CPU or GPU upgrade?

[AquilaRossa]

Turning off SMT helps? I'll try just using Ryzen Master.

Somebody saying they got improvements using a 2080ti makes me think extra VRAM and bandwidth that card offers could actually be beneficial (I definitely see the game stall to load textures using just 8GB of VRAM). Therefore I wonder how a Radeon VII goes on this game with its 16GB of high bandwidth memory. It's only roughly on par with 2080 performance and the cost is not justified considering a 5700XT matches it most of the time, but I just wonder about the VRAM capacity and bandwidth and how it would work with X.

I am just guessing though, because I have not checked my VRAM usage monitoring over a period of game play to see if the textures loading is down to full memory usage, or just the engine not making sure those textures are where they are needed. The most common examples I see is planets being absent for half a second when i go thru a gate, or the other is thumbnails for modules missing in the build editor for a second. I am using a SATA SSD rather than a NVme, so looking at that upgrade may help get bits to the VRAM faster. Maybe.

I also wonder if the 2080ti tried to run my save game it would be crawling with the map open too. Opening the map halves FPS. Maybe crawling a little faster.

I just flew around in the Matrix sector north of the Teladi HQ clearing Ms and Ps that were crowding the gate and not going through it. Frame rates were playable even if a lot lower than game start. The I go to where my HQ is and I am getting 30FPS because there are two massive complexes near the gates. I'd like to share my save to see what other people's systems get in the sector near those complexes. That will help me decide CPU or GPU first, although I am in no hurry and also want to see how the upcoming patch helps.

That makes me wonder how they will deal with the existing factory spam and huge numbers of ships in current save games. I think a liquidator script to get builders to dismantle and remove factories that are failing, but what if they are all thriving? i dunno.

[ballti]

SMT off can help in some games. X4 = up to 8 core, do test FTW. Also oweclock 4+4 vs 8+0, 4+4 core use less power, so eazyer for owerclock.

[AquilaRossa]

I just tested SMT off on the same save. With it on 60-70 FPS drops to 20-30 FPS with the map open. With SMT off the result is exactly the same and the input lag for mouse clicking menu items felt the same too. Bummer. I thought it might work.

Still undecided whether to get GPU or CPU first. It depends on which drops in price first i guess. Both will help with X4 only a bit until the patch, so no hurry. CPU will be a 3700X. I like power savings. Not sure about GPU but probably a Sapphire Pulse 5700XT. The other option is boycotted out of protest at The Great Turing Robbery. I have had only their stuff since the very first generation back when the alternative was stuff like 3DFX, but if a friend tries to rob you, do you stay loyal to them? Nope.

p.s. Re overclocking. I found it is not really worth it with a Ryzen 2700X, unless a few percent matters to you and you may or may not get it. PBO etc seems to get the most out of the CPU automatically. Only advantage of better cooling I found was that throttling does not happen in gaming workloads so far at least (Noctua NH-U12A). So the cooler was definitely worth it, even though the stock one actually did okay. The only thing I want to tweak is vcore and perhaps tackle trying to improve on XMP settings for the memory. I think the vcore runs higher voltages than it may need to, I see 1.5v from stock settings at times. That seems very high to me. Perhaps mobo makers trying to give their boards an edge in benchmarks, so they get loose with the voltage to get there. It's a Strix B450. I was sucked in by the YouTube review. It's just a bog standard B450 really with added bling. Bling and looks are not highest on my list, as my use of Noctua indicates. I do like RGB though. I thought it was just a silly gimmick until I saw it in person. Now I'd like my whole apartment to have it

[ballti]

XMP will not show potential of your CPU. Turning off SMT and some core may disable worse core and alow further owerclock.
Upgrading on "intel OC 5.2Ghz" is wey 2 expensive just for 1 game.
Upgrading on "R 3000" can benefit ~same like "OC R 2000".
2700x is not bad cpu even for RX 5700x.

[AquilaRossa]

2700x is not bad cpu even for RX 5700x

Going to a 3700X from a 2700X does feel like a luxury purchase for sure, especially considering spinal injury has ended my career and forced me into early semi-retirement. I do not have a lots of spare cash to throw around. I justify it by the 65W TDP lowering my power bill and the rest being covered by selling the 2700X. That's my story and i am sticking to it

Gamer's Nexus was saying just the other day how XMP does not unlock the full potential of memory with Ryzen. I will have a look at timings etc. I do not enjoy tweaking as much as I used to. I am also wary of pushing things too far, so i do not increase memory voltages. Broken parts are more difficult for me to replace these days. I also like to be able to keep a straight face when I sell a part and assure the buyer I have not overclocked the hell out of it.
My old 2500K overclocked well, but I think they deliberately built all that headroom in by setting stock speeds quite low compared to very high potential. Ryzen goes almost as fast as it can out of the box. I remember when I had an AMD Athlon C Thunderbird back in about 2001. Pencil lead in between two points on the CPU would unlock the multiplier. Good old fashioned rough and ready overclocking. The most free performance I got from a CPU was back in the day of OC'ing Core 2. Wolfsdale if I remember correctly.

But to get X playing smoothly when the map is open in late game, I need twice the performance and even then i am not sure it would help. HOP would just build 700 more factories and I'd be back to square one. I do not mind performance drops if I build 1000 ships and fly in the same sector. But that is optional. Can't be long now before we see that beta patch and how Egosoft decided to tackle the problems. It reminds me of Hearts Of Iron by Paradox. They try to fix one thing and it creates new problems. In NZ we call that 'pushing sh!t uphill'.

[h1ght]

playing on a i5 7600k@5ghz and vega 64 at 1440p/windowed. lowest graphics settings getting cpu limited and gpu chills between 10-70% usage so far. i try to test other settings. there is some other bottleneck.

regards

edit//
max settings, gpu limit of course. cpu chilling at ~40-70%.

btw. is there a chance to enable freesync and to limit the refreshrate? e.g. max. 90hz and fps limiter nativly? the few games i tested worked better with ingame fps limit and refresh instead of forced fps via driver. less stuttering. wasnt successfull to set refreshrate for x4.

[ballti]

Ah, Steve and similar ppl need to stey between users and companys, they are very limited whit tricks. If you limited whit $$ then downgrade on r5 1600 and save some $$. Weit for new gen of cpus, this performance is old few years. Even i7 9700k in "i5 shape" is vey too old single core performance..

[AquilaRossa]

I made a decision.

I was waiting for the 2.60 beta to see how it affects performance. Initially it was a huge improvement as they took lots of objects out of our save games. Easily 30% when flying in space in the average sector. FPS still dropped by nearly half when in map mode at a station, but that improved a lot too, i.e., half of 30% better. However the gains evaporated as new stuff got built. Not quite back to square one, but almost. A very slight increase in performance overall. -- maybe.

I tested all graphics quality preset settings at 1080p in a quiet sector. Ultra to low made almost zero difference to FPS. That is almost always a CPU bottleneck, although I think a better GPU than my Nitro RX590 will help in the sectors where lots is going on and textures do not load as going through gates etc (it's a good mid priced card and imo better than the GTX 1060 that was the alternative. Cheaper too. But it has its limits for sure).

I decided to get the CPU first, but only because the Sapphire RX5700XT Nitro is not out yet and will be very pricey for a few months ahead. I just ordered a Ryzen 3700X and expect it to run the game at least 10% better. Sell the 2700X here in NZ. I did not want to upgrade within a year, but the hell. Ryzen 9 is almost twice the price and would be a stretch on the wallet right now (four extra cores would not help X4 either). Divorced now, so no other half to have to negotiate with over money. Happy days, although I rather not have this damn spinal injury and be back out at sea. I do not like being a land lubber. I like exploring, which i guess is one reason X games appeal to me.

If anybody is interested I can take notes of before and after with a save game and post if FPS increases. That might help folks if they are trying to decide the same.

[linolafett]

Enjoy your new toy then!
I guess its great to get some results from you, how much faster the game then runs.

[alt3rn1ty]

I recently got a new laptop which has a hex core CPU, I7 8750H 4.1ghz
6 physical Cores, 12 logical cores

After monitoring the current 2.60 Beta 1 quite a lot, all 12 logical cores are being utilised very nicely. If you watch one its usage percent (MSI Afterburner) ranges between 0-40% (I have seen a couple of them rarely spike up to 60%), which fluctuates rapidly while playing. The same behaviour and percentage of use happens for all 12 logical cores no matter where you are in the game.

So the beta in my experience is utilising all cores very nicely to spread the load. Compared with my previous laptop which only had 4 cores, my current laptop is managing to run the game with a lot more CPU horsepower to spare.

I dont know how an I7 compares with an AMD CPU, but it seems to me the more cores on a more recent CPU the better the game uses them, and performance benefits very nicely.


Edit : Forgot to mention - The I7 dynamically adjusts its power use between 2.2 - 4.1ghz, I have seen another game take it up to 4ghz once (while I was doing silly things with settings), but X4 has never taken it above 3.7ghz with the distributed usage percentages mentioned above.

[Imperial Good]

After monitoring the current 2.60 Beta 1 quite a lot, all 12 logical cores are being utilised very nicely. If you watch one its usage percent (MSI Afterburner) ranges between 0-40% (I have seen a couple of them rarely spike up to 60%), which fluctuates rapidly while playing. The same behaviour and percentage of use happens for all 12 logical cores no matter where you are in the game.

Which is likely the result of the OS rescheduling the threads between cores. Hence why no core is hitting near 100%. Two cores hitting 50% utilization does not mean that an application is well multithreaded, it can mean it is just single threaded with the thread spending half the time on one core and half on the other for an average of 50% on each.

I think the scheduler does this with Intel CPUs to get around boost duration limits or spread thermals. On Ryzen third generation CPUs the Windows scheduler seems much more willing to keep applications pegged to only 1 or 2 cores because there is no duration limit and instead specific cores are faster than others.

Edit : Forgot to mention - The I7 dynamically adjusts its power use between 2.2 - 4.1ghz, I have seen another game take it up to 4ghz once (while I was doing silly things with settings), but X4 has never taken it above 3.7ghz with the distributed usage percentages mentioned above.

This can be for a number of reasons. For example the power settings of the OS, or the application deadlocking, or the CPU hitting a limit such as thermals.

However in your case it almost certainly is because 4.1 GHz is your "Max Turbo Frequency" which according to Intel...

Max turbo frequency is the maximum single core frequency at which the processor is capable of operating using Intel® Turbo Boost Technology and, if present, Intel® Thermal Velocity Boost. Frequency is measured in gigahertz (GHz), or billion cycles per second.

This is the maximum single core frequency, for 1 thread running on the entire processor. Multi threaded applications usually cannot hit this frequency because all core clock speed is lower. With stock settings this frequency can only be run at for a short period of time, a couple of minutes max, before the CPU must revert to a lower frequency. The OEM might have disabled this time limit if they felt the cooling was sufficient, however only really expensive laptops would consider such a thing due to how thermally constrained they are.

In your case the multi core profile is...

4,100 MHz (1 core),
4,100 MHz (2 cores),
4,000 MHz (4 cores),
3,900 MHz (6 cores)

Hence 3.7 GHz likely means some other limit is being reached. Assuming measuring fastest core (not average).

CPU mostly becomes a bottleneck late game where there is a lot of activity happening. Early game graphics or frame limits usually are the bottleneck.

[AquilaRossa]

Okay. 2700X removed and 3700X CPU installed. Before and after test on a save in late game with lots going on are done. 2.60 beta patch.

I tested this way:

Windows environment normalized for both CPUs, i.e, same background tasks like Steam still running but no applications like Firefox open.
RX 590 with default software settings in the latest drivers. Game running high preset at 1080p with all AA turned off. Adaptive sync off in game, but Freesync is enabled in Radeon settings. Monitor is native 1080p with 75Hz refresh.
16GB Trident Z running 3200 MHz XMP.
Both CPUs running default bios settings apart from XMP on and Realtek audio disabled.
Sound is provided by an Audient iD14 recording interface at 24/96 resolution via USB 3.1.
$team directory and Win 10 pro OS on separate SATA 3 SSDs.
Noctua NH-U12A for CPU cooler.
EVGA GD 650w PSU.

I benchmarked each CPU by loading a save after freshly relaunching the game from Steam (sometimes after long playing sessions FPS will increase when shutting down the game and relaunching, so I had to remove that variable). In the game save I am standing still on a HAT Free Port platform in Silent Witness XI. Then I open the property map mode to record how much FPS drops. I then teleported to my HQ that is very built up and did the same.

Numbers given are averages and typically the swing was one FPS up or down. Game is modded a little (no trading mods though). Save is seven game days in. Lots of stations and ships built, so it is well past that point we see where everything slows down to crawling speed. FPS in a new game start will be a lot higher than these results (it would typically be twice as high on this computer, or more now). I did not think to also test a new game start and I am not keen enough to know to do a CPU swap over again.

Results for 2700X:

HAT Free Port: 40FPS. 30FPS in map mode.
Grand Ex. 1 HQ: 20 FPS. 12 FPS in map mode

Results for 3700X:

Hat Free Port 48 FPS. 37 FPS in map mode
Grand Ex. 1 HQ: 22 FPS. 22 FPS in map mode.

Conclusion:

In a quiet sector the CPU was a greater bottleneck and the 3700X gave a significant boost in FPS. At a mega complex with hundreds of ships there was only about a 2 FPS increase, but a 10 FPS increase in map mode. I do my fleet and station managing while in quiet sectors. The FPS increase will make the map mode and building mode run smoother. In very busy sectors with lots of objects the CPU will help with processing instructions for their AI, while a stronger GPU is needed to draw all the objects faster and at higher quality settings. The answer to the CPU or GPU question is both will help, so the decider became money. I chose CPU first because it was slightly reduced in price and the Sapphire 5700XT I want will not go on sale for some time yet (I am allergic to paying the early adopter's premium, which on top of saving money has the bonus of sparing me the teething problems new hardware and software often has).

Was it worth it? I think so for me at least. I run a digital audio workstation on this PC and it will run better now. My DAW software called Reaper and also Cubase really likes Ryzen and uses the threads when i run lots of VST plugins etc (my old 2500K would get maxed out and cause audio drop outs and latency due to not being able to set the buffer size low). In Cinebench R15 multi-core it beats a 9900K and is only ~2% behind in single threaded, yet the 3700X is about 60% the price of the 9900k here in NZ, even after Intel dropping the prices. I got improved FPS in X4 and the computer feels a little more snappy. Building a clone of a vintage Marshall JMP 100 'Plexi' guitar amp will have to wait- yet again. I can sell the old CPU so the cost of the upgrade is reasonable. Upgrading the CPU while being able to stick with my Zen+ motherboard is icing on the cake.