backtop

Print E-mail del.icio.us 31 comment(s) - last by Samus.. on Feb 6 at 4:42 AM

BIOS screenshot courtesy of Dave Graham
ASUS L1N64-SLI boots with AMD Opteron 2218 and standard DDR2-667 memory

AMD insider and occasional DailyTech writer Dave Graham posted images of an AMD Opteron 2218 processor running on an ASUS L1N64-SLI motherboard. This is intriguing as the ASUS L1N64-SLI is a motherboard designed for AMD’s Quad FX enthusiast platform. The motherboard BIOS detects the processor correctly and functions properly.

It also appears that the system is using standard DDR2-667 MHz memory instead of registered ECC memory typically required by Socket F Opteron-based motherboards. Although it would seem memory compatibility problems would arise, Graham has images of the system passing Windows Vista’s memory diagnostic tool without error.

With compatibility with AMD’s Opteron 2200-series processors, the ASUS L1N64-SLI may be the holy grail of dual-socket socket-F motherboards. Although the motherboard is pricey, the compatibility with AMD Opteron 2200-series processors opens up AMD’s Quad-FX platform to more budget-conscious buyers as the dual-core AMD Opteron 2200-series lineup starts with the 1.8 GHz Opteron 2210. The Opteron 2210 is also priced around $200 USD per processor, which is $150 lower than the cheapest Athlon 64 FX-70 processor.

The ability to use standard DDR2 memory with AMD Opteron processors may also appeal to workstation users that have resisted AMD Opteron systems due to registered ECC memory requirements. A quick search on Newegg shows the cheapest DDR2-667 modules costing around $70 while the equivalent DDR2-667 ECC registered module costs $130.

Another interesting item to note is the screenshot of the Mainboard tab in CPU-Z. The ASUS L1N64-SLI is based around NVIDIA’s nForce 680a SLI MCP but it is detected as the nForce 570 SLI, revision A1.

Comments     Threshold


This article is over a month old, voting and posting comments is disabled
bluff
By mino on 2/2/2007 4:29:49 PM , Rating: 4
What this means is that FX is really just a rebadged Opteron. It also means that for single moduel/channel usage, the opteron should be just fine with regular DDR2...

Good for customers, sad for AMD...




RE: bluff
By Jack Ripoff on 2/2/2007 5:22:56 PM , Rating: 4
I think it's interesting how AMD is, how may I put it, operating a bit more desperately now (I don't know if this is the best word thou), just like Intel was doing when AMD had the performance advantage (with all that HT and EE stuff and how they were so hurried to ship Prescott it ended up consuming more power than Northwood).


RE: bluff
By Samus on 2/3/07, Rating: 0
RE: bluff
By davegraham (blog) on 2/5/2007 12:10:36 PM , Rating: 2
It's worth noting, however, that the Opterons continue to proceed through extra validation for their roles as "SERVER" products.

Also, this has nothing to do with product announcements or anything else. It was simply a field test of viability. As it stands, I heartily dislike this L1N64-SLI WS board. I much prefer Tyan's excellent S2915 (n6650W) workstation board due to it's design, compliance with SSI spec, and widespread availability.

cheers,

dave


RE: bluff
By Samus on 2/6/2007 4:42:15 AM , Rating: 2
Tyan and supermicro boards have always been lacking complete PCI-X compatibility. I've had two experiences where a RAID controller had to be run at reduced speed or was incompatible because of the AMD HT1000 bridge used on many serverworks platforms.


RE: bluff
By Samus on 2/6/2007 4:42:42 AM , Rating: 2
Correction, AMD-based Tyan and Supermicro boards.


RE: bluff
By Goty on 2/2/2007 6:12:56 PM , Rating: 2
How is that sad for AMD? People have known that A64s are identical to Opterons (and vice versa)for a looong time now... like since the launch of the product family.


RE: bluff
By TomZ on 2/2/2007 6:42:18 PM , Rating: 2
In my opinion, it is deceptive to market the same product as two different SKUs with supposedly different capabilities. The customer that pays more than they should have for the imaginary extra capability eventually feels deceived once the lie is revealed.

Also, remember, not everyone knows, and not everyone can be sure. For example, when I bought my Althlons years ago, I had to buy the special MP version that supported SMP. Even though there were rumors at the time that the SMP and non-SMP SKUs were actually the same product, I couldn't really risk "trying it out" in case it damaged the processors, MB, etc.


RE: bluff
By Furen on 2/2/2007 8:27:57 PM , Rating: 2
Huh? What are you talking about? AMD's Opterons are exactly the same as AMD's K8s, Intel's Xeons are exactly the same as its Pentiums/Core 2s, it's always been this way. Sometimes we see special SKUs (like the Gallatins and, more recently, Tulsa) but for the most part the server parts are rebadged (and repackaged) desktop parts (or the other way around, in the case of the K8, since AMD's focus is on the server side).


RE: bluff
By thecoolnessrune on 2/2/2007 10:00:57 PM , Rating: 4
Mustn't forget however that usually server chips are the highest quality bins you can get. Not that it particularly matters but its still a point.


RE: bluff
By mino on 2/4/2007 2:08:04 PM , Rating: 2
Well, how can you know that the core is not tweaked in some way(or something internally enabled on it) to be able to work in SMP enviroment ?
(i.e. multiple-bus in case of Athlon MP or asymetrically-loaded bus in case of Xeon DP/MP's)


RE: bluff
By MonkeyPaw on 2/4/2007 4:36:15 PM , Rating: 2
I believe the SMP Opterons have extra HT-links enabled that 1-way Opterons and Athlons do not. This is what prevents people from running cheap, 1-way K8s in SMP motherboards. In that case, it's not that AMD added something extra to the "better" CPUs, but rather they disabled something in the cheaper ones. However, I'm not terribly surprised to see that Opterons can run non-ECC memory, as I'm sure AMD has designed the IMC to handle both memory types so they could use their CPUs in any product line. ECC sacrifices speed for added stability, and really most memory is very stable for the average user. ECC memory just adds extra assurance for critical computing environments.


RE: bluff
By davegraham (blog) on 2/5/2007 12:14:53 PM , Rating: 2
There's less of a speed sacrifice than you think (usually on order of 3-5% at a given clock speed). Again, the Opterons benefit more from lower latency than anything else. So, having a CAS 4 DDR2-667 ECC/Reg. part (none exist) would be more beneficial than a CAS 5 part. *shrug*

Opteron 2000 series parts have 2 HT links (processor to processor and processor to NB/SB). Opteron 8000 series parts have 3 links. The price adjustment has to do with the ability to maintain signal integrity across ALL links. It's not a penalty as such. Have you seen how much Xeon 7000 series parts are selling for? *grin* For n-way builds, I'm more than delighted to spend the extra $$$ to ensure that I/O capabilities aren't bogged down by MCH translations, etc. Point to Point is a very, very efficient topology to use.

cheers,

Dave


RE: bluff
By decapitator666 on 2/5/2007 10:02:53 AM , Rating: 2
I still have my asus am266 SMP board working with 2 athlon XP 1900+ processors. Those palominos XPs(?) still work fine eventhought they were officially not workinng in SMP mode


RE: bluff
By Zandros on 2/2/2007 8:03:26 PM , Rating: 2
Not really. The Sledgehammer and Clawhammer cores differed in that the Sledgehammer had a dual channel memory controller where the Clawhammer did not.


RE: bluff
By Furen on 2/2/2007 8:30:26 PM , Rating: 2
Clawhammer also had the dual-channel memory controller, and all the extra HT links, they were just disabled because of the packaging (it doesn't matter what you throw into the die if there are no pins for it). AMD didn't redesign the core just to get rid of the extra memory channel or the extra HT links, it just disabled them afterwards.


RE: bluff
By mino on 2/4/2007 2:04:29 PM , Rating: 2
It means that they were not able to tweak the core a bit to work with 2 modules unbuffered as X2's do.
It is nice they were not forced to do it since Opteron on itself works with single unbuffered module/channel.

I also makes the whole QuadFX platform just a PR stunt which some people thought it was not until this information.

You know I gave them a bit of credit as the platform (thanks to the ASUS uber-board) is actually pretty usable and makes sense.

However this means they actually did NOTHING except for asking ASUS to make exceptional board for them.

Sad is the fact that they
a) have to resort to such tactics
b) intentionally misled the public
c) the only thing the actually can take credit for is comomming up with the idea

This IS sad for a company claiming leadership in the industry.


RE: bluff
By koss on 2/4/2007 11:54:39 PM , Rating: 2
Not really a fact.
AMD did not misguided anyone, claiming quad is here 'pumped' 2x2. And it ended up favouring their theory.
Although performance were more server oriented, the product was limited to a higher bidget due to the current line pricing/positioning in the market.
And what is wrong with making this opteron compactible. If it fits, right price category range it could be a 'tweak' the gamers, and low pocketed 'rollers' will surely go for.
I think this is just a timed speculation, so it would grab attention before the next product hits the shelves. And it better be fast and good, because Intel is really shining at the moment.
So no fooling around AMD is just there fot the fans.
We love AMD.


RE: bluff
By rebturtle on 2/3/2007 12:40:02 PM , Rating: 2
quote:
opteron should be just fine with regular DDR2...


Since the memory controller is integrated into the processor, it's unlikely that this will be an option.


RE: bluff
By rebturtle on 2/3/2007 12:41:12 PM , Rating: 2
Oops, never mind...


By bunnyfubbles on 2/2/2007 5:29:00 PM , Rating: 2
There will always be people who will pay the ridiculously absurd premium for the FX line, but in order to keep a large core of enthusiasts in your boat you have to offer a reasonably priced solution. $400 for 4 cores is enticing even if it cannot keep up with the likes of a C2Q, the point being it would cost a good $400 less for the processor(s).

At this point only the extreme pro-AMD/anti-Intel zealots are adopting new AMD platforms outside of the extreme budget, the Core 2 line really is untouchable by anything AMD - except for an affordable solution such as this, only it doesn't officially exist...




By Brane2 on 2/3/2007 12:51:15 AM , Rating: 2
AMD's real problem is greediness.

They have hyped whole K8 as a modular solution that finally solves multiple cores problem. Where one core isn't enough, one could simply plug in extra CPU and a couple of RAM sticks.

In reality, at least for average mortal, this was never the case. AMD was charging outrageus fine for en extra HT link, god help you if you wanted two.

Only big players could afford that kind of machines.

And now, when Intel's C2D killed them in core performance, they need that capability, so they are trying to push new chips and boards.

SocketF is finally what S940 should have been in the first place. Flexible, usefull with an RAM and old/new HT links. One socket rules them all.

Had AMD done that with S940 (one scoket fits all), I wouldn't have to shell $$$ for expensive dual socket server board and use registered ECC RAMs I don't need and I could easily stay with AMD.

Just dishing the old 240s and plugging in new 290s or some never equivalent would be solution for me.

But now I have to part with expensive boards and RAMs just to byu new boards for which I don't even know how long will they last.

Just as before, AMD might change its mind again and drop support for some key component in chipset or change power envelope etc.

And btw, having _ONE_ board, costing $400 to show for all "4x4&FX" platform is not exactly promising...

Shouldn't enthusiast components be cheaper than high grade server ones ?
What on that L1N64 board justifies its cost, compared to e.g. Tyan's dual socketF S2927 or even S2932?

http://www.tyan.com/products/html/thundern3600b.ht...

Sure, S2927 is leaning more to the server side, but is _VERY_ decent desktop board, and for a lot less $$. Not to mention S2932, which has roughly 4x L1N64's RAM slot capacity...

So, why is "4x4" so expensive ?


By Alexvrb on 2/3/2007 6:15:41 PM , Rating: 2
Well for starters, the 4 PCIe x16 slots. Much like the board you linked to, they actually only have 16/8/16/8 lanes, but that actually shouldn't matter much as there's little chance of quad SLI being efficient enough to keep them all fed constantly. Anyway, if you're actually going to put 4 fast graphics cards in it, two expensive dual-core (and eventually quad-core) CPUs in it, tons of RAM, and lord knows how many HDs, the mainboard doesn't look that costly anymore.

But for most people, of COURSE the platform is ridiculous. With that being said, I'd still keep it on the radar when there are more boards/chipsets supporting it, and quad-core chips can be purchased for it.


By Brane2 on 2/3/2007 6:48:20 PM , Rating: 2
True, but two extra PCIe interfaces for a $200 extra is a bit thick.

It's just a matter of a bit more modern chipset and a couple of connectors ferchristsake.


OTOH that Tyan is optimised for WS/Server use and it has a couple of extras itself ( like PCI-X ports etc) and still sells for $100 less...

Besides, not everyone needs 4 cards in machine and so for 99% of users Tyan could be perfectly fine (minus registered RAMs)...

What I am saying is that "4x4" is fine, especially with Barcelona, but it carries _hefty_ price premium for no obvious reason.



By mino on 2/4/2007 2:13:39 PM , Rating: 2
Actually the Tyan board falls pretty much in the same category as 4x4. The actuall price as of now is different only because of their different age and different dealer prices.
QuadFX as it stands is OC-capable, huge storage-capable and _cheaper_ solution than Opteron 2000 so you are crying on the wrong grave.


By Alexvrb on 2/5/2007 2:23:05 PM , Rating: 2
It is not just interfaces. No single chipset can yet support that many lanes. So they're got not one but TWO modern Nvidia chipsets. That thing is packed to the brim.

AGAIN, I don't think its worth buying or using (yet? ever?), but its price tag is not unimaginable considering what it is packed with, and its Quad SLI Dual CPU RAID and More RAID target audience.


By JeffDM on 2/4/2007 10:27:03 PM , Rating: 2
It's basically the market. AMD can charge the money because they've got a competitive product on the top end. I'm not an AMD fanboy either, I don't have an Athlon or Opteron system, I'm just saying that AMD does have a system architecture advantage at the very high end. Intel has a processing power advantage but that's no good if you need to pump a lot of raw data through the system.

So very few people need dual socket computers, and that the validation it takes for such a system, it's going to be fewer customers to spread that cost over. That validation is a lot more complicated for an eight-socket computer, for an even smaller niche. I'm not convinced that the enthusiast market really needs a dual CPU socket platform. I really think AMD is better off pricing those into the commercial server/workstation market.


As above..
By Regs on 2/2/2007 5:00:51 PM , Rating: 4
All these AMD articles I just skip over now. If it isn't about Barcelona then I'm not reading anything new. And new is the only good news that will come from AMD right now.




RE: As above..
By Furen on 2/2/2007 8:32:09 PM , Rating: 3
Oh, but it is about Barcelona, to some extent. This means that any 2-way Opterons that are released (in Q2, for example) could conceivably be able to run on this motherboard without registered RAM as long as there is BIOS support.


nForce 680a = 2x 570
By Anonymous Freak on 2/3/2007 1:42:15 PM , Rating: 2
I thought it was public knowledge that the "680a" chipset was really just two 570 SLI chips; one for each socket.

I know I read that in various reviews when the '4x4' platform came out. That's why it has 4 Gigabit LAN connections, four 16X PCI-E connections, etc. It has double EVERYTHING that a "normal" chipset would have, because it is really just two separate chipsets on one motherboard.




Clarification
By davegraham (blog) on 2/5/2007 2:26:07 PM , Rating: 2
I know this is late in coming, but...

just would like to point out that I'm NOT the first person to do this. AndreYang and S7 over at XS provided some early testing on the platform before I got started. However, to my knowledge, I'm the only one testing F2 opterons (with my F3s just showing up on Friday of last week). Anyhow, just wanted to give first credit to whom it is due.

dave




"Young lady, in this house we obey the laws of thermodynamics!" -- Homer Simpson

DailyTech Poll
Which web browser do you use on your primary personal machine? 






44 Comments









botimage
Copyright 2009 DailyTech LLC. - RSS Feed | Advertise | About Us | Ethics | FAQ | Terms, Conditions & Privacy Information | Kristopher Kubicki