Ars Technica's God Box

venturi said:
...and yes QUAD SLI is OFFICIALLY supported in the whql 91.31 64-bit drivers.

Why are you still here? Seriously. Thought you were "signing off."
First of all, not what I said by a longshot. Not that something like that seems to stop you. Secondly, what, do you think cutting and pasting that poorly formatted list from some website instantly makes you the most helpful person here? :rolleyes:
Please, take the hint already. We have enough of this kind of useless crap in the CPU, video and PPU forums.
 
For what it's worth, Tyan has a new BIOS out that's supposed to enable support of the 7950 on the K8WE.

I don't have a 7950, and I'm not home to try out the BIOS...but...oh well.
 
sharkapult said:
For what it's worth, Tyan has a new BIOS out that's supposed to enable support of the 7950 on the K8WE.
I don't have a 7950, and I'm not home to try out the BIOS...but...oh well.

Which is all fine and dandy till you realize the only reason there's any BIOS updates for the K8WE is because of all the furious customers. It's still a board I won't sell, especially when the 7950 worked without changes on the H8DCE apparently.
 
AreEss said:
Which is all fine and dandy till you realize the only reason there's any BIOS updates for the K8WE is because of all the furious customers. It's still a board I won't sell, especially when the 7950 worked without changes on the H8DCE apparently.


Your distaste for the board is well documented.

But, many of us K8WE owners bought it because, at the time, it was the only option available. And while the H8DCE maybe a superior solution now...I'm not going to shell out another $400 to replace a board that is working for my current needs. So....whether it's from furious customers or benevolent engineers (ha!) we'll take whatever updates we can get.
 
God I love Supermicro (had to put that out there).

Can't wait for their damn Opty blades to come out, my rep is tired of me asking...they have been dangling that carrot for WAY too long.

As to these GOD systems...you can't use the power...you can't. You can try...but really...you can't. Yes, the spec list would look nice in your sig...but...you...can't...use...the...power. :D
 
AreEss said:
Which is all fine and dandy till you realize the only reason there's any BIOS updates for the K8WE is because of all the furious customers. It's still a board I won't sell, especially when the 7950 worked without changes on the H8DCE apparently.

You have any more info or links on the 7950 and H8DCE working together?
 
hardwarephreak said:
God I love Supermicro (had to put that out there).

Can't wait for their damn Opty blades to come out, my rep is tired of me asking...they have been dangling that carrot for WAY too long.

Fuck the blades. The 1U 4-ways with HE's and EE's have better density. :D

As to these GOD systems...you can't use the power...you can't. You can try...but really...you can't. Yes, the spec list would look nice in your sig...but...you...can't...use...the...power. :D

Oh, I can. But I'm weird like that. I mean, come on, I used to bitch about having 'only' 256MB on my old P233MMX.
 
Ooh, wasn't aware they had a lineup close to ready'ness.../me takes a look closely
 
Cript said:
You have any more info or links on the 7950 and H8DCE working together?

Still nothing concrete or official yet. I'm waiting to hear back on 4500X2 testing - same damn card, just a 400% markup.

NulloModo said:
If I was going to build my own uber-rig, I would wait for the new dual socket Woodcrest + FB-DIMM boards.

I wouldn't. Price disadvantage is absolutely absurd, especially when you factor in memory costs. The FBDIMMs aren't where they need to be for me to start pushing them to my customers. Not enough advantage and too much cost.

And I could still run a single PCI-E video card on a x8 slot correct? As long as it can handle dual-link DVI I would be happy.

Not unless it's a 16x physical. And I would have to check; last I heard, nVidia disabled dual DVI output on 8x mode operation. Could be wrong, but I'd test first.
 
Well if I was to personally pickup a Woodcrest board I would have to go with this board...

http://www.supermicro.com/products/motherboard/Xeon1333/5000X/X7DA8.cfm

The funny thing is, if you look at some of the FB-DIMM boards , a couple years ago you would have thought someone had photochopped the DIMM sockets on the board...1/4 of the board layout is DIMM sockets.

I haven't really begun to push FB-DIMM stuff either...of course that might change when the chips start becoming more available later this month. ES Woodcrests don't really count since I can't sell what I am evaluating. I am sure there will be a few people interested in picking up a few boxes, but to be honest I don't really know how things will end up. Opterons have been king for awhile now, it will be interesting to see how easy it is to peel people away from their Opterons. I can remember trying to sell customers on Opterons before they were fashionable...Some corporate and government types can only see blue.
 
hardwarephreak said:
I haven't really begun to push FB-DIMM stuff either...of course that might change when the chips start becoming more available later this month. ES Woodcrests don't really count since I can't sell what I am evaluating. I am sure there will be a few people interested in picking up a few boxes, but to be honest I don't really know how things will end up. Opterons have been king for awhile now, it will be interesting to see how easy it is to peel people away from their Opterons. I can remember trying to sell customers on Opterons before they were fashionable...Some corporate and government types can only see blue.

There's no way you're going to get the serious customers away from Opterons; especially GIS, CAD/CATIA, etcetera. The hardware is nowhere near where they need; the idea of anyone trying to push the X7DA8 as a workstation board makes me laugh. Seriously. 16x/4x? That's not even a joke. And won't drive Quadros, either. (They will NOT play at 16x/4x split.) The GIS guys would eat you alive on memory costs alone, before you get to the inability to accelerate multiple displays.

Servers? Isolated yeah, grids? Not a chance. Once you say 'HyperTransport,' they don't even want the Intel numbers for comparison. You give 'em the option of 1056MB/sec theoretical versus 1.8GB/sec proven and they stop caring about the color of the logo right then and there. Not to mention the fact that you really cannot make any legitimate case for an Intel-based grid, even with blades, when throughput or latency are a factor. The second you introduce either or both, there's no sound case for not going HyperTransport/InfiniBand.

Think we'll see Woodcrests making some headway in standalone database, email, web, etcetera, and low performance grids, but Intel's way behind the curve here. It'll be at least several months before they have anything worth considering for high end workstations, at the earliest. And gods only know how long before they finally hop on the grid boat properly. (Blades are not grids!)
 
Well if I wanted a board for myself, just to play around with I would still go with the X7DA8. Don't need SLI for what I would use that board for...

But you are right, until Intel gets some on die memory controllers and some faster interconnects people won't go back. I mean, sure they can do some round-a-bout things to improve on the memory controller bottleneck in the meantime...but without a fast interconnect Xeons just don't scale particularly well.

Oh well, Woodcrest would have been a GOD of a chip, had it had those two weaknesses resolved. Is Intel's high speed interconnect tech supposed to show up late, late next year?
 
We are using Opterons in our shop now because of what they have done for our stored procedures. They totally (and I mean TOTALLY) destroyed our Xeons on stored procedure performance. SQL Server runs like a scalded dog on a 256 or higher class Opteron. Put that in the bank.

We only buy Opterons for our database servers now.
 
Aegeas said:
We are using Opterons in our shop now because of what they have done for our stored procedures. They totally (and I mean TOTALLY) destroyed our Xeons on stored procedure performance. SQL Server runs like a scalded dog on a 256 or higher class Opteron. Put that in the bank.

We only buy Opterons for our database servers now.

I would love to talk to your shop about some of my newest offerings. ;)

Seriously though; that's exactly what all my customers have been saying for some time. When it comes down to anything memory bandwidth intensive, the Opterons crush the Xeons so utterly, nobody can make any legitimate business case for Intel. It's just flat out impossible, and Woodcrest right now, has no hope of changing that. The only place it might show strong is on CPU intensive sorts, but if you're bottlenecking in memory, then you've lost that advantage.
 
Quote:
Originally Posted by mikeblas
Why is it important to be able to open the levers on two adjacent RAM slots at the same time, by the way? Sorry to ask about such a trivial issue, but it's the only specific claim we've got against Tyan so far.
Quote:
Not at the same time.
PERIOD.
You cannot open those DIMM sockets where they meet no matter what. Not at the same time, not one at a time. The end.
AreEss,
you said awhile ago that the s4885 and/or the s4881? could not open the dimm slots one at a time and I got wary about the k8qe (s4885) becuase yes any board that is not mechanically functional is probably not electronically fuctional. But I had the pleasure of finding a s4882 (or the s4882D but I dont remember which one..) quad socket board at my local fry's electronics and I checked the dimm slots. It is possible to open them one at a time and I would have to argue that one would not need to put two sticks in at the same time. Also, if people are have problems with puting in sticks of memory in a 1400 dollar mobo, they shouldnt be doing it in the first place. Putting in memory is about the easiest thing to do to a mobo and if you cannot put in one stick of memory then close the dimm slots, and then put in another stick of memory without damaging the adjacent stick.. just dont tuch computers period. I know that what I saw was not the s4885 but the designs are very simmilar. When I do find a s4885 that I can look at.. I will inspect it and report back.


Quote:
Have you ever worked with PCI, PCI-X or PCI-Express qualification or bus analyzers? No?
Then it's all over your head. The faulty timing, the SERR/PERR glitching, the irregularities on the 3.3V lines, the inadequate VRMs, and so on.

What Im really intested in though is your pcix (16x) analyser readouts by attachment and include the names of the program and hardware you used to measure them with. It would be helpful if you could show the faulty timing and the serr/perr glitching in the attachment.

Also, do you think it was possible that you just had a faulty VRM?

And you do say, "and so on", what exactly are you talking about?

The reason I am asking these questions? I am interested in the s4885 becuase of the dual 16x pcix slots with the octa socket opterons, not to mention the sata 2 controler. IF and thats a big IIFFFF this board checks out.. I think this board would totally rock.

Thx
Some shmoe that would like to know...

software suggestions:
opera 8.52 (the only perfectly safe browser)
xp/ corp ed with xplite with no sp installed but usb 2.0 compliant
(oh yes I installed xp w/o it ever calling home)
rom2
outpost
tcpview
ethreal
total unistall
shredder
more if you want....
 
Back
Top