Cebit, SIS and AGP 8x

[Skeptik]

I found the graph at page 10 slightly amusing
The bars indicates that it's beating the GF4 MX (notice they didn't write out "MX") bigtime, but looking at the numbers it's with less than 10%.

Hehehe, SiS is trying marketing...

But keep in mind the GF4MX runs at 270 mhz, compared to the 200 mhz fast SiS332.
Looks good to me, if the prices are ok (as well as the drivers...)

 

[mboeller]

I found the graph at page 10 slightly amusing
The bars indicates that it's beating the GF4 MX (notice they didn't write out "MX") bigtime, but looking at the numbers it's with less than 10%.

Agreed, thats nothing spectecular keeping in mind that is has 70MHz higher core and also the benefit of being DX8 compliant, but at least we have a lower end DX8 compliant card. Lets hope SiS can improve the driver soon..

Look closer

They compare the SiS332 with the GF4 MX440. The SiS 332 run's at 200 / 200 MHz (core/DDR-memory) the GF4 MX440 @ 270 / 200 (core / memory ). So the clock of the SiS 332 is slower but the chip is faster, at least in 3D Mark 2001

So with the same bandwidth the SiS 332 is faster/got more points than the GF4 MX 440. IMHO that's quite good for an new chip.

 

[Nexus]

So the clock of the SiS 332 is slower but the chip is faster, at least in 3D Mark 2001

Indeed! In a hurry I only looked at the core clock 270 vs 200 and assumed the SiS chip must have the higher clock to achieve this score. I guess my sceptism against this chip shined through, kind of a freudian slip.

 

[midnight creature]

I want to see the higher clocked versions of this chip

Anyone care to extrapolate on sheer bandwidth/clock numbers what kind of performance we will see @ 300mhz-350mhx(overclocked) for the sis 336?

 

[Teasy]

It get's even better! the SiS 33X-Series has 4 x 2 TPU's like the GF3 / Radeon 8500 and they seem to have bandwidth saving features too (otherwise the 9,6GB/sec with 275MHz-DDR-Memory would not make sense ).

That doesn't make sense though unless only the 336 has the bandwidth saving tech. Because the 332 has 200mhz ram and is quoted with 6.4gbps and 200mhz x 16bytes x DDR is approximately 6.4gbps and the 334 has 250mhz ram and is quoted with 8gbps bandwidth which again is correct for 250mhz ram. It actually looks as if SIS have either mis-printed the ram speed for the 336 or mis-printed the mem bandwidth figure for the 336.

So with the same bandwidth the SiS 332 is faster/got more points than the GF4 MX 440. IMHO that's quite good for an new chip.

Yeah.. although the Geforce 4MX 440 scores seems a little low don't you think? On Anandtech the Geforce 4 MX 440 got a score of 5480, although that was on a Athlon XP 2000 system, but still I wouldn't expect a P4 1.9ghz system with a Geforce 4 MX 440 to get almost 700 points less in 3dmark2001 then a Athlon XP 2000 system, or am I wrong here?

Anyone care to extrapolate on sheer bandwidth/clock numbers what kind of performance we will see @ 300mhz-

At 300mhz thats 50% more fillrate and bandwidth then the card in that benchmark so going on pure extra bandwidth and fillrate that would take the score from 5200 to 7800.. which is impressive, especially with early drivers.

 

[Ailuros]

It really starts getting interesting with SiS. More than anything since the roadmap suggests Q1 releases for all 4 variations of it.

I remain sceptikal though until I see the card/-s reviewed and that in real games.

 

[midnight creature]

I've seen shifts lke that

AMD does awesome when it comes to 3d benches and intel does awesome when it comes to business benches. Awesome is defined as +/- 5% of rated performance # from amd compared to intel megahertz. My amd 900mhz scored almost 500 points higher than my gf's 1000 megahertz pentium 3 both with a voodoo 5.

 

[Humus]

Quite interesting information:

http://www.theinquirer.net/16030201.htm

According to this it supports PS 1.3, but no vertex shader (done on CPU).

 

[mboeller]

According to Tomshardware it supports even Vertex shaders !
lokk at the "Day4 at cebit" page

 

[Humus]

Yeah, I just saw that.

http://www.tomshardware.com/business/02q1/020316/cebit-01.html

 

[Skeptik]

Mmh, this chip becomes more and more interesting!
Can´t wait for the first real game benches...

How about the 2D/3D image quality of SiS-Cards?

 

[LeStoffer]

I got mixed feelings about this one, gentlemen!

It will all come down to price of course, but there is already a lot of fast and reasonable affordable DX8-cards out there from ATI and nVidia. I'm just not sure that I can see a large market from SiS' point of view since they are newcomers. I fear horrific drivers etc.

More competition is good and all, but we really need to push DX8 into the valuemarket more than anyhing. If SiS helps out here it's great, but I have my doubts that they will have large succes in this area and will eventually back out of this business in two years or so.

Regards LeStoffer

 

[Idiot]

They are selling very well inside china.

per their invester info the gfx biz is profittable for them.

Also 630 is a very popular notebook chipset, and the 660 (which integrates 330) will likely be too.

(they say they are going to intro dual DDR channel chipset in Q2, dunno if 660 is one???)

 

[Tygrus_Artesiaoa]

sis image quality

Sis 2d quality isn't bad...and their video engines have supported many features for a long damn time as far as dvd goes and they support tv out its 2d speed is also fairly fast enuff...i havent tested them really in ages.
Their 3d IQ sucks though...lots of supported features...but many times filtering is crummy images don't look so crisp and is blurry...
sis 315 doesnt do trilinear right and its tnl isnt fast and AA doesnt work

 

[Ailuros]

Can someone show me where it has been verified that it actually supports vertex shaders in hardware?

/snip from Tom's

In addition to graphics cards with ATI's Radeon 7500, 8500LE and 8500 chips, Gigabyte presented a prototype of an AGP 8x card with the SiS336 GPU. The 300 MHz chip should be available on the market in two months, offering DirectX 8.1 support. According to Gigabyte, the GPU has a pixel shader unit.

DX8.1 support could be claimed easily, with full dx8.1 compliance being a different story. Not a crime at all since vertex shaders can be emulated decently on strong cpu's, but I can't even see the company itself claiming dx8.1 compliance just for the record.

 

[Humus]

Looks like they have removed the reference to a hardware vertex shader unit at tomshardware, which I suppose means that it doesn't have one.