Promise TX4 only 8.32 MB/s in Digiutils

[dolonescu]

I just installed a Fasttrack TX4 on my brandnew Tyan Tiger MPX. When testing it with NERO and other benchmarks it shows a transfer-rate of 85 - 95 MB/sec, which is absolutely great!!! But when I have a look at it in the Digisuite LE settings under storage I only get 8.32 MB/sec single write, which is absolutely pathetic! Also, playback with multiple layers in Speed Razor 2000x does not work, as suspected.

Any ideas?
Help is greatly appreciated.

Raphael

[Dr Mordrid]

That's WAY too slow.

What kind of and how many HDD's do you have in your array?

Have you installed the FastCheck monitoring tool yet? If not then do so, set it to run at startup and on its Options page turn OFF the S.M.A.R.T. feature and re-test.

Dr. Mordrid

[dolonescu]

fast-check activated -> no change

thanx for the quick reply,
i have installed fast-check, but not to run at startup. cannot find the option to switch "run at startup" on the options page. however, things do not change, when i disable s.m.a.r.t.! any other ideas?
of course, the tx4 is running in a 66mhz/64 bit slot and it gives it's performance to other programs, as tested with nero. it's only that the digisuite doesn't take it.
i have four ibm GXP 120 gb drives 7200 rpm attached to it. the system disk, an ibm gxp 60 gb, gives me around 22 mb/sec. in digiutils! strange, isn't it? something about the tx4 somehow slows down communication with the digisuite extremely.

i also have a mylex scsi2 raid 5 with 30 mb/sec, but i haven't installed it yet, so i don't know wether digisuit likes that fellow in the new configuration.

well, looking forward to your thoughts!

all the best

raphael

[Dr Mordrid]

Running at startup is an installation option. When run it should be in the SysTray.

Presumption: you're array is a RAID0

By "digisuite doesn't take it" do you mean that you drop frames when capturing at high bitrate or lossless MJPeg or do you mean that the drive test tool reports the low rate or both?

Some Matrox drive test tools work fine (RT-2x00 for one) while others are drek.

Also, some Matrox drive tests enter their low results in the registry and thereby restrict available capture options as well. Is this the case with the Digisuite or not?

To get an independent view of what the TX4 is doing I'd install SiSofts SANDRA system tool and use its drive test on the array.

SANDRA PRO: http://www.sisoftware.demon.co.uk/sa...html/dload.htm

The result you want to check is NOT the graph number as that is a composite of sequential and random r/w's. Look in the detailed list of results below the graph for "Sequential Write" and "Sequential Read" and you'll get a much more realistic idea of how fast the array is for video work.

Mine does ~90 mb/s sequential writes with four IBM 75GXP's on the RT-2000 system.

If these are still in line you can also try turning down the PCI bus utilization in the FastCheck tools Options page to give the Digisuite more breathing room. It might not have to go down that much.

Dr. Mordrid

[dolonescu]

testing

Hi,
i will be testing the system for another round. I tried to play back a project with two m-jpeg clips at 15mb/sec that have their opacity turned to 50%. i do not question the overall performance of the tx4, other programs have showed a transfer rate of around 85 mb/sec. but i will try sandra nevertheless...
i have thrown the fasttrack utility into the autostart folder and that works fine for running at startup.

however, when i turn down the lever that controls the pci usage of the tx4, the results in the digisuite storage report drop to around 7 mb/sec. somehow it seems to me, that the communication between digisuite an tx4 is awkward, or that 33 and 66mhz pci cannot transfer is corrupt.

any thoughts?

oh, yes it's a raid 0!

raphael

[Dr Mordrid]

It could be numerous things, including the Digisuite's drivers, which I consider likely since other apps play well with it. Let me check with a couple of deep sources about this and get back to you, OK?

I do know that the RT-2x00's and TX4 do play nicely....very nicely.

Dr. Mordrid

[dolonescu]

DETAILS

i have checked with sandra, but not with the pro-version. i guess you have to buy that thing. the regular version doesn't give me the details i want ...

however, i checked the details of digisuite again:
single write: 8,36
dual write: 4,16

single read: 54,65
dual read: 20,66

i can live with the read-speeds, of course, but not with the write-speeds. it seems to me, that writing to the tx4 is rather pathetic.

any news from your contacts?

raphael

[Technoid]

Seems that Windows buffering is weirding your write results.....

[Dr Mordrid]

Maybe, but I'd like to see the report from a program other than the DigiSuite tool.

Earlier he noted that NERO's speed test reported much higher results, which would argue against an OS cause.

Dr. Mordrid

[dolonescu]

NERO is that cd-rom burning software which only reports read-speed, is suppose. it's the test for cache speed used to buffer the data that's burning.

am i right with sandra? because i might just buy the pro package, if i have to ...

raphael

[dolonescu]

sandra

o.k. bought sandra

these are the results

buffered read 95 mb/s
sequ. read 94 mb/s
random read 7 mb/s

buffered write 56 mb/s
sequ. write 52 mb/s
random write 23 mb/s

access time 8ms

sounds great to me, although i expected a far higher write rate ...

but!!! that still doesn't help me very much with good 'ol digisuite. because, i cannot log 15 mb/s from pal digibetas, which i usually do.

so, i would be tremendously thankful for any more ideas, hints, solutions.

thanks for your support

raphael

[Dr Mordrid]

I've got inquiries in going in both directions (Digisuite and TX4) and am in a holding pattern 'til they get back. I'll post as soon as I hear anything.

Dr. Mordrid

[Brian Ellis]

Raphael

Just a thought... What kind of disk cache are you using? Depending on which version of Windows you are using, this can affect results terribly. Win can pinch all your resources. I have 384 Mb RAM and, if I allow it to take what it wants, it puts c. 350 Mb into vcache and leaves precious little for the apps. If your testing vehicle requires large resources in its own right, it will then require to use the swapfile. I changed the vcache to a fixed min/max 30 Mb and even Sandra improved her results. I also have a fixed swapfile of 250 Mb on it's own ad hoc partition, although it is rarely accessed.

Also I recommend all benchmarking be done on a newly booted computer so as to avoid comparing apples and pears due to memory leakage.

[dolonescu]

thanx brian, sounds intresting.

can you tell me where i can find the option for virtual cache. i only know the option menu for the swapfile. any danger with changing the virtual cache? because i remember that i crashed two edit-workstations once, that were running on nt4 when changing the swapfile size.

oh, i have 2gb of ddr-registered ram on a win2000 servicepack2.

raphael

[Dr Mordrid]

VCACHE....forgot about that one since I haven't used Win98SE in weeks at least, but you're using Win2K and it doesn't have disk cacheing problems.

Dr. Mordrid