Load Times for Games

[Tycoonbrad]

I'm trying to figure out what upgrade would make games like Battlefield 1942 load faster. A friend of mine has the same AMD 2000+ XP CPU and the same amount of memory(512mb PC2100) as I do. Yet Battlefield loads the game quicker. He has a 20 gig Maxtor 7200rpm 512kb cache and I have an Western Digital 80 gig 8meg cache. Just based on my hard drive, I should slay him on drive throughput. We ran some Sandra tests on memory and CPU and our results are very similar. I think it could be two things: motherboard or video card. His motherboard is the new nforce2 ASUS boards. It could be that, but I think it's his 128mb DDR Nvidia 4200 card. I know this sounds like a "my-computer-is-faster-than-yours-is", and it is. But, I want to make sure the next time I go over to a LAN party, he notices my computer loads it faster. I know it's childish, but what the hell.

Both systems
AMD 2000XP
PC2100 512 Meg

My system
KR7A-Raid - mobo
WD 80 Meg 8mbCache - Hard Drive
ATI Radeon 64DDR VIVO

His system
ASUS Nforce2 (model#?)
Maxtor 20gig 512kb cache
Nvidia 128mb DDR 4200

 

[caboob]

Looks like it comes down to PCI bus throughput on the chipsets. Apparently the PCI bandwidth of his nForce board is greater than your VIA based chipset (don't know the figures specifically) which has historically been crippled by VIA's quirkiness and they have yet to address this problem. I remember owning a KT266A based board and was plagued by PCI issues such as slow apps loading and sound card compatibility issues (SBlives, audigy). Even if you are RAIDed, there are still latency issues within the VIA chipset that I remember people complaining about.

 

[DaveSimmons]

Get both 512 MB and a ti4200 to really make him envy you

 

[AgaBoogaBoo]

That ATI Radeon VIVO is holding the system down quite a bit, I think. Is that just a Radeon? Or like a Radeon 8500,9100, 9700, ect?

 

[Lord Evermore]

They both already have 512MB of memory.

Both the KT266A and the nForce2 chipsets use southbridges in which the IDE controller is not dependent on the PCI bus throughput, so I doubt that is the issue, as long as Tycoonbrad's hard drive is on the standard IDE ports and not the RAID ports. If it's on the RAID port, then it is passing over the PCI bus. However even VIA's issues with PCI (which haven't ALL been VIA's fault) shouldn't cause a huge noticeable difference in load time.

Hard drive cache also isn't going to make a huge difference here, since it's having to stream a large amount of data rather than being able to cache and burst transfer it. Of course I assume that the system just loads the map straight into memory; if it processes the map in bits and pieces from the files, then the extra cache could be put into use since the hard drive can cache the next parts of the map files before the CPU is ready for them.

The nForce2 does however have an 800MBps HyperTransport bus between the north and southbridge, while the VIA uses a 266MBps V-Link bus. Again though, the bandwidth shouldn't be an issue.

The load time of the game shouldn't be affected by the video card, since the load time is just the time for the map and other stuff to be put into main memory, not for the video card to load anything. Graphics memory is loaded as needed during actual gameplay, not ahead of time.

I think in all likelihood, the memory controller performance is what's coming into play. The KT266A is an older controller, as well as being a single-channel controller. The dual-channel controller of the nForce2 provides both higher memory bandwidth (which the processor may not take advantage of but DMA transfers from the hard drive can) and better latency performance due to the design of the controller. Of course, if you've each only got 1 module, then he's not in dual-channel mode, but the latency hiding of the nForce2 may still be in play. Even if it's not, it's just going to be a newer, better designed memory system.

 

[Insidious]

I think Lord E has it.

I'm on a KG7-Raid board (which is a bit dated) and was amazed at the difference between this
and even the KD7-E board of my friend. (remaining hardware of each machine is very similar)

I'm guessing you can blame the MoBo for a lot of it.

Another thing I would look into is the CDROM setups you each have. If you watch, during the loading
in BF1942 the CDROM is getting hammered.

You aren't sharing a ribbon with the HD and the CDROM are you... If so, get that optical drive off the IDE
controller that is handling your HD.

-Sid

edit:

This is the kind of info I keep finding...

__________________________

The next step is to attach the power and data cables to the new drive. The drive cable is connected according to the drive configuration and jumper settings. For a bare system install, find the IDE 1 connector and run a data cable from it to the new drive. If the drive is being added as a Slave device, then simply attach it to the secondary connector on the IDE cable. When using ATA 66 or 100 drives, use the appropriate 80-pin data cable, but remember that when mixing ATA 33, 66 and 100 drives on the same channel, overall drive speed will still be limited by the slowest drive.

__________________________

This particular article can be found Here at Sharky Extreme

 

[Lord Evermore]

Only older ATA33 chipsets had an issue with drives being reduced to the rating of the slowest drive. All current and recent chipsets have independent device timing.

Of course there is some performance isse due to the "one device active at a time" thing with ATA, so the hard drive has to wait while a slower optical device is transferring data (the delay of course being in the microsecond type of range, an eternity for a computer even though to us they'd seem to be working simultaneously). However the drives themselves would be able to function at their full speeds when they are actually active, so an ATA66 drive wouldn't be slowed down so long as an ATA33 or PIO modem CDROM drive wasn't reading. But of course, with the amount of time it takes a CDROM to spin up and transfer relatively small amounts of data, the hard drive may end up waiting a lot.

I personally see very little CDROM usage when loading maps in BF1942 (basically just the CD check), and really not a hammering of the drive during the initial game loading.

 

[Tycoonbrad]

...Of course, if you've each only got 1 module, then he's not in dual-channel mode, but the latency hiding of the nForce2 may still be in play. Even if it's not, it's just going to be a newer, better designed memory system...

Will you please explain dual Channel mode?

Come to think of it, I think he DOES have two memory modules. I have just one 512mb DDR and I think he has two 256mb DDR sticks. His Sandra memory becnhmark times were pretty close to mine. Hmmm....

 

[Lord Evermore]

Search the forums for previous threads, or look at reviews about nForce2 or Granite Bay boards. Dual channel memory means the memory controller accesses two modules at the same time, giving it effectively double the bandwidth. The Athlon frontside bus can't really make use of it, but other devices with direct memory access can use the memory without taking away bandwidth from the CPU.

 

[mechBgon]

Do you both have the same version of Windows? If not, that could have a lot to do with it too, particularly if you have Win98 or WinME and he has Win2000 or WinXP. Do make sure your drive is running in DMA mode if you're using Win98, or make sure the drive's controller is running in UDMA mode if you're using WinXP or Win2000 (does not apply if you're using the RAID controller's IDE channels instead of the board's native IDE, however). If you're not familiar with how to check into that, give a holler. Welcome to the forums

 

[Mavrick007]

Does he have anything overclocked, like ram or cpu? His memory timings shouldn't be that close to yours if he's loading faster, then perhaps it isn't on dual channel.

Does he play with the cd or use a no cd patch? Do you both have your display settings configured the same? I went to a Lan party on the weekend and my maps to UT2k3 loaded alot slower than the my friend's computer and mine was superior in every respect (GF Ti4400 vs. GF2 TI, 512megs ram at fastest timings vs. 256megs, 1900+ vs. 1700+, etc), but I ran my display at 1280x1024 at all high textures and high settings while he ran his at 800x600 and minimal settings. My display took almost a minute longer to load but once I was in the game I cruised along with all the eye candy.

 

[Insidious]

Originally posted by: Lord Evermore
Only older ATA33 chipsets had an issue with drives being reduced to the rating of the slowest drive. All current and recent chipsets have independent device timing.

Of course there is some performance isse due to the "one device active at a time" thing with ATA, so the hard drive has to wait while a slower optical device is transferring data (the delay of course being in the microsecond type of range, an eternity for a computer even though to us they'd seem to be working simultaneously). However the drives themselves would be able to function at their full speeds when they are actually active, so an ATA66 drive wouldn't be slowed down so long as an ATA33 or PIO modem CDROM drive wasn't reading. But of course, with the amount of time it takes a CDROM to spin up and transfer relatively small amounts of data, the hard drive may end up waiting a lot.

I personally see very little CDROM usage when loading maps in BF1942 (basically just the CD check), and really not a hammering of the drive during the initial game loading.

actually, it was the Sudhian website (KR7A-Raid) that gave me the link to the warning about combining optical drives and Hard Drives on a single controller.

Look in the "How should I connect my hard drive and other storage devices section" This article (Sudhian) was not addressing some older chipset, but the KR7A MoBo specifically.

-Sid

edit: for the CDROM usage, I was addressing the "long" time it takes for maps to change in multiplayer rather than the initial startup of the game.... alot of that initial stuff is just the game posting developer ads, a movie, etc... some of those segments are timed and you will not speed them up.

 

[Smilin]

Stuff 2 try.


Be sure HDD is on it's own ATA cable

Defrag.

Upgrade mobo.

Buy a 2nd HDD and make use of that raid controller

 

[Tycoonbrad]

It appears a new motherboard could be in order. If I do get another mobo, it looks like I need another memory module to take advantage of Hypertrasfer Bus technology. I'm going to read up on the new KT400a article on this site before making a decision on which board to get.

 

[Lord Evermore]

Sudhian is wrong, that was my point. Current chipsets have independent device timing, and will not set both devices to the same lower UDMA or PIO mode. If that's what they were referring to, then they need to bone up on current hardware specifications and stop depending on what they knew 5 years ago (and stop referring to 3 year old articles on other sites as reference for current hardware). If they were only referring to the performance loss due to only one device being active at a time on an IDE bus, then they are right, but the difference is probably unnoticeable in most situations.

 

[Insidious]

Originally posted by: Lord Evermore
Sudhian is wrong, that was my point. Current chipsets have independent device timing, and will not set both devices to the same lower UDMA or PIO mode. If that's what they were referring to, then they need to bone up on current hardware specifications and stop depending on what they knew 5 years ago (and stop referring to 3 year old articles on other sites as reference for current hardware). If they were only referring to the performance loss due to only one device being active at a time on an IDE bus, then they are right, but the difference is probably unnoticeable in most situations.

That's good to know.... Thanks Lord E!

This is why I enjoy AT.... I don't have the savvy to separate the News from the BS

-sid

 

[human2k]

good god, original Radeon DDR running BF1942! What resolution you run at ? I consider Radeon 8500 mininmum for playing BF1942. When I had 256mb of ram in my rig, BF1942 took forever too load (30sec-1min), with a 512mb, it cut the time in half. Get another 512mb!

 

[vash]

Load times, for games, boil down to basically:

1. How fast your disk can access the data
2. How unfragmented the drive is.

To help improve the load times, you coul make sure to defrag your drive first before playing or increase the throughput of the I/O on the system with RAID. For my next gaming rid, I will be going RAID-0 (cheapo drives, not large or significant data storage) for the gaming drive. The faster your subsystem can transfer the data, the faster your levels will load.

Also, that video card of your isn't going to cut it for BF1942 action. This game is DX8 based, you need at least a card that can handle that. I would be willing to bet with 512+ megs of ram and a faster video card, your friend would be itching for an upgrade too.

vash

 

[chizow]

Originally posted by: human2k
good god, original Radeon DDR running BF1942! What resolution you run at ? I consider Radeon 8500 mininmum for playing BF1942. When I had 256mb of ram in my rig, BF1942 took forever too load (30sec-1min), with a 512mb, it cut the time in half. Get another 512mb!

Heheh yep, RAID 0 + 1 gig of RAM for BF1942 = Choice of any vehicle/plane/stationary defense with time to spare to set-up shop at enemy spawns. If you ever wondered how half the map is capped and you're getting shelled by artillery the first time you spawn, the other players have more RAM and/or SCSI/RAID 0.

Chiz

 

[Lord Evermore]

Or you're a crappy player with slow reflexes.

 

[StraightPipe]

I had 256 DDR RAM
Upgraded to 768

Battlefield loads much, much faster.

 

[Lord Evermore]

Comparing performance from 256MB to higher amounts of memory is pointless. Big games like lots of memory, and it's well known that going higher than 256MB is going to make them run better. Aside from the common knowledge of that, both of these guys have 512MB of memory, so saying that someone with 256MB gets better performance by upgrading is unrelated.

I stuck an extra 256MB in my system to see what happens, and I don't notice a significant difference in load times at 768MB. However with 512MB I was using dual-channel mode on the nForce2, and I don't know what sort of impact that had. Load times certainly never seemed shorter compared to a similar system without dual-channel.