So apparently, the SiS655 is the fastest chip out there for P4, by a slim margin. So much for Rambus, at least until SiS´s own R659 with quad-channel PC-1200 RDRAM...
Yeah, where was THG on this one? I mean, isn't it kinda noteworthy that SiS is now the top performing chipset manufacturer for Intel processors? Eh? Anyone?
Well considering how fast they buried the Radeon 9800 Pro artcle with worthless crap on the front page, I am not surprised they "overlooked" giving this any exposure at all.
<P ID="edit"><FONT SIZE=-1><EM>Edited by rain_king_uk on 03/10/03 03:24 PM.</EM></FONT></P>
Last I checked, Anandtech's results of the 655 were crappy. It's yet another failure. Why? Because to me it seems no company is yet able to match the highly tuned i850E interface. We need nVidia in Intel's camp, fast!
--
This post is brought to you by Eden, on a Via Eden, in the garden of Eden.
Eden: Did we read the same review? Because Anand's conclusion was rather glowing:
Quote :
First off, we can safely say that MSI's 655 Max motherboard is the fastest Pentium 4 motherboard we've ever tested, albeit by a relatively slim margin over Rambus solutions based on Intel's 850E chipset. While E7205 boards were more or less the performance equivalent of PC1066 Rambus equipped systems, MSI's 655 Max has clearly taken the performance lead, but again only by a very thin margin.
Motherboards based on SiS' 655 chipset will be marginally faster than E7205, 850E, and 845PE motherboards as long as dual channel DDR400 support is fully implemented in SiS 655 motherboards (MSI's board is just one example). We never would have thought it, but dual channel DDR400 actually outperforms dual channel DDR266, despite the fact that the Pentium 4's FSB doesn't require more than 4.2GB/s of peak bandwidth, which is exactly what dual channel DDR266 is capable of offering.
Similar to the situation with AMD, the PC3200 gives improved performance due to less latency in real time (Cas Latency measured in cycles, shorter cycles mean faster response time). Of course PC2100 at Cas 1.5 should outperform PC3200 at Cas 2.5 in dual channel mode (if all the other memory timings were equally "sped up" ), since any bandwidth beyond PC2100 dual channel is wasted.
<font color=blue>Watts mean squat if you don't have quality!</font color=blue>
I'm sorry, but using DDR400 is not my type of "max performance" I wanted.
It used DDR400 in order to occasionally win and tie.
I have yet to meet any damn Intel chipset for Dual Channel, which with PC2100, maxed bandwidth possible on 533MTs processor, can outperform the PC1066 platform. NONE. All of them close in and sometimes tie, but they still don't beat it despite the known DDR low latency compared to RDRAM.
--
This post is brought to you by Eden, on a Via Eden, in the garden of Eden.
i dont understand your reasoning or what you are trying to imply here Eden you are basically saying that beacuse they are using ddr400 it's not good for you, beacuse intel is using ddr266 and is not winning??? LOL, well even with ddr333 the 655 wins a majority of benches over 850E.
DDR400 (pc3200)is a fact. hell you can even get ddr3500 today from your local computer store.
Oh, and I do agree that intel makes highly tuned chipsets. The i850E has been on the market for several months now (I don´t know exactly) and it it still a top performer. Quite impressive. Intel´s chipsets have always been respectable! That´s what makes me so curious about Canterwood... Boy, I can´t even stop mentioning it in my posts!
what impresses me with Granite Bay's Dual Channel DDR266 is that it more or less ties with 850E Dual Channel RDRAM 1060. Looking forward for 800fsb springfield and canterwood but dont forget <A HREF="http://www.theinquirer.net/?article=7944" target="_new"> 659 </A>
It´s going to be one hell of a year for the P4 platform!
Seriously, though, 9.6GB/s max memory throughtput? Looks absurd! The P4 only needs 6.4GB/s with a 800Mhz Fsb...
But as others have remembered here with 2xDDR266 vs 2xDDR400, having more bandwidth for memory offers performance benefits, even IF it would seem the processor can´t take that much info rate.... Anyway, it´s interesting to think about the final showdown, later this year, with the contestants:
<b>Athlon 64</b>, at last, with <b>64-bit extensions</b> and other features, such as <b>Hypertransport</b> memory technology
<b>Prescott 3.6Ghz</b> (or so), 800Mhz FSB, enhanced-performance, on <b>Canterwood</b> with PAT, <b>dual-channel DDR-400Mhz</b>, Doubled caches
<b>Prescott 3.6Ghz</b> as before, 800Mhz FSB, on SiS659 with <b>quad-channel PC-1200 RDRAM</b> capable of 9.6GB/s memory bandwidth.
Which platform will outperform the others and by how much?...
I think Intel have alot in reserve and can up the stakes whenever they feel like it.
For me, the Athlon 64 is too little too late - I fear it's not going to be as competitive as it would've been had they been able to release it 6 months ago. However there weren't any OS's around to really take advantage of it then anyway.
On the surface, it appears Hammer might be a big disappointment. However, I know AMD has a 400FSB Barton 3200+ in the works. I seriously doubt AMD would release any Barton that's faster than Hammer will be, but who knows? I did read that the problem AMD was having with Hammer's integrated memory-controller was that it did not ramp as well as the core itself, but that the problem was minor and has already been fixed. Supposedly, the integrated memory controller, by itself, gives Hammer a 20% speed boost. I've also read that the additional registers Hammer has for 64-bit operation might help in 32-bit mode as well, though I am totally unable to vouch for that.
Let's speculate that without the integrated memory controller, Hammer is as fast, clock cycle-for-clock-cycle, as the 400 FSB Barton. If AMD can release the Hammer in a 2 ghz version, and given the 20% theoretical boost from the IMC, wouldn't it be as fast as a 400 FSB Barton running at 2.4 ghz? And a Barton at 2.4 ghz would be...what? A 3400+? Maybe AMD has something up its collective sleeve, and will release Hammer in a 2.2 ghz version. What would that be? A 3700+?
I am disappointed in AMD the last year or so, and I'm really hoping Hammer is strong enough to keep Intel jumping. I will say the Canterwood chipset sounds very promising indeed. I am very interested to see what a 3.2 gig P4, with double the L1 cache, and with 1 gig of L2 cache, and improved HT, on an 800 mhz FSB can do. Considering the P4's always seem to get a big boost from improved memory performance, it would seem AMD has a tough row to hoe ahead of them. But who knows? Maybe Intel knows something we don't, and is ramping up the P4 because they know they HAVE to to compete with Hammer.
<-----Insert witty sig line here.
<P ID="edit"><FONT SIZE=-1><EM>Edited by Twitch on 03/11/03 01:09 PM.</EM></FONT></P>
For me, the Athlon 64 is too little too late - I fear it's not going to be as competitive as it would've been had they been able to release it 6 months ago. However there weren't any OS's around to really take advantage of it then anyway.
<sarcasm><i>Other than 64-bit x86 versions of Unix and Linux and any version of Windows for a 32-bit x86 CPU (and technically ANY OS for a 32-bit x86 CPU). Yeah. No OS's at all that could have run an Athlon64. It's not like it was intentionally designed to be backwards compatible with 32-bit x86 software. And certainly everyone who runs a 64-bit server does so using Windows and software specifically for Windows.</i></sarcasm>
We really don't need to be making stupid excuses for AMD. Lack of an x86-64 version of Windows is hardly a viable excuse for the continued delay of the Athlon64.
<font color=blue><pre>If you don't give me accurate and complete system specs
then I can't give you an accurate and complete answer.</pre><p></font color=blue>
Listen to me, I am targetting chipset SAME memory performance.
The goal is to see whether the chipset, when run with the APPROPRIATE SYNCHRONISED memory speeds of 266MHZ, can outperform the competition. Otherwise we are still seeing another Pentium 4-like environment where only brute force MHZ saves it. I don't want that in chipsets. Yes with DDR400, you push it basically far and then it can win. But that is INEFFICIENT, AND PC3200 costs much more than PC2100. The goal for consumers is to get the same platform but much better and for less. If the 655 was truly well designed, it'd have required 266MHZ DDR only, dual channeled, at 4.2GB/sec JUST LIKE PC1066 i850E, and outperformed it. THEN it will have a great value and proven that it is damn efficient, and future DDR400 combinations will provide even more boosts.
That is what I wanted, and I was not given once again.
DDR 266MHZ is ALL IT NEEDS, Dual Channeled, to compete PC1066, and it failed to do so.
--
This post is brought to you by Eden, on a Via Eden, in the garden of Eden.
Indeed, I´d tend to agree with you... Dual DDR400 holds true potential for 800Mhz-FSB. A well-designed intel chip for this configuration, a.k.a. Canterwood/Springdale, should perform admirably for a long time, like the i850E did.
About SiS659, as I remember, the 9.6GB/s it offers should fill up the FSB for the Tejas, the next generation Intel processor after prescott, which should have an astonishing 1.2Ghz FSB On the other hand, I think Tejas won´t be out until late 2004 or 2005 or something - if I´m not mistaken.
slvr_phoenix, the 64 bit version of linux is relatively new - certainly not been available for 6 months. Hardly any applications have been ported to run on it. I said no OS's that could take advantage of it, not no OS's that could run it.
<b>Athlon 64</b>, at last, with <b>64-bit extensions</b> and other features, such as <b>Hypertransport</b> memory technology
On the other hand, I think Tejas won´t be out until late 2004 or 2005 or something - if I´m not mistaken.