Hows it going with the SATA problem?

[Rolle2k]

Some months when the problem was up in the old forum, AG told that DFI has talked with silicon image about the problem with slow write speeds, and that silicon image would try to fix it. Hasn't SI got any solution to the problem, and will there even be a solution? i really would like to know this, because im going to buy another drive soon, and if the write problems still are there i maybe have to buy a add on card, to got those speeds back :/

[RGone]

:confused: I don't think Silicon Image has fixed a thing. I don't see any SATA working like Intel 'hyped' it when they started the push for everyone in the world to go to 'their' idea of a new drive type. Some even brand new Silicon Image stuff works and the other half seems not to work> but that is only in relation it seems to a slower 'write' speed than is expected.

 

I made a quick check of various SATA controllers on various brands of motherboards and none no not one of them impresses me with working as quickly and 'evenly' as the older PATA raid setups I and most are used to seeing run.

 

Native SATA on Via seems problematic based on the brand of drive in use. SI3114 seems roughly the same as it was when we first saw it on the UI and LpB motherboards. Good in some places and iffy in others as relates to even read to write patterns. Intel's ICHR5 seems to be OK>but then it was Intel 'idea' in the first place. They love to be able to tell people what to do. Just normally their chipsets work better than most. Well until they came with a whole 'weeks' worth of ICHR6 chipsets on the new 925/915 mobo's that would run the CMOS battery down and caused all the mobo makers to recall the new boards. Hehehehe.

 

SATA, yeah a paper giant to me at this time and when all is considered> it works, it just does not work in many implementations at the level many are used to seeing. The why is only vaguely glimpsed when the SATA driver and the layer of SCSI control that "most" basically use to control the raided drives. The drivers are not able to turn on WriteCaching; in many cases and that slows the write speeds down. Seems this is not all a bad thing with the O/systems in use today as WinXP and the like do not send a 'flush' cache signal to the SCSI devices at 'shutdown' to keep errors from occuring at shutdown of the computer.

 

I have two SATA maxtors in raid 0 on my LpB and other than the glorious ego boosting benches that I am used to seeing the dang 3114 works. Works fast and fine as I can tell in day to day use. Strange how a benchmark can queer the whole deal and leave a sour taste in one's mouth, though. I had so many other things to look at that I quit running drive benches by and large and the SATA 3114 just kept on working. Maybe one day the hype of SATA will be realized. For now for me in my opinion; Intel SATA by the originator of the idea is still the only one that seems to mostly always get closer to right.

 

It took me 3 days of searching the internet to glean what I have based my opinion on and others may have a different opinion but right now, I just don't see any superb overwhelming value to SATA in the first place and I do not know of SI getting any more speedy nor any less speedy either. I am just running Raid 0 and using it and it seems ok day to day.

 

Sincerely, RGone...

 

PS: dang I just clicked on 'submit reply' and was shocked again at my new Avatar. Oh my gawd>AG has been up late last night again. Hehehehehee. :nod:

[Rolle2k]

RGone: thanx for the fast answer, too bad SI haven't taken the problem seriously then. Will se, maybe i have to buy a add on card.

[uwackme]

I hadnt bothered to check, but using the Sil SATA raid utility... use the old Sil3112 one... "3x1x_x86_win_sataraid-10019" on thier site. The "javaII" PoS they have listed for the 3114 is crap.

 

Write caching IS on with mine....and patch KB331958 from M$oft is therefore CRITICAL to enforce writecache flushing on shutdown.

 

*WARNING* if you use the latest NVidia IDE driver, 331958 will NOT be used and suddenly you will be Mr Corruption. Dont install it, use the default M$oft IDE, even with the new 4.24 driver pack.

 

I get low 80K's write, mid 100K's read in ATTO on this DFI 3114 setup. I got a more symetrical mid-90K's/mid-100K's on the Sil3112 controller (NF7 and PCI). Using 2x Maxtor 80G/8 pata drives on ABit serillel convertors. I forgot to check my bro's raptors.... Ill report back.

 

But it feels fine, plenty peppy so I aint complaining. I would still LUV to see a benchy on 4x sata drives on the 3114.... but Im not gonna rearrange all my 80G drives just for a test... YOU DO IT :nod:

[Angry_Games]

im gonna agree with everything RG says...SI hasn't done a thing to update their weak write speeds.

 

but also, if you read RG's post, you will see him saying something is 'relative' and how a bench can leave a sour taste in one's mouth.

 

I interpret this (correctly I hope) and feel the same way...that speed is relative to the user, and benchmarks, while great for comparing to yer buddy or others (such as a penis measuring contest as it goes most of the time), really mean nothing in the real world.

 

translation? speed is relative to the user in question, and benchmarks can make something that is truly good and fast seem inconsequential and weak when in reality it is not. Say you just came off a P3-500Mhz 256MB RAM to an XP3200 1GB RAM...speed is something you are going to feel right away in this situation.

 

now imagine coming from a P4 2Ghz to an XP3200...speed is going to really feel about the same. So you run benchmarks to figure out how much speed you think you should feel. If the numbers don't mesh to what everyone else on the internet claims to be the holy grail of fantastical speeds, then you grow angry and say 'wtf i want SPEED dammit'. And so it goes again.

 

this doesn't excuse SI from being able to ignore user complaints. This also, in my opinion, doesn't excuse them from resting on their laurels when VIA's integrated VT8327 SB SATA RAID is very much superior to the 3114...both being 4-channel. And the king-daddy ICH5R...but we dont expect anyone to beat Intel at this game, but we'd like a good close race.

 

And DFI picked the 3114 based on teh 3112A's excellent performance, and ability to run 4 channels instead of the 3112's 2 channels. How little did we know that the 3114 wouldn't perform as good in the write department as the 3112...but hey you got 4 channels of SATA right?

 

And another thing is the benchmarks...the TW labs show scores relative to what the 3112A can do, and I have seen others that can closely match those scores. I can match the 3112A in read speeds (127MB/s), but cannot come close to write speeds (getting only 65MB/s). Funny thing though...is that i can't seem to tell any difference between reads and writes so I run a different benchmark...and get same tho slightly different scores...but still it FEELS like its kicking butt and taking names the entire time.

 

Then i run the raptors on this K8M800-MLV VT8237....and it feels suspiciously like the same exact speeds as the Infininty and the 3114. Except that I get writes upward of 131MB/s at times in benches. Yet it feels EXACTLY the same as the 3114.

 

So enter teh king daddy 875B LanParty and its vaunted ICH5R...and guess what...it still FEELS exactly the same, and the scores are almost identical, except the Intel board is much much much more consistent...no funny ha ha yet strange dips down into the 60's for these babies...no, they scream at a constant 120+ MB/s the entire time...but yet again they FEEL no different than any of the other setups.

 

so AG, wtf is your point in all this?

 

SI screwed up but yet I find it hard to really find fault in them as regardless of benchmarks, the drives connected to their controller seem to run equal to my other excellent RAID controllers. this may be sacrilege to a lot of you, and mobs with torches and pitchforks are already forming outside, and the cries of NF7 this and NF7 that can already be heard...but the simple truth is that I cannot tell the difference from it and the other controllers...and I don't know what else to say other than that...and the fact that neener-neener i have 4 channels of SATA goodness instead of only 1 or 2.

 

so remember, speed is relative to the user more than it is to an entire society (and yes, we at the Street can be considered a society, as scary as that is!). This is because an individual has specific tastes and expectations vs an entire society (basic sociology).

 

im still ticked that SI has ingored every request we've made to do something about it...but I'm not so disappointed because the controller works and works well, and nipping at the heels for a couple extra juiced MB/s...I've got more important things to do like keep messing with RG's avatar until he sets himself on fire in a crazy fury hehehe

[uwackme]

Besides.... if you want the bench numbers of the Sil3112A with your LpB/Inf.... just spend $24 @ NewEgg on a PCI 3112 card.

 

Then you'd have SIX sata channels :nod:

[Rolle2k]

hehe, but i live i Sweden, so that will be a expensive card if i order from newegg