SiS 655 *2-18-03 UPDATE* MSI, Gigabyte and ASUS Done

[First]

2-18-03 UPDATE: We just posted our ASUS P4SDX (SiS 655) review here. Enjoy!

The only information I'm waiting on is what the ASUS P4SDX will retail for and whether any A0 boards will be shipped to customers. My review sample was B0.

Anyway, there will be a nice 800MHz FSB processor/motherboard article waiting in the wings in the next few weeks. I'm just waiting on some hardware to arrive, maybe you've familiar of it; Springdale and Canterwood.

 

[Kowan]

Glad to hear we'll be seeing some benches on the SiS 655.
Anyway you could compare the E7205 with it?
Or have a similar setup to your GB review to see how well they do against each other?

 

[poMONKey]

you gotta compare it to the E7205!!!

this is just what i've been waiting for... perfect timing cuz i'm about to buy a p4pe and wait for springdales.

but i'll hold off until this roundup, now...

Evan, are you planning on using a p4 3.06????... please!!!

 

[sumrtym]

Evan,

Which Gigabyte board are you getting...the regular, ultra, or ultra2?

1) Things I'd like to know (particularly on the Gigabyte) is to whether you can hook up front USB and Firewire from an Antec Performance Plus case to them. The pictures I'm seeing of the Ultra don't seem to look like you can (at least, as far as Firewire is concerned).

2) Also, on the Gigabyte (guess the rest of these will be Gigabyte specific), I want your take on the placing of the IEEE1394 location between slot 1 and slot 2 on the PCI bus. Interference issues, and why there? Is slot 1 or 2 usage going to be necessary to avoid IRQ conflicts (say for a soundcard), and thus you couldn't use the firewire because of blockage?

3) If you're getting an Ultra2, I'd be curious for you to try overclocking both WITH and WITHOUT the dual power system to see if you see a difference with the exact same components and the dual providing 6-phase, not just backup. This would test the "increased stability" claims.

4) Also, and I don't know how you'd check this, I believe Gigabyte's implementation supports hyperthreading on the AO chipset only in BIOS, and that there are "limitations" to this approach. I don't know what you could do to test that other than slam some PC2100 and test it against the Gigabyte Granite Bay with essentially the same settings (FSB, memory timings, 1:1 ratio, etc.), same memory, and a 3.06 Ghz HT chip running different benches known to benefit from HT. That would be VERY interesting since they're claiming HT support. I think you might see some of that fall down in that test.

5) <edited in> Almost forgot...since the previous SIS chipset had 9700 issues, I'm curious to see if you find any. (Looking at using an AIW 9700 myself).

6) <I'm going to set a record for most edits to a post> I want to know from what Gigabyte tells YOU, does the naming convention mean that Gigabyte is going to back up this board as working with hyperthreaded 800Mhz FSB CPU's via a future BIOS upgrade...for the record???

Those are the things I REALLY want to know, and I think at least 2-4 would be of interest to everybody (kind of high interest, really, as I haven't seen the #3 and #4 tests, just mouthing of Gigabyte claims). #1 is also very relevant to those of us with front firewire and usb, and #5 to anybody who ever had a 9700/mobo problem. #6 goes to the heart of whether Gigabyte is semi-deceptively marketing this board, and those of us who would like to use it for those chips without having to buy Springdale, thus skipping a mobo upgrade, would love to know that.

 

[mechBgon]

Originally posted by: sumrtym
Evan,

Which Gigabyte board are you getting...the regular, ultra, or ultra2?

1) Things I'd like to know (particularly on the Gigabyte) is to whether you can hook up front USB and Firewire from an Antec Performance Plus case to them. The pictures I'm seeing of the Ultra don't seem to look like you can (at least, as far as Firewire is concerned).

2) Also, on the Gigabyte (guess the rest of these will be Gigabyte specific), I want your take on the placing of the IEEE1394 location between slot 1 and slot 2 on the PCI bus. Interference issues, and why there? Is slot 1 or 2 usage going to be necessary to avoid IRQ conflicts (say for a soundcard), and thus you couldn't use the firewire because of blockage?

3) If you're getting an Ultra2, I'd be curious for you to try overclocking both WITH and WITHOUT the dual power system to see if you see a difference with the exact same components and the dual providing 6-phase, not just backup. This would test the "increased stability" claims.

4) Also, and I don't know how you'd check this, I believe Gigabyte's implementation supports hyperthreading on the AO chipset only in BIOS, and that there are "limitations" to this approach. I don't know what you could do to test that other than slam some PC2100 and test it against the Gigabyte Granite Bay with essentially the same settings (FSB, memory timings, 1:1 ratio, etc.), same memory, and a 3.06 Ghz HT chip running different benches known to benefit from HT. That would be VERY interesting since they're claiming HT support. I think you might see some of that fall down in that test.

5) <edited in> Almost forgot...since the previous SIS chipset had 9700 issues, I'm curious to see if you find any. (Looking at using an AIW 9700 myself).

6) <I'm going to set a record for most edits to a post> I want to know from what Gigabyte tells YOU, does the naming convention mean that Gigabyte is going to back up this board as working with hyperthreaded 800Mhz FSB CPU's via a future BIOS upgrade...for the record???

Those are the things I REALLY want to know, and I think at least 2-4 would be of interest to everybody (kind of high interest, really, as I haven't seen the #3 and #4 tests, just mouthing of Gigabyte claims). #1 is also very relevant to those of us with front firewire and usb, and #5 to anybody who ever had a 9700/mobo problem. #6 goes to the heart of whether Gigabyte is semi-deceptively marketing this board, and those of us who would like to use it for those chips without having to buy Springdale, thus skipping a mobo upgrade, would love to know that.

I've been after AntecCSR for a long time for dongles which convert the Antec multi-pin FW cable into a standard male FW plug. A whole cable that ends in a FW plug would do the job too (it's detachable at the front circuit board). AntecCSR recently PM'ed me that they are now available, and I called Antec and ordered four!

I didn't get them, unfortunately... I got four of the front circuitboard with the USB and FW ports, and four stock sets of cables that terminate in multi-pins. Still waiting to hear back from Antec on whether the dongles are a myth or whether I simply was sent the wrong thing by accident. Since you're interested in this I'll try to remember to PM you what I find out (if you enable your PM's, anyway).

I am a strong contender for most edits on one post, I'm sure... take yer best shot

Evan, you do great mobo reviews, I'm sure this one will be interesting too. Can I suggest using an extra-strong PSU (Antec TruePower 430 or Sparkle Power 400W)?

 

[sumrtym]

*cough, cough* TruePower 430 *cough, cough*


Thanks, mechBgon! I'd appreciate that. I'm one of those guys with a "computer desk" meaning it's buried in, and running through, the desk. Getting at the back isn't an option and I hate to just have a firewire cord plugged in and run to the front (or USB) for when I want to plug things in. I'm wanting to do video editing on the system for burning to DVD.

Being able to use Antec front Firewire and USB connections to a mobo is pretty important to me.

P.S. I'm building my first system for myself now. I'm REALLY hoping that by keeping the door open while working and having the 3 Antec fans off the TruePower fan headers along with two Panaflo L1A intakes on the front that my temps will be ok. I've going to end up with a DVD-Rom, DRU-500a, floppy, AIW9700, and 2 WD1200JB 7200 rpm drives in there. Unfortunately, there will be about a 1/2 to 1" gap ONLY between the side wall of the desk and the side intake fan. There's a fairly big gap going all the way down on the back inside for running cables out that I hope let's enough of the hot air escape.

P.P.S. <edit one on this post, by the way> I almost forgot. Enabled my PM

 

[First]

Nice suggestions guys.

mechBgon, what was that WinZip benchmark you wanted me to run? Btw, I only use a 300W Enermax PSU because I want to show a base FSB overclock one can expect from their board. Most overclockers aren't going to be using any worse than a 300W PSU from a reputable manufacturer. Honestly, not many people are using an Antec 430 or Enermax 550. Besides, you really don't need more than an Enermax 350 for a substantial overclock, unless of course you're looking for every last drop of overclock out of your system (which I admit some are doing ).

 

[mechBgon]

Ok...

• Take some arbitrary large amount of files/folders and put them in a folder on the system partition. I used Disc 1 of classic Unreal Tournament because it's something people are likely to have.
• Right-click the folder, choose WinZip > Add to Zip file...
• Choose Maximum (slowest) compression and begin timing when you click the Add button
• Stop timing when the progress bar disappears

Here are some results so far:

mechBgon's results:

• Duron 1GHz@1.33GHz on K7S5A with 256Mb of PC133 at aggressive timings, using Win98SE: 4 minutes 53 seconds
• AthlonXP 1700+ @ stock speed on A7V333-R with 768Mb of PC2100 at aggressive timings using Win2000: 3m 19s
• AthlonXP 1700+ @ stock speed on A7N266-VM with 512Mb of PC2100 at SPD using Win2000: 3m 26s
• AthlonXP 1600+ @ 1750MHz 166/166 on A7V333-R with 512Mb of PC2100@PC2700 at SPD timings using Win2000: 2m 58s

I also did enough testing to rule out the significance of 256Mb versus 512Mb, IDE versus SCSI and slow SCSI versus 15000rpm SCSI. Operating system may have a bit of an impact, with Win2000 being better than Win98SE.

Wingnutz's results:

• P4 2.4B on i845G (presumably with add-in video card?) with 512Mb PC2100 @ PC2700 SPD timings using WinXP: 3m 39s

jbond04's results:

• P4 2.53 on i850E with 1Gb PC1066 using WinXP Pro: 1m 27s :Q

jamesbond007's results (not sure if I got his username 100% right):

• Dual AthlonMP 2200+ on 760MPX with registered PC2100, I forgot the OS and don't have his PM: 2m 16s

jamesbond007 also reported that a P4/RDRAM-equipped buddy of his was outrunning him in WinZip. He tried the equivalent benchmark using WinRAR (a benchmark I see you used in the 3.06GHz P4 HT CPU review) and again got smacked down by the P4/RDRAM combo. And that?s PC800 RDRAM.

Ajaykay didn?t give system specifics but has a P4 2.8 on his Gigabyte 8INXP and reports ?2 ½ minutes?

Conclusion: the P4 does curiously well at compression stuff when equipped with RDRAM, even compared with dual-channel DDR, if these results are correct.

 

[THUGSROOK]

Evan ~ id like to know if the 655 is future proof and can run the 800fsb cpus.
a P4m could be used for testing.
as you can see by my sig ~ even a p4b533 can run 223fsb (+/- a few)
any board that cant run 800fsb cpus are quickly gonna become "low-end".

thx again for your excellent reviews!

 

[Elvis2]

Bump for an eagerly awaited review

 

[ErikaeanLogic]

Ditto with what THUGSROOK said, Evan. We have yet to see even an E7205 board do 200+ fsb, and that is the deal breaker for me as I am one of the folks running a P4m at over 220 fsb now, and all on an i845E motherboard (Asus P4B533-E). I admit to feeling a little silly, having rattled the sword in the scabbard while anxiously awaiting the GB chipset boards which are only now are trickling onto the market, with higher release prices than I've ever seen attached to a desktop board at release; then to find out that the board isn't going to be "future-proof," as THUGS calls it, is a major disappointment. The way I see it, this would be a great opening for SiS to release a solid (read: stable) chipset with native DDR400 support AND that offers the superior bandwidth which the dual-channel platforms can provide. SiS nearly siezed an opportunity like this with the 645DX chipset but Intel quickly countered with the releases of the i845E/G, followed by the PE/GE. SiS tends to be considerably less expensive than the Intel platforms and keeps Intel on their toes with fresher releases for us, too.

So, in short, I'd like to underscore the request to see some 200+ fsb testing on the SiS655 boards in your review. Keepin' my fingers crossed, and looking forward to another excellent review!

 

[jbond04]

Originally posted by: mechBgon

jbond04's results:

• P4 2.53 on i850E with 1Gb PC1066 using WinXP Pro: 1m 27s :Q

I remember that...scared the living daylights out of me...I thought for sure it was wrong. My computer continues to surprise me sometimes (except when I render global illumination solutions in 3ds Max...I need a freaking render farm )

 

[First]

*Go to second page for clarification on MSI boards*

 

[Tummy]

I'd also like to see how high a fsb the 655 chipsets can do!

 

[RideFree]

Evan,
That is some, "Seriously need-to-know" info on the MSI. This has serious implications for me.

Did I say that I took your information seriously?

Just so that there is no mistake, this is a seriously complementary comment on your previous post!
No more confusion for me...I'm giving it up for Lent.

 

[Yozza]

I'd like to see a number of things performed if at all possible:

• Memory Bandwidth and latency comparisons in read, write, copy, and modify operations, on the SiS655, i850E and E7205, at DDR266, 333, and 400 speeds on the SiS655 (the latter two obviously being asynchronous). It would be interesting to see how SiS655's memory write bandwidth fares, espcially with i850Es far superior, and E7205's luckstre write speeds. How asynchronous operation affects bandwidth would be especially useful, especially in seeing how asynchronous operation is offset by lower memory bus-turnaround latencies resulting from the higher clock granularity. This should help DDR RAM in heavy random read/write operations considerably, where RDRAM has a significant advantage thanks to its higher clock granularity and very low bus-turn-around latencies.

• Performance in both 2 * 64 bit memory controller mode, and a single 128 bit combined channel

• Stability and Performance at high FSBs - does the SiS655 operate reliably at 167MHz+ FSB? The previous chipsets have been shown to possess very poor overclockability in the FSB. Will it have the possibility of supporting 200MHz FSB Pentium 4 CPUs? In 1:1 ratio with the memory?

• PCI/AGP clock fixing at independent speeds - does it work reliably, and not just at 33MHz/66MHz? What are the limits?

• Support of HyperThreading in the A0 chipsets - Does Gigabyte's BIOS-based workaround work? What are its limitations?

Many thanks. This list contains my initial thoughts on comparisons and details I'd like to see. I will add to these as I come up with more.

 

[RideFree]

Evan,
I know you've invited, "I'd like to see this or I'd like to see that".
However, in all honesty, so-far you've done just fine for the majority. I don't see any need for change. And for all that has to be coming up on your plate, I don't see how 24 hour days will suffice.

All those finite questions are best left for our home labratories<sic>. After all, we've got to have something to report or there wouldn't be a forum.

 

[TheCoz]

Originally posted by: THUGSROOK
Evan ~ id like to know if the 655 is future proof and can run the 800fsb cpus.
a P4m could be used for testing.

That's a great idea. You could have a 'maximum FSB test' using a P4-M and see how high a FSB speed the MCH can manage. Those P4-Ms should be easily capable of 220MHz+ and that'll probably find the limit of the MCH before the CPU.

 

[sumrtym]

Hey, I personally think those were all good questions (and not just because I really want to know the answers before deciding on that board).

ocworkbench is reporting that the name has been changed form the ultra 2 to the SINXP1394. To me, that seems to suggest that Gigabyte knows the board won't do 800 Mhz and didn't want irate customers accusingly them of deceptive advertising of the board (inferring it was good for 800 Mhz FSB when it wasn't).

 

[Conscript]

even, any idea when the Gigabyte 655 board is goign to be available? My next computer has it in it's future...

also, what's the difference between ultra and ultra2? Just the gigabit lan and extra power supply thingy? Hopefully they have the same bios options...


thanks evan, keep up the good work!

 

[memories2002]

i was choosing between the 2 platforms, deciding to go for the nforce2 athlon combo..guess i will have to stop and think again...my post:
http://forums.anandtech.com/messageview.cfm?catid=28&threadid=955459

 

[Elvis2]

So, assuming MSI can't get me a stable 655 board and AOpen doesn't send me their 655 board until next weekend, I'm just going to go ahead and write up an individual Gigabyte 655 review, which should be done in a matter of days.

Right on Evan.
Keep up the good work

 

[sumrtym]

If that turns out to be the case, then I'll guess you'll have time to test and answer ALL the questions I asked, eh?

lol

 

[THUGSROOK]

Tell me how to improve my motherboard reviews. Email me at....

i was gonna say ~ you need motherboards




j/k of course

 

[mechBgon]

Originally posted by: THUGSROOK

Tell me how to improve my motherboard reviews. Email me at....

i was gonna say ~ you need motherboards




j/k of course

LOL!