet deal, what do you recon?

The G is APU, so he won't need a GPU, which he really doesn't for FM anyway, so yeah, that's a very good choice. But how much is the 3300X? 120 quid? Plus a GPU, so if it's purely FM and nothing else for years, then yeah go with that. I don't know the ST performance of that CPU though, so it may or may not be a sweetspot. let me check

EDIT: 3300X is 21% faster for single-thread. So it may be worth it to shell out slightly more. Plus, 3300X will benefit fully from RAM OC/tweaking, which no other Ryzen will. 

[FM20 Performance Benchmarking Thread]

7 minutes ago, Brother Ben said:

Depends on use case, see the discussion above. There are pros and cons to both.

If you want the absolute best performance for FM then it’s gotta be Intel, the stats on the first page bear this out but it will come at a cost and upgradability as mentioned.

that said my 6 year old i5 is still performing admirably and makes short work of 20+ leagues

 

The fastest Test A score is 0:53 seconds for non-overclocked. My 3300X does 1:08. That is a 28% improvement. The cost differential is 300%, ie. 4 times the price. Maybe, this will help put things into perspective. 

[FM20 Performance Benchmarking Thread]

4 minutes ago, shai82 said:

so you recommend AMD for the long term? id dont want used or previous gen. 

maybe i will wait for zen 3.

thanks.

No one can see the future and there are no guarantees. But, comparing Intel and AMD's past track records of recent times, AMD is doing much better. They are delivering on their promises. The 3rd Gen Ryzens are going to be pretty good. My 3300X is a transitional CPU between 2nd and 3rd gen, it's not like the other Ryzens. 4th Gen, which probably will be supported on B550 and X570 boards (though no guarantees) should be a significant upgrade over a typical Ryzen today (2600, 3600, etc.).

Honestly, if you don't want to spend that much unnecessarily and minimize future-proofing risk, then just get the 3300X for $120 now. Get a B450 board that is bargain-priced but perfectly fine in terms of features and performance. I hear MSI Mortar and Tomahawk get recommended a lot for budget builds. If there is a socket change for 4th gen Ryzen, you won't have invested as much into a X570 high-end board like I did. If you're willing to spend more and intend on overclocking a future 3rd gen CPU, then yeah a mid to high-end X570 or B550 will be needed. 

If you're willing to wait for 3rd Gen Ryzens to hit the stores, then that is very reasonable. If you find that a particular CPU is reasonably priced and offers a noticeable increase in performance, then get that. If not, you can always fall back to the 3300X, which will have become even cheaper by then. 

[FM20 Performance Benchmarking Thread]

43 minutes ago, Brother Ben said:

True but I wasn't basing on cost or upgradability.  Of course Ryzen is the better processor on almost every metric you can think of.  Just not the typical FM set up.  That's why I always ask what someone plans to do with their PC if i'm recommending processors. 

You're probably correct about 8th or 9th gen cpu's used.  Trouble is not a lot of people want used gear.

Maybe i'm biased.  I just don't have any software that would even come close to needing all the cores Ryzen has to offer.  My most demanding piece of software is FM. As soon as AMD can reliably hit 5 GHz i'll make the switch.  I'm skint though so that's by the by 

It's not a bias, it's being reasonable. If you're only going to game + internet, etc. Is there a single reason to get a 3900X? Even a 3700X is overkill. That's where the 3300X comes in, it is essentially a proper gaming CPU that is super-competitive in pricing, no bloat in terms of idle cores.

My present AMD lean has less to do with AMD or Ryzen than with the AM4 socket. Intel has an excellent reputation for changing the socket for no reason at all; with pathetic performance improvements over generations. My X570 board will support 3rd Gen and very probably 4th gen, and possibly 5th gen. If one gets a 10600K now, they will be stuck with it  unless they are willing to spend significantly more for a possible hypothetical future flagship CPU that is only 10-15% or so better in (gaming) performance (the 10900K, even, is not even 10% better in single-thread). Otherwise, they'll almost certainly face a motherboard upgrade as well.

Two years ago, especially with an upgrade as opposed to full rebuild, the 8700K and 9700K would have been no brainers. They're still not cheap today second-hand, I can't imagine what they cost back then. BUT you would have gotten what you paid for, with no better alternatives that were reasonably priced. In hindsight, I'm lucky I didn't upgrade back then, as I would have been stuck with those CPUs and having invested so much in that dead-end system. My trusty i2500K (stock) got me by just fine. 

[FM20 Performance Benchmarking Thread]

13 hours ago, Brother Ben said:

Test 3 has more than 20 leagues running in full detail and unless he plans to do this, which 99% of people won't (you're better off using that processing power for more active leagues) then it's hard to recommend Ryzen at all purely for FM.  I agree that for the future it may well be a better choice but there's no getting around the fact that in basic terms clock speed is king on FM and unless AMD can address that then the typical user, represented here by benchmark A, is better off with Intel

Big fan of your Polish database by the way, played the FM19 version a fair bit 

Yes, that makes sense. BUT, when you factor in costs, CPU+MB, and upgradability, then I don't think 10600K is that great an idea. I'd consider 8700K or 9700K used, if you want intel and a short-term build. If not, an Intel build makes zero sense unless you like throwing money around.

[FM20 Performance Benchmarking Thread]

4 minutes ago, shai82 said:

hope its ok to ask here, can someone recommend me for a new CPU?

i only play FM but i want it to be fast enough with 20 +- leagues.

i thinking about one of those CPU's:

Intel Core i5 10600K

Intel Core i5 10600

AMD Ryzen 5 3600

AMD Ryzen 5 3600K

AMD Ryzen 7 3700X

 

thanks

do you need many cores for other workloads? Is it mostly gaming you're interested in?

[FM20 Performance Benchmarking Thread]

4 hours ago, AlWakRa said:

I tried it, it was slower, did A and C, I made a re-run with current settings as benchmark, did it faster than the original benchmark, 1:08 and 7:04, but no point in trying for the best possible times as these only a benchmark not a drag race, without under-volting, 1:12 and 7:41.

 

Thanks for the insight, I was very suspicion of these speed reporting, as I used asus integrated app and task manager, and they report different speeds, anyway, my understanding on the matter, the chip performance can vary, and not all chip produced equally, there is some margins to consider when they do the default setup, they consider the variation and go for something safe for all chips, it is the same logic to overclocking, not all chip able to reach same speed with same setup. And you are right, to get the sweetspot of performance takes time and lot of stress tests to make sure everything is stable. 

Yes, exactly. After every tweak, you need to run something like Cinebench CPU stress test. Not only to test stability but to get a score. If your score is lower, it's not optimal. If high temps is a problem, then you sacrifice slight drop in performance. If it isn't you need to find the right max Voltage that gives you as good a score as stock. This is just for undervolting to reduce temps. Overclocking is a different matter but similar in methodology. Don't rely on clock speeds being reported in software, only bernchmark results, not even FM benchmarks, as it may not put a full load on the CPU continuously.

Also, don't use two hardware monitors at the same time. They will interfere with eachother. And with most monitors you have an 'observer's effect'; while monitoring the software is stimulating your CPU and skewing the results. Use Ryzen Master for temps and CPU clocks, etc to get the most precise measurements.

[FM20 Performance Benchmarking Thread]

On 19/06/2020 at 22:18, AlWakRa said:

 

Those Ryzens are monsters, I tried to convince myself to get a Ryzen 9 4900HS based laptop, they were cheaper than the one I got, but the screen was the deal breaker for me, nothing with good color coverage for photo editing, but their power are no joke for video editing and productivity, looks like they aren't treated with much love from laptop manufacturers, or maybe it is early days and in the future there will be better options. 

 

As for undervolting, I was reading the available reviews for my laptop while waiting and one of them suggests undervolting to makes the processor sustain over 4.0Ghz on all core for longer time, which proved to be 4.3 in my case on all cores, without doing it, it can jump to higher clocks for a moment but it will heat up quicker and throttle down, if I have time, I may try without undervolting, as my temps weren't very high, to see if there is any difference, but usually a 5.0 Ghz turbo speed doesn't mean all cores can reach it, if I know which cores are capable of that, I can assign them to fm2020 and see how it fares, but this is taking it a bit far with thinking and tinkering  which I don't want to spend time on, at least if I can turbo with higher speeds for long time, it will benefit other tasks.

 

Anyway, I am happy with my purchase in general, as I can work on all tasks more efficiently, in fm case, where I spent only brief time, it didn't lag or stutter like before, it is more smoother experience.

Regarding undervolting; when you set a certain clock speed, it can appear to be reaching it when you check it through software such as Ryzen Master or HWMonitor. This may be falsely reported by such software. They are being tricked into seeing those clocks when actually they aren't hitting those speeds. The only way to be sure if you are taking a performance hit is a full-load benchmark. 

Basically, you undervolt but keep your clocks as high as you can, thinking you're gaining reduction in temps at no performance cost. The reality is that your CPU will not hit or maintain those clocks when they are required. 

That said, there probably is a sweetspot of voltage/real cpu performance, but finding it may not be as easy.

[FM19 Performance Benchmarking Thread]

I see. But I did set it up with almost the exact player count, database size. Anyway, so do you think the extra 8 cores is the difference between 2:42 and 3:12 times on the 102 league bench?

[FM19 Performance Benchmarking Thread]

1 hour ago, Brother Ben said:

Sounds about right for Ryzen on FM19, there's a 3900X on page 2 that gets 2:42.  bear in mind Benchmark A is different for FM20

I did state that I used FM20 bench A settings for FM19. I equaled the 9700K at 1:00. Using FM19's Bench A, I got 3:12. I am aware that the 3900X gets 2:42, which is the very reason that I made the post. I was expecting it to be close. I'm confused as to the discrepancy. Could it be that loading 102 or so leagues totally taxes the 4-core, while the 12-core is able to handle it much better? FM20 Bench A has only 20 leagues loaded, so that could explain why I was able to beat the 3900X there. Also, it seems FM20 has benefited Ryzens more, as they seem to have closed the gap between the top Intels.

[FM19 Performance Benchmarking Thread]

Hi guys. I did benchmark A. My system is Ryzen 3300X, a bit undervolted but @3.4 GHz all-core, 16 GB 3200CL14 RAM, 1080 TI, SSD. I got 3:12 and I'm a bit underwhelmed. I was expecting better. Is my result normal? Using the FM2020 Benchmark A settings on FM19 I got a really good score, beating all Ryzens and tying the 9700K for 5th place. 

[FM20 Performance Benchmarking Thread]

Yeah, that's why I only did the Bench A using your 2020 settings. I chose Spain for starting league and only this is playable. All other leagues are view only. Then I also reduced detail to 'none' for the other leagues. Spain is La Liga and Copa full, others are none. Is this correct? And also, when you go back or forth a year, the day is not the same for a particular date. Both 19 and 20 benchmarks A start at 8-20, but greatly less or more matches might be played depending on the coinciding day (TUE, WED for CL, THU for EUROPA, WED and SAT for INTL). This would greatly affect bench times accross two versions. Am I right in my thinking?

[FM20 Performance Benchmarking Thread]

i5-I2500K (4/4) @3.3 ghz, 8GB DDR3 1600 mhz, Integrated GPU, SSD got me 1:58 for Bench A.

[FM20 Performance Benchmarking Thread]

I've decide to give it a go on FM19. Can you list which nations were chosen for the 10 and 51 nation saves. Does it matter if I pick just random nations to increase the count to 51, as long as the player count is the same?

[FM20 Performance Benchmarking Thread]

4 hours ago, Brother Ben said:

I don't think it will work on the demo i'm afraid, the demo only allows you to run one nation iirc

For what it's worth its definitely worth buying the game.  If you decide to then pop back and give it another go

Yeah. My intention was just to try it out of curiosity for my old 2500K  to see how it fares and also to contribute to this benchmark project. I have 2019. Is there a chance you can make the equivalent saves for FM19? Or I can make my own, though I'm not sure which settings to choose when setting up the game; there is some info in the OP. Thanks.

[FM20 Performance Benchmarking Thread]

Thank you for the civility. I wasn't aware of who the OP was, not that it matters. I never said there was something wrong with the save files. Also, I have had huge issues installing Win 7 on Ryzen, these past days, spent hours, no solutions worked. I am generally pissed at thinbgs not working for no reason. Maybe that contributed at my frustration.

That said, I didn't mention that my FM is a demo version from steam. It SHOULD NOT matter, but I'll mention it for what it's worth. Fresh install. Is clean cache necessary? Thank you.

[FM20 Performance Benchmarking Thread]

13 hours ago, Brother Ben said:

Too right mate, I bet the bloke who started this thread messed up the upload.  The idiot. You just can't get the staff 

My advice though is to download the file put it in the save game folder, load the game, clear the cache, then shut down the game, then reload the game. 

Smug, thinly veiled comments like this reveal that the person isn't as bright as they think they are.

1. I wasn't necessarily blaming the OP and his files.

2. Your advice isn't helpful. It is a fresh install, with the saves in the proper folder, appearing in the load game screen, with multiple shutdowns and reboots of the game. THEY WON'T EFFIN LOAD. Yeah it must be my fault. How stupid I must look now. FFS.

[FM20 Performance Benchmarking Thread]

I'm trying to participate in this but I'm getting a ******** savegame could not be loaded message. Does ANYTHING ever properly work in this world. The fcckkknnng incompetence I have to face daily.