Sign in with
Sign up | Sign in
Your question

SiS 645DX Review

Tags:
  • Motherboards
  • Sis
  • Systems
Last response: in Motherboards
Share
March 11, 2002 4:48:22 PM

That motherboard looks pretty solid, aside from the lack of USB 2.0 and such. I am disappointed that there was no attempt made to overclock it or compare it to overclocked memory systems. I would say the PC1066 - PC1200 RDRAM speeds at which we i850 owners are running our systems would likely slaughter anything running at stock speeds, including this new motherboard.

-Raystonn


= The views stated herein are my personal views, and not necessarily the views of my employer. =

More about : sis 645dx review

March 11, 2002 5:37:37 PM

I was disappointed that they chose to 'shortchange' the i850 board with only <b>256MB</b> of PC800 RDRAM (2 x 128MB) while stuffing <b>512MB</b> of PC2700 DDR into the SiS645DX board. While that disparity may not have effected all of the benchmarks, IMO it did effect some of them (such as Sysmark 2001). Don't you agree?
March 11, 2002 5:53:00 PM

I had not noticed that before. Thank you for pointing it out. The review is no longer simply a comparison between chipsets that use different types of memory. It is now a comparison between motherboards that use varying amounts of memory as well. I believe we all know that 512MB of memory will offer more performance than 256MB of memory, especially on operating systems such as Windows 2000, on which these tests were run. The disparity between the amounts of memory used pretty much invalidates all of their comparisons. To correct the situation they would need to provide benchmarks that use two 256MB RDRAM modules.

-Raystonn


= The views stated herein are my personal views, and not necessarily the views of my employer. =
Related resources
Anonymous
a b V Motherboard
March 11, 2002 6:05:09 PM

*ALL* computer manufacturers "under-RAM" their machines. This is one reason why I chose to build my own box!

If I had my way, anything coming out today would have a *MINIMUM* of 1GIG of RAM.
Anonymous
a b V Motherboard
March 11, 2002 6:47:20 PM

im personally beginning to have some doubts in my mind about these reviews on Tom's. no matter what the conclusion is - pro DDR or pro RDRAM - the benchmarks always support it. hmm... odd, don't you think?
March 11, 2002 6:49:21 PM

Actually I liked that review, it showed that this chipset moves the P4 forward in both IPC and DDR RAM. Face it, if DDR was dual-channeled, it would definitly allow the P4 to excel. Intel is only limiting it right now. Whether or not 512MB affected, it was CL 2.5, so SAY it was 256MB, it may have been around 4% less performance, but if it was CL2, I can imagine it being better than even 512MB RDRAM anytime. Regardless, until Dual Channel DDR comes in, the P4 is never gonna see the daylight. In these benchs I noticed that the RDRAM would only win in bandwidth operation, where its dual channel comes in to kick a 500MB less bandwidth RAM chipset.

--
For the first time, Hookers are hooked on Phonics!!
March 11, 2002 6:55:41 PM

I posted a thread in the <i>correct</i> forum to discuss this article. Feel free to move the discussion over there :tongue:

<font color=orange>Quarter</font color=orange> <font color=blue>Pounder</font color=blue> <font color=orange>Inside</font color=orange>
Don't step in the sarcasm!
Anonymous
a b V Motherboard
March 11, 2002 6:57:05 PM

Facts & Figures are in the eye of the beholder, really.

The only way to truely know is in your home, on your own machine.

I build older model AMD machines, and as you see from another thread, even I can get confused & forget to attach a ribbon cable properly!

LOL!!

:) 
March 11, 2002 6:58:54 PM

Quote:
I would say the PC1066 - PC1200 RDRAM speeds at which we i850 owners are running our systems...

Ray, would you kindly identify your particular CPU and Motherboard. Also, I presume your RDRAM is PC800. How are you achieving 1066 -1200? Thanks!
March 11, 2002 7:00:50 PM

Did you notice the very old version of Scisoft they used? This has to be the sorriest excuse for a front page story yet. They had to work real hard to make the results match what they wanted the outcome to be. I wonder if they even set the RDRAM to 3X? And of course they didn't use Samsung rimms even though they admit PC2700 is still in short supply. And of course the already stated 256Mb RDRAM vs 512Mb DDR. They just should have only done a DDR board comparison instead of shortchanging RDRAM just for a headline.
Ray, I thought Intel went DDR only for servers for the 1gig modules? They make it sound like RDRAM is now dropped for desktops too?
March 11, 2002 7:06:04 PM

Grab a Pentium 4 1.6A, a TH7II or P4T-E with the newer DRCG components, and two double-sided Samsung 256MB PC800 RDRAM modules. With these components I guarantee your system can get up to _at least_ PC1066 speeds.

-Raystonn


= The views stated herein are my personal views, and not necessarily the views of my employer. =
March 11, 2002 7:08:42 PM

RDRAM is not being dropped for desktops. Anyone who says otherwise is trying to sell you something; probably a DDR-SDRAM solution.

-Raystonn


= The views stated herein are my personal views, and not necessarily the views of my employer. =
March 11, 2002 7:27:47 PM

Yes, I beleive I also saw in the SiS article, something saying that RDRAM will really not be anymore needed later on, just look at the article, around 4rth paragraph or so.
It's clear RDRAM is really not that much unless Dual Channeled, and yet it suffers. I am going to hold on till I see Dual Channel DDR to see how much Intel coulda won since.

--
For the first time, Hookers are hooked on Phonics!!
March 11, 2002 7:49:30 PM

Dual-channel DDR is about as difficult to implement on a motherboard as octal-channel (8 channels) RDRAM would be. Just because they are named similarly with the 'dual-channel' label, do not make the mistake of assuming that dual-channel RDRAM and dual-channel DDR SDRAM are in any way comparable technology. The dual-channel DDR is a 128-bit solution. The dual-channel RDRAM is a 32-bit solution. The dual-channel DDR solution requires a great deal more engineering effort to implement on a motherboard than the RDRAM solution. It requires 4 times the number of traces (basically wires) on the motherboard.

-Raystonn


= The views stated herein are my personal views, and not necessarily the views of my employer. =
March 11, 2002 7:58:58 PM

I don't look at the "hard way" or "engineering effort", I just look at what would be the outcome. At 128bit, I think the bandwidth on the P4 is served well. Face it, as long as Intel is running the show with Rambus, the P4 stands about 40% less than what it could do. I just reread today the P4 Introduction article by THG from 2000, about all the structure inside, and indeed it shows how much Intel is missing out on the real deal, just for a Rambus Deal.
I've yet to see the RDRAM PC1066 prices come to Canada, but I am sure they are gonna be something to look at and never look later on.

--
For the first time, Hookers are hooked on Phonics!!
March 11, 2002 8:48:57 PM

Quote:
I don't look at the "hard way" or "engineering effort", I just look at what would be the outcome.

The outcome of what? There is no dual-channel DDR chipset available for the Pentium 4 right now. What outcome are you looking at? If we are going to start looking at the theoretical, how about PC1200 RDRAM?


Quote:
Face it, as long as Intel is running the show with Rambus, the P4 stands about 40% less than what it could do.

This is completely unsupported by current benchmarks. The current 645DX benchmark aside (which compares against a crippled i850 [half the memory]), you need RDRAM to get the best performance.


Quote:
I've yet to see the RDRAM PC1066 prices come to Canada, but I am sure they are gonna be something to look at and never look later on.

PC1066 RDRAM is not yet for sale.

-Raystonn



= The views stated herein are my personal views, and not necessarily the views of my employer. =
March 11, 2002 9:31:00 PM

Quote:
Grab a Pentium 4 1.6A, a TH7II or P4T-E with the newer DRCG components (etc.)

Ray~ What are DRCG components?

<b>God bless the <font color=red>U</font color=red><font color=white>S</font color=white><font color=blue>A</font color=blue></b>
March 11, 2002 9:39:25 PM

Direct Rambus Clock Generators. It's what generates the AC wave that gives you the clockspeed (e.g. 100MHz, 133MHz). Newer ones will allow higher clockspeeds without causing errors.

<font color=orange>Quarter</font color=orange> <font color=blue>Pounder</font color=blue> <font color=orange>Inside</font color=orange>
Don't step in the sarcasm!
March 11, 2002 10:17:59 PM

Ah so... how can I be sure I'm getting a recent board (with the newer DRCGs)?

<b>God bless the <font color=red>U</font color=red><font color=white>S</font color=white><font color=blue>A</font color=blue></b>
March 11, 2002 10:42:41 PM

Look at the DRCGs?

<font color=orange>Quarter</font color=orange> <font color=blue>Pounder</font color=blue> <font color=orange>Inside</font color=orange>
Don't step in the sarcasm!
March 11, 2002 10:52:07 PM

-Outcome as in what would theoretically happen. DDR PC2700, if dual-channeled, provides 5.4GB. RDRAM at PC1066 which is not out yet, gives 4.2GB. RDRAM at PC1200, still is not close to PC2700 Dual Channeled. This means in either RDRAM coming soon, the DDR Dual would have helped the P4 a lot.
-Crippled or not, RDRAM has a bigger bandwidth in those tests and SHOULD have yeilded better results. Ok Ray, I agree this isn't a fair test, so if someone knows how to contact the writers and tell them to redo the tests, I'll hold any comments until then. Bare in mind that the DDR RAM was also CL2.5, and is not at its best potential too, so there is room for more performance yet.
I would however like to see some tests with PC1066 against that new chipset.
IMO I beleive Intel will eventually wake up and do Dual Channel DDR, it just is better. Now I wish someone else here could come and tell you that I am right, I seem to be leading a one man army for nothing.

--
For the first time, Hookers are hooked on Phonics!!
March 11, 2002 10:57:41 PM

And a dual-channel DDR solution would take up <i><b>four times</b></i> the amount of space on the motherboard as the RDRAM platform would.

Yes, you are right about dual-channel (current) DDR against current dual-channel RDRAM. However, for motherboard manufacturers it is much easier to get the same and more bandwidth from RDRAM.

<font color=orange>Quarter</font color=orange> <font color=blue>Pounder</font color=blue> <font color=orange>Inside</font color=orange>
Don't step in the sarcasm!
a b V Motherboard
March 11, 2002 11:17:13 PM

It is the fault of motherboard and chipset manufacturers that we do not have DUAL CHANNEL DDR, in which case PC1600 would even beet these benchmarks, do to it's doubled buswidth. So we're stuck waiting for DDR-II+ to compete with PC1066-PC1200.

What really stinks is that motherboard manufacturers treat SiS as a bottom end chipset manufacturer, and assign them to their sloppier, cheaper designs, as seen with the 735/745. So the chances of us seeing this chipset pushed to its true potential on production boards is slim.

What's the frequency, Kenneth?
March 11, 2002 11:33:25 PM

Ok since a straight and<i><b><font color=blue> USEFUL</i></b></font color=blue> answer wasn't forthcoming I'll ask again for the benefit of the masses.. how does one determine which DRCG one is getting on a motherboard, is there a number or a revision etc..
any actual answers ??
lagger

<b><font color=blue>Checking under my North<font color=red> AND</font color=red> South bridges for <font color=green>Trolls</font color=green></font color=blue>
March 11, 2002 11:36:29 PM

Here is my <A HREF="http://fugger.netfirms.com/report2709.htm" target="_new">Abit TH7-II</A> pushing my single sided Samsung 128MB PC800 to beyond PC1066 spec.

I will have one of the 845DX boards very soon, Ill punish it like the rest. Good call on the 256/512MB memory desparity, I also test with max memory and all slots filled. as we already know on some AMD boards this can be a problem cutting memory bandwidth in half or non operational is worst case.

As I exceed the recomended insertions into the socket 478, the socket does get weak in terms of grip holding the CPU in. I have to be carefull when I pull off the HSF now, the CPU will stay attached to the HSF on removal. Pins are not bent but its something to be carefull of.

So far the only RDR I have not been able to crank is the Kingston remarketed Samsung ram. Its funny cause Kingston sells its ram with original Samsung heat spreader and samsung.com stamped onto in.

netfirms.com (my hosting) is doing upgrades atm, so link might be slow or not working. try again in a few.

Proving once again that <A HREF="http://www.zombo.com" target="_new">anything is possible</A>.
March 11, 2002 11:37:14 PM

couple of things...

1. i thought it was very poor they didnt have the same amount of ram. just looks bad for a comparitive study.

2. dual channel DDR cant be too hard... nvidia did it with the nforce.

3. there are two chipsets on Via's roadmap, the P4X600 & 800 that are supposed to be dualchannel ddr for the p4 later on this year. with luck the legal issues will be sorted and the chipset wont be so hard to find as the current p4x266a

4. im hungry. need chips

I love helping people in Toms Forums... It reinforces my intellectual superiority! :smile:
March 12, 2002 12:06:56 AM

Number 2 says a lot.
Number 3, IMO is gonna be the only solution for now if Intel and Ray disagree. I am somewhat sometimes starting to like Via's rebel ways, because they defy and push further. Granted the P4X266A is not the best chipset out there, but I think those you mentioned are worthy to keep attention. I simply don't care if Intel wants to play stubborn, at least some others are experimenting stuff that just might, help out the P4 in the end. We'll see though, but I definitly want VIA to take this step further.

--
For the first time, Hookers are hooked on Phonics!!
a b V Motherboard
March 12, 2002 2:42:20 AM

VIA can shove a few sticks up their arse! Hey everybody, we stole a CPU bus, now buy our product so you can download WAREZ with it, and steal everybody elses products! Oh, and please nobody mentioned how we've screwed our end users in the past. And don't hold it against us for forcing an SiS embargo. After all, even if we hate you, we love your money!

SiS or ALi should do DC-DDR first.

What's the frequency, Kenneth?
March 12, 2002 2:50:41 AM

*laugh* So true...

-Raystonn


= The views stated herein are my personal views, and not necessarily the views of my employer. =
March 12, 2002 4:05:19 AM

hehehe aint it the truth?

"no sir... there is nuffin wrong with our kt133 chipset... [later] whoops, seems there is a teensy problem with the MONST POPULAR soundcard out there, here, have the via 4in1's ver 4.28 they are just peachy"

TOP things im unlikely to see in my time
1. via releasing a bug free board
2. intel releasing their top of the line flagship at something less than strasphospheric (AU$1000+) cost
3. tech support lines treating you as if u really do know something
4. melty posting something interesting, valid, non combustable, unbiased and un-exclimationmarked.



I love helping people in Toms Forums... It reinforces my intellectual superiority! :smile:
a b V Motherboard
March 12, 2002 4:17:21 AM

More like:
Nothing wrong with our MVP3 chipsets, they simply don't support your hardware...
Nothing wrong with the Kt133, it just won't run at 133
Nothing wrong with the Kt133A, it just won't support either kind of top soundcard
Nothing wrong with the Kt266, ooops, forgot to fix those KT133A probs
Nothing wrong with the KT266A, REALLY, er, maybe?
Trust us, the Kt333 is the best thing since sliced bread!


What's the frequency, Kenneth?
March 12, 2002 4:28:06 AM

i believe the correct statment for the kt266 is:

nothing wrong with the KT266... all you need to do is pay alot more for ddr ram. of course i assume you would only like a 5% performance increase for all that expenditure?

and
nothing wrong with the kt333, despite it having MAJOR flaws, and once detected we shipped them anyway, just telling the mobo makers to 'disable' the buggy features rather than actually correct the problem. no sir, we wouldnt do that!

I love helping people in Toms Forums... It reinforces my intellectual superiority! :smile:
March 12, 2002 11:05:17 AM

Number 1 is KT266A, fully stable on my comp.

--
For the first time, Hookers are hooked on Phonics!!
March 12, 2002 12:48:30 PM

I'd just like to comment that the DDR was at 2.5 because as of yet certified cas2 PC2700 RAM that actually works in motherboards has been next to impossible to find, as far as I've seen and heard. Yes, cas2 PC2700 DDR SDRAM is being sold, but has anyone managed to actually run it at cas2 yet?

Another thing to note about the article as well: I didn't see it specified if the RDRAM was single or dual sided. As I've heard, single-sided has less latency, so that in itself could have made a difference to the article as well if they used double-sided RIMMs.

All-in-all, this has been one of the most disappointing articles from THG yet.

<pre>If you let others think for you, you're the only one to blame when things go wrong.</pre><p>
March 12, 2002 1:23:01 PM

Everyone has said you can hit that with the current P4 1.6A. How about the other P4s? Do the 1.8A, 2.0A, and 2.2A have the same overclocking room? Or is that where Air cooling falls short?

I do not like it Tom you see,
I do not like green PCB.
March 12, 2002 1:37:03 PM

I think that they are not as good for overclocking because you can only change the FSB, not the multiplier. So with the higher multipliers of the 1.8a, 2.0a, 2.2a, etc. you are running the risk of clocking the core too far for air cooling. (Or else you have to settle for a worse FSB which could negate the performance increase of the chip compared to a higher-FSB but slightly lower-clocked equivalent.)

Anyone please correct me if I'm wrong.

<pre>If you let others think for you, you're the
only one to blame when things go wrong.</pre><p>
March 12, 2002 4:31:46 PM

My opinion is, when there's a conflict between the Live and the 686b, who should fix it? Tough call, but the facts are that only one company has put any effort into fixing it. It's it Via's crappy product, or Creative's crappy product? I say Creative's, since I was very happy with my KT133a board (Abit KT7a-RAID).

<font color=orange>Quarter</font color=orange> <font color=blue>Pounder</font color=blue> <font color=orange>Inside</font color=orange>
Don't step in the sarcasm!
March 12, 2002 5:25:38 PM

But considering that Creative's crappy product worked just fine in other chipset motherboards it makes your opinion suspect. Compound that with the fact that Creative wasn't the only company to have this problem with VIA, and it makes it even more lop-sided. Thus it is my opinion that the problem was VIA's fault, and thus the onus falls upon them to fix it.

<pre>If you let others think for you, you're the
only one to blame when things go wrong.</pre><p>
March 12, 2002 6:35:33 PM

If i believe right someone made a PCI patch for the 686b a while back but it cripples the proformace on your comp.

THGC, saving 1 pc user from buying a GeForce4 MX at a time.
March 12, 2002 6:41:34 PM

Considering that other cards work great on Via chipsets, it appears to be Creative's problem. Couple that with the fact that Creative cards have problems on other chipsets, and your opinion is even more lopsided.

<font color=orange>Quarter</font color=orange> <font color=blue>Pounder</font color=blue> <font color=orange>Inside</font color=orange>
Don't step in the sarcasm!
March 12, 2002 6:43:34 PM

Unfortunatly if any bugs appear on my chipsets from Sound card, it's Creative's fault for not doing good WinXP drivers. Let alone the annoying DEVLDR that won't close on Shut Down, and the occasional EAX effects not applying, this is Creative all the way for not well writing XP drivers. Everyone has those errors I mentioned and blame them.

--
For the first time, Hookers are hooked on Phonics!!
March 12, 2002 7:06:58 PM

That's why I have a Hercules Game Theater XP... even the Fortissimo II was great (though it had issues with some gamee like B&W).

-SammyBoy
March 12, 2002 8:13:22 PM

Yeup, I love my Game Theater XP :cool:

<font color=orange>Quarter</font color=orange> <font color=blue>Pounder</font color=blue> <font color=orange>Inside</font color=orange>
Don't step in the sarcasm!
March 12, 2002 8:28:44 PM

I only see this

Memory 1 2 x 256 MB, RDRAM, 400 MHz, Viking
Memory 2 512 MB, DDR-SDRAM, 166 MHz, CL2.5, Winbond

seems they did a typo and fixed it. SiS 645dx ruxxors :D . To bad im using an Athlon.

THGC, saving 1 pc user from buying a GeForce4 MX at a time.
March 12, 2002 11:58:28 PM

a 79% effectivness

Do you like your samsung RDRAM.

cheap, cheap. Think cheap, and you'll always be cheap.AMD version of semi conducteur industrie
March 13, 2002 12:13:06 AM

um...did i miss something. on the set-up page <A HREF="http://www.tomshardware.com/mainboard/02q1/020311/sis64..." target="_new">Convinient Linkie)</A> Memory 1 is listed as <b>"2 x 256 MB, RDRAM, 400 MHz, Viking"</b> and Memory 2 is listed as <b>"512 MB, DDR-SDRAM, 166 MHz, CL2.5, Winbond"</b>. Have i done my math wrong? they both look like 512 to me...


<i>EDIT</i>: sorry rcf...didn't see your post. beat me too it.
no vestige of a beginning, no prospect of an end, when we all disintegrate, it'll all happen again.<P ID="edit"><FONT SIZE=-1><EM>Edited by jollygrinch on 03/12/02 09:17 PM.</EM></FONT></P>
March 13, 2002 12:36:08 AM

Hmm seems like Ray's point is no longer valid and that indeed the SiS with DDR does pack a punch, not to mention it ain't even CL2, looks like DDR is more than what it promises.

--
For the first time, Hookers are hooked on Phonics!!
March 13, 2002 3:50:04 AM

Sorry, this was a type in the Test Setup, all setups ran with 512 MB of Memory.

Fredi
Tom's Hardware Guide
!