Note from Vijay re: SMP server issues

[Tobit]

Below is a quick note from Vijay Pande in response to all the noise being made by donors regarding the availability, or lack thereof, of the primary SMP server @ 171.64.65.56

We're working on this on several fronts: new SMP WUs to go onto other servers to spread the load and an improved WS software to handle the load better. Both are moving to resolve this issue, but are taking longer than we expected. I hope that several PG members will be beta testing new SMP WUs shortly (hopefully later this week). Once they're through the QA process, the load will get *a lot* better on this server. Also, the new WUs will likely run off of the new WS software, which should help too (the new WS has code to specifically handle the situation when the # of connections saturates).

 

[Nathan_P]

Thanks for that Tobit, I haven't been affected as much as some of the guys but its good to know. For those that can run them, any update on the lack of -bigadv units???

 

[Tobit]

For those that can run them, any update on the lack of -bigadv units???

Three factors most likely (IMO):

• The current SMP WS issues causing bottlenecks up the chain.
• More people than ever before crunching bigadv.
• Slow i7's and X6's not returning results fast enough so new WUs can be generated. If you look at the server stats, there are close to 3000 bigadv WUs currently being crunched.

There is a new bigadv project coming that may help alleviate some of this but I do not have an ETA.

 

[Doozer]

Does this mean that I may actually pull a 3rd bigadv with my L5640 one day?

 

[Tobit]

Does this mean that I may actually pull a 3rd bigadv with my L5640 one day?

Just think happy thoughts and you will be rewarded..

The bigadv'll come out
Tomorrow
Bet your bottom dollar
That tomorrow
There'll be bigadv!
Just thinkin' about
Tomorrow
Clears away the cobwebs,
And the 670x's
Til there's none!

 

[nwrtarget]

Three factors most likely (IMO):

• The current SMP WS issues causing bottlenecks up the chain.
• More people than ever before crunching bigadv.
• Slow i7's and X6's not returning results fast enough so new WUs can be generated. If you look at the server stats, there are close to 3000 bigadv WUs currently being crunched.

There is a new bigadv project coming that may help alleviate some of this but I do not have an ETA.

Only here does any i7 get called "slow"!

SMP WU seems to have gotten more stable for me recently although it could be that evga is scaling back down and not burning up the connections as fast as they were. Their points have dropped off a ways from the peak.

Also if bigadv units become much more common evga may have a much harder time getting in front or staying in front based on the number of bigadv machines here vs. their more gpu heaviness.

 

[Tobit]

Only here does any i7 get called "slow"!

Slow as in i7's that aren't OC'd [H]ard enough to turn the bigadv in faster. There are many out there who think just because they have 8-threads, they can run bigadv but ignore their low clocks.

 

[nwrtarget]

Slow as in i7's that aren't OC'd [H]ard enough to turn the bigadv in faster. There are many out there who think just because they have 8-threads, they can run bigadv but ignore their low clocks.

Oh I know I won't run -bigadv on my single i7 920 at stock clocks but for most any other application even at stock clocks the i7 920 is a monster! Hence my comment.

Of course most people can't fathom running dual hex cores and then overclocking the heck out of them either. That is why we are different as we actually need all of those cores and cycles we can get.

Anyway the i7 has been a monster for some time and since AMD is now finally going to challenge them they have to move on and release the next step Sandybridge. They have had plenty of time on it.

LGA 1356 is the socket of interest it would seem which according to this roadmap
http://en.wikipedia.org/wiki/Sandy_Bridge_(microarchitecture)
isn't available until 2012 and is marked as tentative. This doesn't look good for the next generation SR2 as it would have a separate socket not allowing for the easier upgrade path we have now.

I was hoping for a quad channel desktop board but that looks improbable for the next few years.

 

[flyz]

Oh I know I won't run -bigadv on my single i7 920 at stock clocks but for most any other application even at stock clocks the i7 920 is a monster!

Here's my story.

Before I got back into Folding (I last folded in 2007?), I too thought that a 4GHz 920 was a monster. It was amazing when it was released in Nov 08 and still is 2 years later... for most things.

In the short 2 months that I've been folding again, I upgraded to an Extra Special chip, found my old 920 a home by upgrading my file server and just today purchased a 970 to shuffle my Special chip to my file server.

Never in a million years would I think that I would need 2 highly clocked hex core machines

 

[Vaulter98c]

Or two OC'd hex's in one rig

 

[sirmonkey1985]

Oh I know I won't run -bigadv on my single i7 920 at stock clocks but for most any other application even at stock clocks the i7 920 is a monster! Hence my comment.

Of course most people can't fathom running dual hex cores and then overclocking the heck out of them either. That is why we are different as we actually need all of those cores and cycles we can get.

Anyway the i7 has been a monster for some time and since AMD is now finally going to challenge them they have to move on and release the next step Sandybridge. They have had plenty of time on it.

LGA 1356 is the socket of interest it would seem which according to this roadmap
http://en.wikipedia.org/wiki/Sandy_Bridge_(microarchitecture)
isn't available until 2012 and is marked as tentative. This doesn't look good for the next generation SR2 as it would have a separate socket not allowing for the easier upgrade path we have now.

I was hoping for a quad channel desktop board but that looks improbable for the next few years.

no the next socket thats of interest is the LGA-2011 sandy bridge coming out in the 2nd quarter of 2011.. it will have the quad channel DDR3 to compete against AMD's G34 socket quad channel DDR3 and bulldozer.. but we all knew LGA-1366 was a dead socket almost 6 months ago..

 

[DWolvin]

Just think happy thoughts and you will be rewarded..

The bigadv'll come out
Tomorrow
Bet your bottom dollar
That tomorrow
There'll be bigadv!
Just thinkin' about
Tomorrow
Clears away the cobwebs,
And the 670x's
Til there's none!

That is the best maul of a song I have heard all day! +1(dunno- internet/kitty/bj)