Is there going to be a connection limit on this Test?
Unlikely, unless we get a volume of connections that could place the test in jeopardy
i still get a black screen when im undocking and delay which is irritating still
i got a queue connecting to the server, so there is some kind of limiter there. Was eventually able to get on though
Had the same thing happen when i tried to log another account on as well
How long is the test going to run for? What time will it end?
Duality does not appear on my Launcher today (nor yesterday).
User Error … I had not enabled Duality on my launcher
Looks like we got there, lol.
Initial INSTALL process is amazingly slow.
Low cpu, memory, disk and Network transfer speeds?
Fun to see Tidi again & get podded a few times 
My observations after mass test:
Client has memory leaks or multithreading issues (also this was also before x64)
- During fight fps dropped to ~5 and memory was at 4.7gb
- Warped out to asteroid belt (no other ships or npcs were there)
- Fps become ~7. Memory is still on 4.7gb. CPU utilization is 17% of 8c 16t processor (so it using full 3 threads at maxium for empty scene).
- After near 10 minutes memory slowly dropped to 4.2gb, fps rize to ~10. All this on empty asteroid belt.
- After about additional 10-15 minutes (sorry dont get exact moment) memory dropped to 2.5gb and fps get back to my 60 (capped). CPU utilization got back to 5%. (Which is somewhere near normal).
My assumption is that client is not freeing memory (at all or not fast enough) and still process tons of objects that already should be removed.
Also this behavior was also before introducing x64 client but due to crashing it was harder to observe.
1 Like
so to note I’m running a radeon 7 16 gb on onboard ram and 64gb of desktop ram
6.5gb used at most, with max graphics on, that said 4 frames per second.
submitted a bug report as it appeared I never left grid despite boing 37AU away
Now someone told me to boson not sure who but yea… no one is saying it was aimed poorly, but alas my bad sorry for nuking 40% of the online testers.
That seemed like a total mess today…
Hope you gathered some decent data, but i’ve got a hunch we’ll be doing this again.
Maybe the next one will run smoothly without all the children unable to follow directions causing problems.
On my end I had to run my client in brick mode - memory and CPU was staying almost maxed, GPU wasn’t even taxed like everything was bottle-necking at the CPU. The fleet fight was almost impossible for me to participate it. I don’t have the latest and greatest wet-ware machine but with the current 32 bit client I can run 30+FPS in high visual mode in all but the most complex mission environments and large scale fights. The previous fleet fight from the last test on Singularity was nice and smooth running on medium visual settings.
2 Likes
Just finished testing, followed all the steps of the test, in the end I was killed 
. But it was fun 
, this is my first event. So many ships and such huge, I have not seen. thank 
2 Likes
The test seemed to go pretty well on my end; client performance seemed decent even in TiDi and more responsive in general. Memory usage on one of my clients went up to 7.5GB, and 3.8GB on the other (my rig has 64GB).
It was very nice to not just see the client hit the 32bit memory limit then crash, like in some other TiDi fights I’ve been in 
Looks like you’re making some nice progress
Edit: Full System Specs:
Windows 10 Pro 1809
Core i7 7700k
64 GB Corsair Vengeance 3200 MHz DDR4
ZOTAC GeForce GTX 1080 Ti Mini
Samsung 860 EVO 500GB SSD (System Drive, EVE installed on this)
1 Like
@CCP_Goliath @CCP_Antiquarian i ran the mass tests with logging, what do i do with the data generated? i had a crash during the mass test not associated with the node death but due to max system memory exceeded that may be of use.
I’m actually really really worried. With minimal client settings (the proverbial ‘potato mode’) A mere 450ish person fight spiked my per client memory use up to 5.5 GB. Scale this to several more clients and 1000 people on grid and most desktop computers will simply run out of memory to allocate, with unpredictable results (yes I know some of you have like 512 GB of memory installed on your home pc and will be sure to inform us but the average is probably 16 or 32). Sadly that kind of thing is notoriously difficult to design a test for and we probably won’t actually know how the client will behave until a huge battle actually occurs.
it wont scale linearly, the game loads assets into ram to be utilised in the scene, not all duplicated models use the same amount of ram, but the more different ships there are and the higher the settings and especially all the skins on the ships would add more and more assets required into ram.
There is likely an issue with speed of prefetching assets and clearing assets not needed anymore but that comes with further optimisation that the 32bit client sorta blocked the issues from really arising.
I missed the tests itself but the 64 bit client does seem to run well.
Participated the test 1AM in the morning, CNTZ.
My hardware
AMD 1700x with Asus C6H motherboard.
Nvidia RTX 2080
2* 8G DDR4 RAM @ 3200
Samsung ssd 850 evo.
All settings were set to max for sure. 1080p @ 60 FPS, limited by my monitor.
(I have a double monitor setup but for EVE I use single)
if you don’t mind Chinese here is the link for the whole battle recording, I did streamed & recorded at the same time
It went very well in the first few minutes and then the FPS starts to drop.
I felt the modules are not as responsive as it should be, for eg some modules activated few seconds after it should (without tidi in local)
FPS was reduced to around 3 by the time I finally died.
RAM occupation was around 9GB. CPU did not heat up at all, GPU is very calm.
Warped out, tired full screen and windowed mode, the client is very slow to react, and did not help FPS.
Jumped a gate, everything’s fine again.
Hope this will help.