Sneak Preview: Intel's Alderwood/Grantsdale Chipsets Reviewed

L

Labrador

[H]ard|Gawd
Joined
Jan 19, 2002
Messages
1,930
http://www20.tomshardware.com/motherboard/20040301/alderwood-05.html

wow this review from Th, makes it sound like the new Alderwood + 775pin Prescott, wont be a major leap forward you would expect ? I mean the benchmarks make it look just ok ?

I may stay with my current mobo/cpu in my sig, and just pop for either the NV40 or R420, which ever is the best bang for the buck :) I am an adult, not some lame ass tool fanboy:rolleyes:
 
Why would it be a major leap anyways? There's no architectural changes made to the Socket 775 Prescott, just a change in packaging. Additional memory bandwidth would do little in the vast majority of benchmarks and real-life applications. What's left? Some nice features and a much longer upgrade life. For buyers of new systems in a few months, that's reason enough to go with an Alderwoood/Grantsdale system.
 
whoa... the latencys don't look too good on that DDR2, huh...

other then that... I'm excited to see how all this new stuff coming out at once will fair.
 
I'm disappointed, Tom doesn't mention anything about if S775 makes Prescott cooler or not, only mentions superflow of the BTX's fan form factors.

As for DDR-II, its one step forward in bandwidth for two clock steps back for latency. The DDR-II implementation is like settling for 3.73 gears in your 400hp hotrod that you race on weekends; everyone else has 4.99's but you made your choice for the gas mileage and highway cruising speeds during the rest of the week... A 4-bit prefetch is not going to hide DDR-II's dreadful 0-60 time. Sure its a hotrod in the 1-mile, but its a rather boring performer when it comes to quickness. Sounds like Intel wants to push memory technologies that cripple their competition's performance and runs very fine on their upper end products that use long pipelines.
 
It does seem like "the competition" doesn't have DDR-II in its short-term plans, though.
 
AMD has no choice in the matter if the memory producers sway to DDR-II.
 
Originally posted by xonik
And that will take how many years exactly?
Click to expand...

.75. AMD will have DDR-II support in the fourth quarter most likely. They're in no hurry though, seeing as DDR-II will be of no gain at this time over DDR.
 
No, I'm talking about how long it'll take for the memory producers to sway over to mostly DDR-II production. I would expect AMD to be a little more timely, but I doubt the transition to DDR-II is a very pressing matter to them, considering the technology's flaws right now.
 
thank god there's gonna be grantsdale w/ddr1 support..
But i wonder... how much will be 915 crippled then.. to make
925 look as a better product?
based on this data, if 915 = ~875 then where does that leave
ddr2-only 925?
IMO ddr2 will start to make sense from ddr2-667 speeds..
and that requires 333mhz fsb speed..
any fsb bump planned before tejas? not that i've heard of :(

But i DO like the pinless cpu and new socket..
and pci express, too.

Intel should've been continuing supporting rdram.. for uber-high end desktops..
that XDR thing looks promising.. and 64-bit rdram (quadchan) aswell..

But industry likes ddrII.. with 0.11 process, u'll get more chips
per wafer, with 2x lower clocks u'll get higher yield..
and with 2x higher prices.. u'll get more profit :)
 
looking at what ddr2 has to offer makes me wish rambus took over with the 820 chipset 4 years ago :/
 
Intel 775-pin = mainboard makers cringe

Intel basically is passing on the problem with bent pins over to the mainboard manufacturers. Its not Intel's problem if their pins get bent. I cringe thinking of how much of a PITA the mainboard RMA process can be...
 
The Intel Extreme Graphics 3 look interesting. Something good to have on any board if you ask me. I dont know at the times i've had to come up with a new AGP card cause the fan was pissing out on my other one. But with integrated graphics i could just switch it over to that till i got the video card replaced.
 
Originally posted by burningrave101
The Intel Extreme Graphics 3 look interesting. Something good to have on any board if you ask me. I dont know at the times i've had to come up with a new AGP card cause the fan was pissing out on my other one. But with integrated graphics i could just switch it over to that till i got the video card replaced.
Click to expand...

This is exactly why integrated graphics will be an option in Alderwood (something Intel has never done with a flagship chipset): easing the transition (for users) to PCI Express.

Surprise #2: Intel makes GigE standard, but eschews the prpprietary (but scarily-effective) CSA for the PCI Express bus. However, the new iCH6W southbridge is very interesting as it also supports all three speeds of Wi-Fi while retaining the wired-GigE option of both ICH6R and plain ICH6. Will some motherboard makers include both g-class WiFi *and* wired GigE for desktop WAPs that double as desktop PCs?

Non-surprise #1: Neither Grantsdale or Alderwood are revolutionary in any sense; both can be said to be *evolutionary* as opposed to revolutionary (The only real ground broken is standardization of Gigabit Ethernet at the low end.).

Non-surprise #2: LGA775 is not a surprise as this effectively closes off Prescott to non-Intel chipsets (while VIA, and SIS currently make S478 chipsets, and VIA's PT800 even supporting Prescott, I doubt that either company can legally support LGA775). Quite honestly, I like that the new processors are essentially *pinless* (the pins are in the socket, as opposed to being on the processor), as the pins themselves will likely be thicker and less likely to bend due to the new location, and it will actually make processor upgrades (in the same form factor) *easier* for newbie DIY builders (and unless the big companies like Dell and Gateway start gluing the processors to the retention mechanism, you will actually be able to buy a faster Prescott for your BTX Gateway *cow-box* without having to send the computer back to Gateway or paying their ridiculous markup).
 
I doubt that Intel and VIA do not have an agreement for the S775. VIA and Intel agreed on a patent exchange to settle their long running feud over the P4's bus.
 
Yup, they all got licence for 800 PSB.
Any fsb bump will require licence renewal..


Originally posted by MadRat
I doubt that Intel and VIA do not have an agreement for the S775. VIA and Intel agreed on a patent exchange to settle their long running feud over the P4's bus.
Click to expand...
 
I doubt VIA's technology exchange did not cover the new S775 or any other format for the next ten years.
 
guys read the review the second to last page details via and sis chipsets to compete with alderwood and grantedale. so even though i dont recall it saying that they definatly have lga rights. its a pretty good bet they doi.
 
"LGA rights"--just what does that mean to you? Did you know that Land Grid Array packaging has been available for a while now in other applications? It's just an electromechanical interface between two large circuits (the CPU and motherboard). As long as Intel gives the electrical specifications of the CPU and socket (which it will do for its motherboard partners), chipset manufacturers will be able to adapt to socket changes without needing "rights" or whatever.
 
Originally posted by PGHammer
Surprise #2: Intel makes GigE standard, but eschews the prpprietary (but scarily-effective) CSA for the PCI Express bus.
Click to expand...

CSA was a just a second copy of the Hublink bus dedicated to GigE because of the limited bandwidth available (266mb/s). It was just a stopgap until Hublink could be replaced.
 
does anyone else have the feeling that intel just ran into a huge wall?

there's no way im putting a prescott into a confined case, same goes with that ddr2 sham.
 
Originally posted by Finsta
I wonder if ICH6W will have the features of ICH6R(SATA and RAID wise)
Click to expand...

The plain-jane ICH6 supports SATA standard (as did ICH5), but does not support native RAID like ICH6R and ICH6W (in addition to ICH5R). The ICH6 southy is primarily aimed at Gransdale (not Alderwood); Alderwood northies should appear with either ICH6R (or W) soutnies (R adds native SATA RAID, and W further adds native WiFI in all three speeds).
 
Yup... I can't help but see Intel as a bug hitting the windshield on car going 100 kilometers/hour.

DDR2 looks like it might be another Rambus, expensive as heck and not any faster. PrescHot is slow, but at least its not expensive. The jury is still out on PCI-Express and BTX, but personally I still like ATX, and don't see a huge need to moveover.

The LGA775 is really going to piss off the mainboard manufacturers. Yeah, the complexity of pins will now be up to the mobo manfacturers to tackle. Under the new system, if you break/bend pins you will have to replace/repair the motherboard instead of replacing/reparing the CPU..

On the upside, it will definitely keep all the computer repair technicians happy (they get to charge much more labour to replace a mobo than a CPU)
 
Originally posted by xonik
"LGA rights"--just what does that mean to you? Did you know that Land Grid Array packaging has been available for a while now in other applications? It's just an electromechanical interface between two large circuits (the CPU and motherboard). As long as Intel gives the electrical specifications of the CPU and socket (which it will do for its motherboard partners), chipset manufacturers will be able to adapt to socket changes without needing "rights" or whatever.
Click to expand...

why the assholeish tone, anyone reading that should have known exactly what i was meaning. that from reading toms article both via and sis have rights to make chipsets compatible with the new socket.why is it some people around here just like to post shitty posts to show how smart they think they are.
 
Originally posted by ZenOps
Yup... I can't help but see Intel as a bug hitting the windshield on car going 100 kilometers/hour.

DDR2 looks like it might be another Rambus, expensive as heck and not any faster. PrescHot is slow, but at least its not expensive. The jury is still out on PCI-Express and BTX, but personally I still like ATX, and don't see a huge need to moveover.

The LGA775 is really going to piss off the mainboard manufacturers. Yeah, the complexity of pins will now be up to the mobo manfacturers to tackle. Under the new system, if you break/bend pins you will have to replace/repair the motherboard instead of replacing/reparing the CPU..

On the upside, it will definitely keep all the computer repair technicians happy (they get to charge much more labour to replace a mobo than a CPU)
Click to expand...

The same thing was said about ATX (by most techs) compared to AT (remember, before ATX appeared in droves, you could get Slot 1 AT formfactor mainboards, in addition to Socket 7/Super 7).

Have any BTX formfactor tower cases (of any size) appeared yet?

Quite honestly, it was the sheer size of the ATX midtower (compared to an AT mid) that convinced *me* to switch (and because I generally don't overclock the CPU, I have *never* lost a processor due to heat).

ATX midtower cases are larger all around (and have more hand room in the interior) than an AT case of the same general size-type. The biggest problem with ATX cases isn't even the case's fault: it's the wide latitude ATX *mainboard* designs differ from each other or Intel's Reference Standard Design (and even Intel has broken the rules it laid down a few times). The biggest problem (again, it's more a mainboard design issue than a case design issue) is socket/connector placement (especially for memory and drives)...some mainboard designs are so badly implemented that they force you to pull the board out of the case to add/upgrade RAM.

BTX, on the other hand, standardizes things more than ATX, which should alleviate some of these *bonehead* design ideas.
 
Originally posted by jeef
looking at what ddr2 has to offer makes me wish rambus took over with the 820 chipset 4 years ago :/
Click to expand...

Well the 820 had horrid latency issues. Until Intel figured out that Dual Channel technology would help to alleviate most fo that.
 
No, the i820 didn't have latency issues, the first generation RDRAM it used did. A dual channel configuration does nothing to help or hurt latency; it just spans the memory bus across the bus widths of two memory modules.
 
i820 had the latency issues with three memory sticks. They disabled all but two memory slots to compensate for the problem.
 
You learn something new every day...as trivial as it may seem...
 
I actually understated the problem. Since it had to do with trace lengths from controller to memory, the double-sided sticks had problems, too. The pair of memory slots had to be populated with the cheap stuff else they got lost in logic delays built into the controller that Intel built in to overcome worst case scenarioes. Moving to dual-channel allowed them to squeeze more memory into each stick and the controller was redesigned to handle each pair of slots independent of the other. That is why RDRAM has had to be used in matched pairs.
 
You must log in or register to reply here.
Back
Top