Improving ftb load time

[zemerick]

A bit late for most of the people, but for any future reference: If you have any of Reikas v25z mods in a 1.6.4 pack you need to add his URL fix: http://www.mediafire.com/download/jtefpa1cxgnp5p1/URLFix_for_v25z.jar

The version checker changed URLs, and was taking a very long time to timeout, forcing the load to wait for it. That mod fixes this problem.

 

[ChemE]

No what I've never understood is why can the executable with all items and textures not just be built once each time any mod or config updates and if nothing has changed since the last startup just load the old executable. We don't download the whole modpack every time we start modded minecraft, so why build the house from individual bricks each time we come home?

 

[Hambeau]

I've been trying Direwolf 1.7.10 recently and it takes an average of 6 minutes to load which is an increase of about 3 minutes from the DW 1.6.4 mod pack. If this is the trend I'm not very enthusiastic about it. I mean, where does it end? Will it not be long when the load time is 10 minutes? I found it irritating when I was waiting 2.5 to 3 minutes for Monster & DW 1.6.4 to load and now it's just ridiculous.

I have a pretty good PC and it has a SSD drive so the SSD drive solution is not the thing. The best solution for me is to reduce mods from the pack especially since I only play SSP and don't need most of the mods in the pack. However, trying this is not always an easy thing. Sometimes removing a mod will cause errors which will cause a fail to load. We are not all mod pack and error log gurus that can sift through the jungle of messages and decipher what the actual problems are.

I am curious what the load times are for other folks with DW 1.7 and Infinity 1.7.

To be sure I was only timing the pack load I started with the client already running and timed between the launch button click and the appearance of the Minecraft screen.

I checked my load time using DW20 pack version 1.1.1 and Infinity version 1.2.1

DW20 = 3 minutes 7 seconds
Infinity = 2 minutes 32 seconds

My system:
----------
CPU: AMD FX8350 @ 4Ghz 8-core (may have an effect since the client does call OS functions which support multiple cores)
RAM: 8GB 1666Mhz DDR3
HD: 1TB 10k RPM
GPU: NVidia 9600GTX 512MB dedicated GDDR3 ram

Windows 8.1 Pro, 64 bit
Java 1.8.0_25, 64 bit

Keep in mind that in addition to loading the pack the client is also having to authenticate your login and does so through a separate program that has to access the Mojang server(s) and be authenticated itself as opposed to automatically authenticating as when you log onto Mojang's launcher... You should also take into account your internet connection speed as well.

I'm on Comcast and my current local speed, according to Speedtest ( www.speedtest.net )

is 124.80Mb Download and 12.18 Upload

 

[zemerick]

To be sure I was only timing the pack load I started with the client already running and timed between the launch button click and the appearance of the Minecraft screen.

I checked my load time using DW20 pack version 1.1.1 and Infinity version 1.2.1

DW20 = 3 minutes 7 seconds
Infinity = 2 minutes 32 seconds

My system:
----------
CPU: AMD FX8350 @ 4Ghz 8-core (may have an effect since the client does call OS functions which support multiple cores)
RAM: 8GB 1666Mhz DDR3
HD: 1TB 10k RPM
GPU: NVidia 9600GTX 512MB dedicated GDDR3 ram

Windows 8.1 Pro, 64 bit
Java 1.8.0_25, 16 bit

Keep in mind that in addition to loading the pack the client is also having to authenticate your login and does so through a separate program that has to access the Mojang server(s) and be authenticated itself as opposed to automatically authenticating as when you log onto Mojang's launcher... You should also take into account your internet connection speed as well.

I'm on Comcast and my current local speed, according to Speedtest ( www.speedtest.net )

is 124.80Mb Download and 12.18 Upload

Most of your speed probably comes from the 4ghz of the CPU. Other things that do help though are windows 8 and java 8 ( pretty sure you meant 64-bit there too...not 16bit )

Download and upload speed are pretty irrelevant to the authentication. You're talking about bytes of data. Ping will affect it, but is still likely quite insignificant. Milliseconds out of the minutes.

 

[Artea]

I have two SSDs in raid-0 and infinity still takes around 3 minutes to load, including texturepacks.

 

[Celestialphoenix]

I've heard a chunk of it- certainly for bigger mods- is due to Forge being somewhat poorly optimised with how it loads/handles recipes. This is fixable, but would require a significant investment of dev hours.
(source: Gregtech dev thread)

 

[Hambeau]

Most of your speed probably comes from the 4ghz of the CPU. Other things that do help though are windows 8 and java 8 ( pretty sure you meant 64-bit there too...not 16bit )

Download and upload speed are pretty irrelevant to the authentication. You're talking about bytes of data. Ping will affect it, but is still likely quite insignificant. Milliseconds out of the minutes.

I happen to know how remote authentication code works, having written a few in my years as a network administrator... It does take time... In fact, you can see it by how long the authentication line, "creating YggdrasilAuthenticationService" remains visible in the console before the pack actually starts loading.

This is a separate program that has to load, launch, contact Mojang's AuthSeerver and wait for the token that says you are a legitimate Minecraft owner before it even starts loading your pack.

The time between the appearance of that line and the rest of the log appearing is what I'm talking about. On bad network days I've seen that take up to 30 seconds to complete. I probably was running at 50down/8 up at that time, I think they recently improved my service.

Yes, it may just be a matter of bytes, but that has nothing to do with network transmission speed. I had to prove to an Engineer once that a 10Mhz DOS PC was outperforming a 25 Mhz Sun Unix workstation because the PC only does one thing at a time and actually sent 960 characters/sec (at 9600 baud) while the Sun, being Multi-user was only sending 9600 baud for maybe 25% of each second while that user was being processed, thus allowing a buffer to clear while the PC was overflowing it.

Thanks for catching the 16 bit What can I say? it was late, my eyes were bleary, yadda yadda