Game Load Time

[Doomtrooper]

I simply pointed out there is a difference with ATA 133.

 

[Doomtrooper]

IMO all three: CPU, RAM, HD.

Would you imagine that going from 512 to 768 decreased loading time in CoD by 5-8 seconds (which is quite a lot)

Why yes I do

 

[KimB]

I simply pointed out there is a difference with ATA 133.

Except there isn't any significant difference. You showed a whopping 10-15% difference in burst transfer rates. Burst transfer speeds do not determine the typical speed of most hard drives. It's the sustained read/write speeds that really matter, since the only time you're going to be waiting on your hard drive is when there is a lot of data being transferred.

And regardless, modern OS's frequently make hard drive caches almost obsolete since hard drive information is cached in system RAM anyway.

 

[Doomtrooper]

I disagree, lets leave it at that

 

[Hyp-X]

And regardless, modern OS's frequently make hard drive caches almost obsolete since hard drive information is cached in system RAM anyway.

OS's don't make the drive cache obsolete, since that can store data that was never transferred to the CPU in the first place.

You know that access time = seek time + turn(?) time.
The latter is spend with the head sitting over the right track waiting for the requested sector. All sectors encountered during that time can be freely stored in the cache.

If the application reads relatively large amounts of data this sectors are the ones that it will require after getting the requested one.
Also because the data is read in different order than it's transferred, buffered transfer speed it also important.

 

[tobbe]

I simply pointed out there is a difference with ATA 133.

It's the sustained read/write speeds that really matter, since the only time you're going to be waiting on your hard drive is when there is a lot of data being transferred.

Under Windows I'd say access time is much more important for most applications. Or did you mean app load times specifically?

And regardless, modern OS's frequently make hard drive caches almost obsolete since hard drive information is cached in system RAM anyway.

Since most ppl are running Windows anyway that point is moot

 

[KimB]

If the application reads relatively large amounts of data this sectors are the ones that it will require after getting the requested one.
Also because the data is read in different order than it's transferred, buffered transfer speed it also important.

Well, obviously hard drive cache does improve performance somewhat, but its effect is nowhere near as high as that of the ability to actually read data off the disk.

And remember that when loading a level, much of the data that is read is read in a sequential order (particularly if you've kept your hard drive defragged enough), and in large chunks. This means that the limiting factor will typically not be access time or maximum transfer speed, but rather the actual speed the hard drive is reading off of the disk.

After all, the largest caches on hard drives today are about 8MB. My current hard drive can read 8MB of data straight off the disk in 1/5th of a second. So, when you're talking about delay times from loading, the cache is going to be a pretty small contributing factor.

 

[Bambers]

System ram can make a huge difference, more than a lot of you seem to be giving it credit for.

I've seen my PC loading desert combat (0.39k i think it was) el alamane level next to a friends.

My spec:
XP2000+
soltek kt333 based mobo
1x 512mb pc2700
128mb 9800pro
60gb maxtor ata133 5400rpm in a relatively defragged state not running in quiet mode

his spec:
athlon tbird 1300MHz
some mobo probably k133 based
1GB pc133
gf2 pro
a hdd from the same time as the rest of that and knowing him it was probably going to register at 90+% file fragmentation.

Both running XP pro his DC loading time was almost twice as quick as mine. Of course once we were in game he was struggling at 800x600 in low detail while I was running at a nice 1280x960 4xAA 16xAF 8).

The reason for this is fairly obvious when you alt tab out of bf1942 and have a look at taskmanager. XP chews up about 120mb of ram on my system (about 180 normally but it lets some of that go) the remaining 380MB are completely used by bf1942. 512MB is barely enough for some games today.

 

[ET]

Or maybe system memory is a factor (I have 256MB).

256MB is quite a bare minimum these days. I know that UT seems to eat any memory you throw at it. You might want to consider upgrading to 512MB at least.

 

[Grall]

True, but ATA 133 drives are faster than the same drive with a ATA 100 interface, in fact some ATA100/ATA 133 drives perform as good as Serial ATA drives.
I've tested this myself with a utility called 'HDTACH'

Minor differences in burst speed is completely irrelevant in the real world. Even big jumps barely register when it's seek time and rotational latency that are the two biggest performance factors (RAM access speed measured in nanoseconds, harddrive access speed in milliseconds; orders of magnitudes in difference), and after that, cache and firmware optimizations.

HDtach's burst speed test is mostly for trivia use, why do you think Storagereview does NOT use it?

*G*

 

[NeoCool]

The amount of RAM you have in your computer is a HUGE signficant imperative playing factor in loading times in games on your computer, along with your CPU, HD, and OS, as well.

 

[Tim Murray]

IMO all three: CPU, RAM, HD.

Would you imagine that going from 512 to 768 decreased loading time in CoD by 5-8 seconds (which is quite a lot)

Course it did. Windows is such a memory whore that having less than a gig now on an enthusiast machine (or an AV machine, or something along those lines) is silly.

 

[KimB]

256MB is quite a bare minimum these days. I know that UT seems to eat any memory you throw at it. You might want to consider upgrading to 512MB at least.

Yes. 256MB is quite small today.

Everquest, for example, has horendous load times if you have only 256MB of RAM (we're talking a few minutes to change zones). With 512MB, EQ almost stops swapping...

 

[Doomtrooper]

Minor differences in burst speed is completely irrelevant in the real world. Even big jumps barely register when it's seek time and rotational latency that are the two biggest performance factors (RAM access speed measured in nanoseconds, harddrive access speed in milliseconds; orders of magnitudes in difference), and after that, cache and firmware optimizations.

HDtach's burst speed test is mostly for trivia use, why do you think Storagereview does NOT use it?

*G*

That is not true, as hardrives cache increases burst speed is very important as the cache information is transferred much faster then reading from the disk.

If the program, map, game is 8 megs or smaller it will load faster on a a 8 meg cache drive..and as you can see cache sizes continue to grow.
Does it mean alot for loading huge files, not really...but there is a 10 mb/s difference in the ATA 100/ATA 133 formats...and that is what I wanted to point out.

I stand by my opinion that RAM an CPU speed offer the best improvement for Windows XP, but having a ATA 133 drive isn't 'hurting' anything either.
Plus as I said before I prefer Maxtor Drives...so does most of the Satellite companies offering PVR (personal video recorder) units, they all come with Maxtor Drives.

 

[Grall]

That is not true, as hardrives cache increases burst speed is very important as the cache information is transferred much faster then reading from the disk.

Actually, what *you* say is not true. Load times are limited by the access speed and sequential transfer rate of the drive, not burst rate, and no matter how fast you make burst rate you still won't see any improvement just because of this. Even greatly improving sequential transfer rate barely registers in overall performance figures simply because most tasks don't consist of shuffling very large quantities of data at a time.

If the program, map, game is 8 megs or smaller it will load faster on a a 8 meg cache drive

No it won't. I/O is almost never done by grabbing an 8MB chunk of level off a drive, instead you'll have lots of seeks reading tons of separate textures, bsp tree, sounds etc usually from one rather large pk3-style archive. There'll be lots of seeks and lots of rather small I/Os.

Anyway, cache size doesn't matter at all if you're just reading an arbitrarily sized file off the disk with zero queue depth ( = no other outstanding I/O requests). You could just as well have no cache at all and you'd never notice the difference until you need the exact same data chunk one more time (and then you'd probably find the OS cached the data anyway and never bothered to send any requests at all to the harddrive).

but there is a 10 mb/s difference in the ATA 100/ATA 133 formats...and that is what I wanted to point out.

But that difference means almost no performance difference in real-world tests! Storagereview has written whole articles about this, it's pointless to argue against it. You've already been proven wrong before you even posted!

> but having a ATA 133 drive isn't 'hurting' anything either.

It's not 'hurting', but when it comes to gains, this is a stupid gain to go for since other factors have far greater impact on loadtimes than the interface speed. The Hitachi Deskstar 7k250 is AFAIR the fastest 7200RPM IDE drive bar none right now, guess what? It's got an ATA100 interface! (Unless it is the SATA version of course.)

Also, previous leaders were all WD drives, not Maxtors with ATA133 interfaces. As you can see, your argument is pointless. Give up already will ya!


*G*

 

[Doomtrooper]

Hmmm...I will throw in the towel on this one.

 

[rwolf]

The amount of RAM you have in your computer is a HUGE signficant imperative playing factor in loading times in games on your computer, along with your CPU, HD, and OS, as well.

I think RAM is the big problem. Warcraft takes 127MB by itself and the system is loading it in only to swap it out again. I recall someone saying that hl2 needed 1GB to run smoothly.

Perhaps AMD is on the right track and 64bit addressable memory is the way of the future.

 

[jandar]

HDD: - take a single ATA HDD and compare it to a RAID 0 HDD setup. Better yet, compare that to a RAID 0 SCSI setup.
When my brother and I had identical machines, his RAID 0 setup loaded UT 2-3 times faster than my single HDD did.


CPU: - makes a difference in loading the game. Many background apps and services can slow things down.


RAM: - makes a difference in 128MB versus 256MB versus 512MB. 1GB does help, but a clean running system with 512MB can outperform a 1GB system.