Jump to content

Analysis of RT senario loading times


Recommended Posts

Hey all – may get moved to tech support, but wanted as much exposure/input as I could get, so perhaps Phil will leave this out in the main section for a couple days…

I purchased a Lenovo IdeaPad Y410P laptop last year, and was really encouraged by its Combat Mission load speeds; so I built my new tower this year around the latest Intel components to emulate the performance. But I’ve been surprised by the numbers.

So - 3 different computer systems, tested for loading times of an RT action turn during WEGO vs. computer play. Test #1: Myth of Invincibility (13.2 MB) mid-battle. (I originally kept this turn to show my brother an SU-76 getting KO-ed by another SU-76 cooking off.) Test #2: Augustow Plague Boil (26.2 MB) action turn, done after Old Tower was converted to New Tower.

Mods folder (a ton of Aris & other mods) is standardized across all three platforms. All 3 use the same ESET NOD32 antivirus software. All units were freshly booted for a couple minutes before starting CM:RT. Samsung Magician SSD software in play (optimized for max speed) with HD over-provisioning in place for the Sammy SSD’s.

Old Tower: Phenom II 965 BE, stock 3.4 GHz/OC 3.7 GHz, Mushkin Chronos 240 GB SSD, 8 GB 1866 DDR 3, AMD 7870HD GPU 2GB, Win 7 Pro 64, 1080p.

Laptop: Intel i7 4700QM, 3.4 GHz stock, Samsung 840 256 GB SSD, 8 GB 1600 DDR3, Nvidia G750M GPU 2 GB, Win 7 Home 64. Single screen mode via HDMI to 1080p monitor.

New Tower: Intel 4790K, 4.0 GHz stock/OC 4.4 GHz, Samsung Evo 500 GB SSD, 16 GB 2400 DDR3, AMD 7870HD GPU 2GB, Win 8.1 64, 1080p.

Load Times:

Unit ----- Test #1/Best Balanced Test #2/Best Balanced

Old Tower - 5:07 1:39 N/A N/A

Laptop - 1:13 0:28 0:52 0:25

New Tower - 3:32 1:05 2:06 0:43

Of Note: Removing the entire Mods folder from each of the builds only gains you 4-5% loading time. I thought it would be more.

So, my new 2nd-gen Haswell tower, which I was planning on using to assume control of the Free World, gets absolutely spanked in scenario load times by a 1st-gen Haswell (mobile) laptop. Now, once in the 3-D environment, the New Tower shines – the 4790K combined with a more powerful 7870 video card clearly renders out the battlefield faster, no question. I am aware that (the last time I checked) OpenGL favors Win 7 over Win 8 by a noticeable margin. But the difference in load times, while better than the previous Tower, is massive.

Comments? Could the Samsung 840 (non-Pro, mind you) really be doing something that the Evo isn’t? Using the Magician bench mark, the 840 does have better read scores, but only somewhat (say 10%). Look, it was time to refresh the tower anyway, and I’ll get a new video card at some point – I’m not disappointed with the overall performance. But this scenario load-time thing is really mystifying.

Build notes:

The Lenovo’s original Win 8 1 TB HDD was not clone-able (proprietary format); so it was re-built with the Samsung 840 SSD & Win 7 64-bit from scratch.

Using Magician’s RAPID mode only nets you 4-5 sec. on a fresh load.

Link to comment
Share on other sites

  • Replies 64
  • Created
  • Last Reply

Top Posters In This Topic

Load Times on Hackintosh.

Both Tests run in Best/Best, shadows, shaders, etc. on. Maximum eyecandy settings. Z Folder = 3.22 GB

Myth of Invincibility (7.9 MB) = 14 seconds

Augustow Plague Boil (24.7 MB) = 21 seconds

I do not know what is holding up your load times on the new tower.

"Now, once in the 3-D environment, the New Tower shines – the 4790K combined with a more powerful 7870 video card clearly renders out the battlefield faster, no question." This is a good thing and reason to upgrade for sure.

My Hackintosh has me plenty of vigor for anything CM related I toss at it. Apple's OpenGL screw up is another story :rolleyes:

Phil and Battlefront have CMRT running in stunning war game glory :D

Good Luck with dropping the tower load times Hpt. Lisse .

System Specs:

Hackintosh - ‘EVO Black Resurrection’ = EBR

Intel Core i7-4770K - 3.89GHz not over-clocked in OS X

Gigabyte Z87X-UD3H motherboard

EVGA GTX 760 ACX 4GB GDDR5 RAM

16GB G.SKILL Sniper Series DDR3 2400 memory at 1600 MHz in OS X

Intel 335 SSD 240GB

Cooler Master HAF XB EVO LAN Case

DELL U2713HM-2560 x 1440

Apple OS X 10.9.4 (13E28)

Boot ROM Version: MultiBeast.tonymacx86.com

Link to comment
Share on other sites

interesting, I just built a new system: i5-4670k, AMD R9 290, ASUS Z87 mb, 16 gb DDR3, Adata SP920 SSD 256 gb, win 7 64 bit and I have load times similar to Hpt. Lisse.

From the ResMon, I have one CPU running at close to 90% while loading, so I assumed the CPU was the bottleneck, but the loading times Buzz posted would point to a software issue.

Anyone else getting fast load times like Buzz using Win 7? I would like to know what to tweak. I have used the Adata tweaking software, but it did not improve load times.

All my other games show vastly improved loading times, "Rise of Flight" used to take up to 2 minutes to load and it now takes less than 15 seconds. CMx2 is the only one that is an issue.

Link to comment
Share on other sites

I'm the farthest thing from a tech geek, but I believe that with CMRT additional computations have been front-loaded into the initial load sequence in a (successful) effort to speed up realtime play. So any wait time you gain at startup is more than made up during gameplay in higher framerate and reduced lag.

Apparently map size makes a significant difference too. Maps keep growing bigger and bigger with each new title. Originally BFC planned to greatly expand max map size for CMRT but found the load time increase exponentially to the point of diminishing returns. If an extra km map equaled an extra 5+ minutes startup wait that was just too high a tradeoff.

Link to comment
Share on other sites

I have experienced what I consider to be very slow load times in all three games and have noted this in other threads on the same topic. My system should be able to smoke through the load times

My system specifications:

  • ASUS P9X79 Deluxe (Sandy Bridge-E) with Intel i7 3930K @4.2 Ghz
  • 64 GB DDR3 RAM @1685 MHz
  • ASUS HD6970 Direct CuII (2GB DDR5 RAM @890 Mhz/1375 MHz)
  • Storage - OCZ RevoDrive (OS) and various SSDs
  • Screen Resolution - 2560x1440 (this is what game is run at)
  • Win 7 x64 Ultimate

Link to comment
Share on other sites

What MikeyD states ..."CMRT additional computations have been front-loaded into the initial load sequence in a (successful) effort to speed up realtime play. So any wait time you gain at startup is more than made up during gameplay in higher framerate and reduced lag." .... makes sense to me.

I am much further down the tech geek totem pole than MikeyD is for sure :) CMRT feels really fast and smooth on the Hackintosh. The CMBN/FI scenarios I can play (despite Apple's OpenGL mistake) load and play well but it "feels" like CMRT is faster and smoother ... just zings along loading, saving, computing turns, etc. I usually play Real Time vs AI. The load times I noted for Hpt. Lisse on Hackintosh were done as Elite Turn based saves.

Checked Gog and Magog scenario largest stock CMRT scenario.

Gog and Magog loads in 18 seconds.

I had previously tested the same scenario in 10.9.3

Gog and Magog loads in 22 seconds.

Maybe Apple polished something in 10.9.4 to improve graphics?

*Looking forward to CMBN/FI Patch & Upgrades coming soon I hope*

Link to comment
Share on other sites

So what are the ideal machine specs than can speed up load times and also provide the best in-game play experience?

I had thought that assuming one had a decent motherboard and video card it would simply be a question of having an SSD plus a lot of RAM. However, the above discussion makes it unclear.

I know that CM only benefits from single core, so does having multiple cores actually slow the game down, or are multiple cores not the issue any more.

Link to comment
Share on other sites

I do not know the “ideal machine specs”. I have learned from building my 1st computer, my hackintosh.

1st was a powerful GPU makes for a much smoother, fluid 3D game experience. This is probably a more consistent “frame rate” but I don’t have a FPS counter. Graphic settings can be set to best / most eye candy and CM looks great.

I moved my GPU out of the older Mac Pro into the new hackintosh in my build. The new motherboard, CPU, memory, etc. in this build made a huge positive impact on CM’s load and times. Everything in CMRT is fast with other programs running at the same time.

I would not think having multiple cores is a problem for CM. If the game has enough GPU power, enough memory, increasing your CPU power should speed up the game loading and saving. This is what appears to have happened in my hackintosh. Having multiple cores allows me the headroom to have other programs running and still have a great CM experience. I have even had all 3 games CMBN/FI/RT playing scenarios at the same time w/o a problem!

The 36% Apple OpenGL load problem is a very good example of how driver level malfunction can impact the game. No amount of hardware can fix that problem. Phil Culliton resolved the Apple OpenGL issue in Red Thunder with an an explicit workaround that makes CMRT Mac a true joy to play.

Link to comment
Share on other sites

IIRC Steve said that they now assume that everyone at least has a dual core machine and that they are starting to use a second thread for some things during the loading screen.

Subjectively the loading times went up with 2.12 (and the game itself got faster) and went down again with 3.0 (when the second thread was used).

Link to comment
Share on other sites

The game is a joy to play without working around Apple :D

A major reason I built my EVO Hackintosh Ian :D I have not ruled out buying CMPC games as well. Buying into the "Evil OS" ;) could feed my CM "Recreational" Addiction when the ZED Wars arrive:) Currently having a blast playing CMRT Mac. ChrisND posted patches and upgrades are in testing so... I can wait... sort of.

Link to comment
Share on other sites

Ok, so it will be useful to get a multicore (quad)... that's what I was wondering about.

My opinion....

I don't think a multi-core (quad) CPU would hurt your CM performance. The cost difference relative to the # hours daily serious CM sportsman use their CPU would be minimal. A multi-core (quad) CPU is a wise prchase.

See Below

-------

We are certainly working on this. As you guys say, by now it's pretty safe to cut off support for people without at least 2 cores. So we are exploring ways to allow limited use of other cores when it offers a big payback in performance. For example, we have identified a specific part of game loading that typically maxes out a single core. By rewriting just that portion of the code we can speed up the loading process significantly and quite "safely".

Extending multicore support to runtime game elements is a lot less "safe". It has to be done carefully and tested like crazy. But there are definitely things we have in our crosshairs. A couple of things having access to other cores could make a big difference.

One thing we will not do is rewrite the entire game engine to be compliant with multicores. This is absolutely not viable for us. It's a massive code base and we'd basically have to stop all work on everything for a very long period of time. And for the most part it won't do much to speed up the game more than the selective approach we're exploring.

If any of you guys lived through the transition of the old Motorola based Mac processors to RISC PowerPC the above will sound familiar to you. For several years the MacOS, and most of the programs that ran on it, were selectively optimized for the newer architecture. Things got progressively faster without a huge gap of time when there was no progress at all.

In general our approach to CM improvements is like this.

Steve

------

Link to comment
Share on other sites

Yep, confusing stuff - the only difference I see is New Tower SSD is a Sammy Evo and Win 8.1 64; the Laptop an 840 and Win 7 64. The tower outclocks it by a full GHz, to say nothing of its full-featured Asus Z97 Pro mobo, faster RAM, etc. I'll throw a 3DMark 06 at both after I feed the ankle-biters here, and post it... and may try to post this stuff in a Samsung forum to see if anyone bites...

Link to comment
Share on other sites

I vaguely recall someone was attempting to assemble a multi-km map with 3-4 battalions on each side. I believe in that instance load time was very long indeed. :)

That's the thing about benchmarks and optimals and such. If you're working well within the game's parameters then then fast-slow-high-low makes remarkably little difference. If you're attempting things at the outer envelope of the game's capabilities then things get problematic rather quickly. A battalion+ on a larger forested map is a battalion+ on a large no matter what CPU you got.

Link to comment
Share on other sites

So, 3DMark 11 rates my laptop at P2880 and my New Tower at P8600 (performance test, at 720p). Individual scores were (graphics - physics - combined) as follows:

Laptop 2629 - 7691 - 2361

New Tower 8254 - 11316 - 8233

I wonder if, being that the laptop SSD is plugged directly into the mobo affects anything (but it still goes through an SATA controller, right? Same as the tower through a cable & SATA port?) Haven't posted at Samsung today, but will later...

Link to comment
Share on other sites

Buzz et al: I'll throw another piece of evidence into the pile - I kept this out earlier as to not confuse things...

When I initially set up the New Tower, I was using the AsusMedia SATA port to do so - the four Intel SATA ports wouldn't recognize the SSD. Predictably, the Magician benchmark was pretty bad - so viola, right? That's why Test Battle #1 was loading so slow (3:37). Turns out, the mobo CD (Asus Z97 Pro) didn't load a current driver for the Intel SATA ports. Great - find driver, manually install it (kinda archaic for 2014/Win 8.1 era, but...) and sure enough, our SSD benchmarks connected to the Intel port is up to snuff. Sequential Read (MB/sec) rises from 314 to 505; Random Read (IOPS) similarly jumps from 53314 to about 97000.

But that only dropped the loading time by 5 seconds (3:32).

Yes, I understand that programs like Samsung Magician are synthetic benchmarks, and may not directly correlate to how fast a saved RT battle loads... but this is just damn peculiar now. Either there is some major differences on how a Samsung 840 & 840 Evo run data queue depths; the manner SSD's are handled between Win 7 and Win 8.1 is different; or some mobo setting in the tower is off.

I will get to the bottom of this. One of my brothers has offered to buy the Evo off of me if I want to get a brand-new Samsung 850 Pro, but that's a $400 experiment (minus whatever he'd pony up for the Evo...)

We still have a huge (3:32 vs. 1:13) loading disparity between Tower of Power and its lesser laptop brethren. There has to some reason for it.

Link to comment
Share on other sites

“The four Intel SATA ports wouldn't recognize the SSD.”

Did you check for latest Asus Z97 Pro Intel SATA drivers from Asus support?

GPU Drivers are up to date?

Another shot in the dark is check BIOS to assure disk drive numbers correspond to the SATA channel numbers.

Samsung 840 & 840 Evo are considered very good SSDs. If everything else is loading fast it should not not be a bad SSD.

Sorry I can’t help more as I am a novice hackintosh guy

Good Luck.

Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Unfortunately, your content contains terms that we do not allow. Please edit your content to remove the highlighted words below.
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.


×
×
  • Create New...