Deadlines way too short on 8001, 8004, 8011 [Not]

Moderators: Site Moderators, FAHC Science Team

7im
Posts: 10179
Joined: Thu Nov 29, 2007 4:30 pm
Hardware configuration: Intel i7-4770K @ 4.5 GHz, 16 GB DDR3-2133 Corsair Vengence (black/red), EVGA GTX 760 @ 1200 MHz, on an Asus Maximus VI Hero MB (black/red), in a blacked out Antec P280 Tower, with a Xigmatek Night Hawk (black) HSF, Seasonic 760w Platinum (black case, sleeves, wires), 4 SilenX 120mm Case fans with silicon fan gaskets and silicon mounts (all black), a 512GB Samsung SSD (black), and a 2TB Black Western Digital HD (silver/black).
Location: Arizona
Contact:

Re: Deadlines way too short on 8001, 8004, 8011

Post by 7im »

Hyperlife wrote:And the deadline using the actual ratio, as Bruce demonstrated, is much more than 3.5 days. So again, to answer your question: no, that is not enough time.

Until a new policy is implemented, the current ratio using a P4 2.8GHz as the benchmark should be used for the calculation of all uniprocessor WUs.
No need to use that, new policy already in place. See these:

http://folding.stanford.edu/English/FAQ-Points
http://folding.stanford.edu/English/FAQ-PointsNew
How to provide enough information to get helpful support
Tell me and I forget. Teach me and I remember. Involve me and I learn.
Hyperlife
Posts: 192
Joined: Sun Dec 02, 2007 7:38 am

Re: Deadlines way too short on 8001, 8004, 8011

Post by Hyperlife »

7im wrote:No need to use that, new policy already in place. See these:

http://folding.stanford.edu/English/FAQ-Points
http://folding.stanford.edu/English/FAQ-PointsNew
There is no strict deadline formula under the new points system -- the deadline can be lengthened or shortened, and the use of the k factor keeps the PPD similar by normalizing it to the benchmark machine.
We may vary the deadline length between projects (some projects require fast completion and thus have short deadlines)
And the benchmark machine, a Core i5, is still very close to the performance of a P4 2.8GHz:
According to our projections, this new benchmarking standard will result in point yields for a 2.8 GHz P4 that are slightly above the typical uniprocessor values
There still needs to be a way to ensure that uniprocessor WUs will not terminate too early on a standard machine. The old formula ensured that didn't happen -- nothing in the new formula does if there's a quick return bonus. So the old deadline formula is the only guide we have at the moment.
Image
7im
Posts: 10179
Joined: Thu Nov 29, 2007 4:30 pm
Hardware configuration: Intel i7-4770K @ 4.5 GHz, 16 GB DDR3-2133 Corsair Vengence (black/red), EVGA GTX 760 @ 1200 MHz, on an Asus Maximus VI Hero MB (black/red), in a blacked out Antec P280 Tower, with a Xigmatek Night Hawk (black) HSF, Seasonic 760w Platinum (black case, sleeves, wires), 4 SilenX 120mm Case fans with silicon fan gaskets and silicon mounts (all black), a 512GB Samsung SSD (black), and a 2TB Black Western Digital HD (silver/black).
Location: Arizona
Contact:

Re: Deadlines way too short on 8001, 8004, 8011

Post by 7im »

Actually, there is a strict formula. Read that page again!!!

The "We vary deadlines" is an exception to the rule, not the rule.

In what bizaro universe is an i5 close to a P4?

Are you really suggesting A4 work units, WUs that run on both UNI and SMP systems, be benchmarked on a P4? Might as well remove the QRB while you're at it!!!

Dude, you're not making any sense...
How to provide enough information to get helpful support
Tell me and I forget. Teach me and I remember. Involve me and I learn.
PeterA
Posts: 59
Joined: Mon Dec 03, 2007 2:28 am
Hardware configuration: PC: AMD Sempron(tm) Processor 2800+, 1024 MB RAM, Microsoft Windows XP (Home Edition) SP3
Laptop: Mobile AMD Sempron(tm) Processor 3500+, 896 MB RAM, Microsoft Windows Vista (Home Premium)
Location: Blaine, MN

Re: Deadlines way too short on 8001, 8004, 8011

Post by PeterA »

One thing that we can't overlook is that some of us don't have 24/7 internet. I downloaded a 8011 yesterday and when I tried send back the results today. I received a " Unit 0's deadline (January 18 13:17) has passed." message. I have to find a library or a coffee shop to access the internet. I'm not going to do that several times a day. I would rather process a WU that has a 2 week deadline then one that has a <24 hour deadline.
bruce
Posts: 20824
Joined: Thu Nov 29, 2007 10:13 pm
Location: So. Cal.

Re: Deadlines way too short on 8001, 8004, 8011

Post by bruce »

bruce wrote:
In the old FAQ, FAH wrote:How do you set the deadlines for the work units?

Each work unit is benchmarked on a dedicated 2.8 GHz Pentium 4 machine with SSE2 disabled. For most work units (although there may be exceptions, described in the next paragraph), we apply this equation:

timeout = 20 * (daysPerWU) + 2 deadline = max(30* (daysPerWU) + 2,10)
timeout = 20*(18 hours/24) +2 = 17 days.

That policy was established when a P4 was a pretty common machine. They're pretty rare now. If we assume there will always be part-time donors, but most of them have faster machines, it's not unreasonable for the Pande Group to establish a new policy that tightens up the deadlines somewhat. IMHO, 17 days is too long and 3.5 days is too short, but either way, it requires a policy change.
As others have said, there's already a published policy for points based on the newer/faster i5 CPU (rather than the P4) which gives about the same uniprocessor PPD as the older policy. I have not seen a published policy for the deadline, though apparently one is being used. Presumably the 20* factor would change, based on a faster CPU. The +2 factor provided extra time for you to connect at the nearest coffee shop and essentially took care of office machines that are turned off for the weekend (suppose a WU was almost finished on Friday night). If you can't finish the WU in whatever the 20*(daysPerWU) gets changed to, your machine isn't fast enough.
Hyperlife
Posts: 192
Joined: Sun Dec 02, 2007 7:38 am

Re: Deadlines way too short on 8001, 8004, 8011

Post by Hyperlife »

7im wrote:Actually, there is a strict formula. Read that page again!!!
The formula is strict, of course; the additional variables cause, well, variation. PG is free to substitute different k values to balance out a change in the deadline time. The old formula had no option to include a k value to counteract a change in the deadline -- you got the same deadline every time you ran the formula.
7im wrote:The "We vary deadlines" is an exception to the rule, not the rule.
So? I'm not pointing out it always happens -- I'm explaining that the option exists. Obviously that option was used with the 80xx WUs, and may have been used with others (I haven't run the math on other WUs). The point is this: variation is allowed under the new formula. It wasn't allowed under the old formula.
7im wrote:In what bizaro universe is an i5 close to a P4?

Are you really suggesting A4 work units, WUs that run on both UNI and SMP systems, be benchmarked on a P4? Might as well remove the QRB while you're at it!!!
Did you even bother reading the quote about PPD equivalency between a P4 and an i5 that I included? For your benefit, I shall quote it again; I hope you read it this time. I'll even include another line from the FAQ that makes the comparison even more explicit:
The new benchmark machine is a Core i5-750 with Turbo Mode off. We compare single-core performance to the old benchmark machine, a 2.8 GHz Pentium 4.
According to our projections, this new benchmarking standard will result in point yields for a 2.8 GHz P4 that are slightly above the typical uniprocessor values
I am not suggesting that A4 WUs be benchmarked on P4s. I am restating what the FAQ states: a 2.8GHz Pentium 4 will get slightly more than 110 PPD from WUs that are benchmarked on an i5-570. Hopefully that is now being made clear enough to you -- apparently others in this thread figured it out without difficulty.
7im wrote:Dude, you're not making any sense...
I'm not making any sense? :roll:
Image
Hyperlife
Posts: 192
Joined: Sun Dec 02, 2007 7:38 am

Re: Deadlines way too short on 8001, 8004, 8011

Post by Hyperlife »

bruce wrote:As others have said, there's already a published policy for points based on the newer/faster i5 CPU (rather than the P4) which gives about the same uniprocessor PPD as the older policy. I have not seen a published policy for the deadline, though apparently one is being used. Presumably the 20* factor would change, based on a faster CPU. The +2 factor provided extra time for you to connect at the nearest coffee shop and essentially took care of office machines that are turned off for the weekend (suppose a WU was almost finished on Friday night). If you can't finish the WU in whatever the 20*(daysPerWU) gets changed to, your machine isn't fast enough.
Makes sense to me. Sounds like a rational way to approach it.

If there is a deadline policy, we should probably include it in the FAQ; if there isn't, it would be helpful to have one so these errors don't occur in the future.
Image
7im
Posts: 10179
Joined: Thu Nov 29, 2007 4:30 pm
Hardware configuration: Intel i7-4770K @ 4.5 GHz, 16 GB DDR3-2133 Corsair Vengence (black/red), EVGA GTX 760 @ 1200 MHz, on an Asus Maximus VI Hero MB (black/red), in a blacked out Antec P280 Tower, with a Xigmatek Night Hawk (black) HSF, Seasonic 760w Platinum (black case, sleeves, wires), 4 SilenX 120mm Case fans with silicon fan gaskets and silicon mounts (all black), a 512GB Samsung SSD (black), and a 2TB Black Western Digital HD (silver/black).
Location: Arizona
Contact:

Re: Deadlines way too short on 8001, 8004, 8011

Post by 7im »

There is a hybrid deadline policy for a4s because they are sent out to both single and multi-core systems. Shorter than previous CPU work unit deadlines, longer than previous SMP work unit deadlines. I've already contacted the appropriate parties about getting that information updated/published.
How to provide enough information to get helpful support
Tell me and I forget. Teach me and I remember. Involve me and I learn.
GreyWhiskers
Posts: 660
Joined: Mon Oct 25, 2010 5:57 am
Hardware configuration: a) Main unit
Sandybridge in HAF922 w/200 mm side fan
--i7 2600K@4.2 GHz
--ASUS P8P67 DeluxeB3
--4GB ADATA 1600 RAM
--750W Corsair PS
--2Seagate Hyb 750&500 GB--WD Caviar Black 1TB
--EVGA 660GTX-Ti FTW - Signature 2 GPU@ 1241 Boost
--MSI GTX560Ti @900MHz
--Win7Home64; FAH V7.3.2; 327.23 drivers

b) 2004 HP a475c desktop, 1 core Pent 4 HT@3.2 GHz; Mem 2GB;HDD 160 GB;Zotac GT430PCI@900 MHz
WinXP SP3-32 FAH v7.3.6 301.42 drivers - GPU slot only

c) 2005 Toshiba M45-S551 laptop w/2 GB mem, 160GB HDD;Pent M 740 CPU @ 1.73 GHz
WinXP SP3-32 FAH v7.3.6 [Receiving Core A4 work units]
d) 2011 lappy-15.6"-1920x1080;i7-2860QM,2.5;IC Diamond Thermal Compound;GTX 560M 1,536MB u/c@700;16GB-1333MHz RAM;HDD:500GBHyb w/ 4GB SSD;Win7HomePrem64;320.18 drivers FAH 7.4.2ß
Location: Saratoga, California USA

Re: Deadlines way too short on 8001, 8004, 8011

Post by GreyWhiskers »

PantherX wrote:
GreyWhiskers wrote:...I'm up to 82% now - let's see if it completes....
You did do it successfully:
Hi GreyWhiskers (team 0),
Your WU (P10722 R0 C3382 G0) was added to the stats database on 2012-01-23 03:10:19 for 3009.2 points of credit
8-)
Thanks, PantherX. It's good to see that WU FINALLY finished. By far the slowest one I've ever seen.
:roll: Started on Dec 31 - finished on Jan 23. With "Quick Return Bonus" :!: What a project :!:
Interestingly enough, the next WU on the old Uniprocessor was an 8011 (2,631,39). So, at the moment under v7.1.43 on two computers,
- the old HP Pentium 4/HT 3.2 GHz uniprocessor is getting ~22-24 minutes TPF (projected at 82.2 ppd with a small QRB). Stock clocks with a Core 11 GPU WU in the background on the ATI HD4670 AGP GPU. The timeout is about 51 hours - which we should make OK. Win XP.
- the new i7 2860QM laptop is also processing an 8011 (1,690,26) with SMP 8, with a Core 15 proj 6804 running in the background on the underclocked GTX560M GPU - FAHControl is estimating 1 min 15 second TPF for the 8011, or 7,803 ppd (pretty slow for the SMP WUs). The timeout here is also about 51 hours, so no sweat. It should complete in about 2 hours. Win 7 Home Prem.

That's a ratio of almost 100 to 1 in ppd (94.9:1 to be exact) for the i7 SMP8 laptop vs the Pent IV Uniprocessor running the 8011s. And, that's with the laptop. With a desktop i7, or better, the ratio will be even higher.
rwh202
Posts: 410
Joined: Mon Nov 15, 2010 8:51 pm
Hardware configuration: 8x GTX 1080
3x GTX 1080 Ti
3x GTX 1060
Various other bits and pieces
Location: South Coast, UK

Re: Deadlines way too short on 8001, 8004, 8011 [Not]

Post by rwh202 »

GreyWhiskers wrote: That's a ratio of almost 100 to 1 in ppd (94.9:1 to be exact) for the i7 SMP8 laptop vs the Pent IV Uniprocessor running the 8011s. And, that's with the laptop. With a desktop i7, or better, the ratio will be even higher.
Well, with a desktop i7 getting 30000ppd @ 4.7GHz, the ratio is almost exactly 365:1 !!

Is a day's folding of 30 WUs on an i7 really the same value as steady folding 230 WUs over a year?? I guess it must be!
Post Reply