Getting 4GB, what about more?

Obviously there will still be some 'level-load' involved :rolleyes: However it will be massively reduced - the i-ram is a rubbish indicator of performance when it comes to speed compared to Ramdrive based solely on ram (the pci slot limits its speed!)

Ramdrive:
ramdrive.jpg


I-Ram:
hdtune-avg.gif


I-Rams were actually pretty rubbish at tramsfer (Pci bus limitations) - a good Raid 0 leaves them standing! Safe to say, with a storage solution running 20x faster you will see a dramatic improvement in performance in tasks reliant on storage speeds (like loading levels)
 
Transfer speeds don't improve game loading times times. Hence why raid 0 gives no improvement over a single drive and can even increase load times. It's the latency that gives the boost. So the limited bandwidth is hardly a limiting factor of the i-rams gaming performance.

http://www.overclockers.com/articles1063/index02.asp

http://www.anandtech.com/storage/showdoc.aspx?i=2101&p=10

http://techreport.com/articles.x/9124/6

I would love to be proved wrong, but alas, I think it is a long time before I will see these massivley reduced game loading times you are talking about.
 
Last edited:
Transfer speeds don't improve game loading times times. Hence why raid 0 gives no improvement over a single drive. It's the latency that gives the boost. So the limited bandwidth is hardly a limiting factor of the i-rams gaming performance.

:eek:

That's going in the scrap-book - thanks - always appreciate a good laugh! :D

I know why you think this - pretty sure you saw this:
time-load-farcry.gif


TBH there is something seriously wrong there - if they have gone from a single drive (caviar) to a 4 disk Raid 0 with no improvement they've hooped it up somewhere! (Or farcry is badly fragmented/poorly set out)
 
Last edited:
TBH there is something seriously wrong there - if they have gone from a single drive (caviar) to a 4 disk Raid 0 with no improvement they've hooped it up somewhere! (Or farcry is badly fragmented/poorly set out)

As you can see from the other sources I posted. Raid 0 commonly gives no improvement in game loading. I'm still waiting for some proof of these amazing load times you speak of.
 
Why have we been side-tracked by Raid 0? We were discussing improvements running a Ramdisk - access time <0.1ms. It may well be that the speed increase I have seen is because of the low access time of my scsi array (3.5ms) rather than their array as your links would indicate (still not convinced for BF2 thought ;) ) But that's irrelevant when you're talking about a Ramdrive build entirely out of system memory - the load times will be blisteringly quick as the transfer speed is tiny and the latency is minimal ;)
 
I-Rams were actually pretty rubbish at tramsfer (Pci bus limitations) - a good Raid 0 leaves them standing! Safe to say, with a storage solution running 20x faster you will see a dramatic improvement in performance in tasks reliant on storage speeds (like loading levels)

The IRAM takes power from the PCI bus and nothing else. All data is over SATA150.

For access times they cannot be beaten by another a SATA device, they are orders of magnitude faster than spinning disks for latency, and are capable of sustaining that 120MBps transfer, unlike a RAID setup which will fall off towards the edge of the disk and be severely affected by fragmentation.
 
You're absolutely right about the pci bus being used only for power, forgot about that one, but the device is still limited by it's Sata 1 interface (150mb/s) I guess it's running about the same access time as straight ram (0.1ms) so it would be interesting to see if it does make a difference.

Found this on another thread though:
12disk8mbnw3.png


:eek:

Go on Mi9chy - take one for the team and try it out on your rig - get some benchies done and end this pointless debate!
 
Why have we been side-tracked by Raid 0? We were discussing improvements running a Ramdisk - access time <0.1ms.

Which as I have shown with i-ram benchmarks, reduced the loading times by a few seconds, but nothing significant. Until I see otherwise I will assume a ramdrive gives the same performance, due to the latency being almost the same.

In any case, I wouldn't reccomend the op buying 4GB of ram without any evidence of the supposed benefits.
 
Last edited:
Back
Top Bottom