New A3 Project - 6040

E

EvilAlchemist

2[H]4U
Joined
Jan 11, 2008
Messages
2,730
New Project released to -advmethods

Project 6040
- 1395 points,
- 4.63 k-factor
- 6 day preferred deadline
- 8 day final deadline

These projects are 4x larger then current A3 units.

Those users with dual cores processors may take several days to complete these.
(Example - Xeon 3075 (Core 2 Duo @ 2.66Ghz) takes 0h, 38m, 28s per frame / 1,946 PPD (0.37 WUs Per Day)

Those with Core i7 systems should see a small boost in PPD with these units.
 
How's you dually do on these?
Posted via [H] Mobile Device
 
Not a bad addition to the list of WU's we already are working with.
 
One step closer to A3 -bigadv.

I'm going to have to get a second DSL line for all this crap.
 
Kendrak said:
One step closer to A3 -bigadv.

I'm going to have to get a second DSL line for all this crap.
Click to expand...

Yeah, all the uploads are killing me :(

2 Big Adv, 4 SMP2 & 10 GPU2 clients. They are just uploading all throughout the day :(
 
capreppy said:
Yeah, all the uploads are killing me :(

2 Big Adv, 4 SMP2 & 10 GPU2 clients. They are just uploading all throughout the day :(
Click to expand...

The -bigadv shut down my connection for the most part.
 
Yeah, pretty much sucks when I am in the middle of a TDM in MW2 and I don't check if there is an upload about to hit. Sad Panda, zero bars and I'm getting picked off like chicken feed.

ON topic, are these bonused like the other A3's? I've not seen one yet, but am looking forward to them.
 
capreppy said:
ON topic, are these bonused like the other A3's? I've not seen one yet, but am looking forward to them.
Click to expand...

From keeping an eye on Evil's Stats it seem these give a boost of 4-5k ppd over the standard smaller units. (compared to what he was running prior)
 
Cool... My current wu is getting ready to upload. Maybe I'll get lucky. :D

Maybe it will make up for the 24+ hour a1 I just completed this morning. :(
 
1500 ppd drop on my Phenom 2 with these units compared to other A3s according to HFM. Is HFM not counting bonus points correctly for these, or are they just brutal in AMDs?

 
musky said:
1500 ppd drop on my Phenom 2 with these units compared to other A3s according to HFM. Is HFM not counting bonus points correctly for these, or are they just brutal in AMDs?
Click to expand...

I have noticed these units love hyper-threding cores. My Core 2 Quad based systems did seem a very small drop.

I did read some reports that Intel hardware does perform better on these.

I am pretty sure they are still using the intel compiler for the Windows Core.

For AMD hardware, you *may* want to test a linux based option since it is open source.

Not sure if it will help or not.
 
EvilAlchemist said:
I have noticed these units love hyper-threding cores. My Core 2 Quad based systems did seem a very small drop.

I did read some reports that Intel hardware does perform better on these.

I am pretty sure they are still using the intel compiler for the Windows Core.

For AMD hardware, you *may* want to test a linux based option since it is open source.

Not sure if it will help or not.
Click to expand...

Keep in mind this was a P2 system that got 5200 ppd normally, so this is an extreme drop in ppd. I don't really care since I am scrapping both my AMD systems in favor of Intel-based systems. It just seems like kind of a crummy move on Stanford's part to cripple AMD owners even more than they already are with F@H.

 
Project ID: 6040
Core: GRO-A3
Credit: 1395
Frames: 100


Name: Daviserver - Dual Xeon E5530 w/ HT no Turbo, 2008 R2
Path: \\daviserver\folding\smp
Number of Frames Observed: 300

Min. Time / Frame : 00:09:22 - 16,183.6 PPD
Avg. Time / Frame : 00:09:57 - 14,781.5 PPD


Name: HTPC - AMD X4 620, win7 x86
Path: \\htpc\folding\smp
Number of Frames Observed: 300

Min. Time / Frame : 00:25:49 - 3,536.7 PPD
Avg. Time / Frame : 00:26:18 - 3,439.7 PPD
 
musky said:
Keep in mind this was a P2 system that got 5200 ppd normally, so this is an extreme drop in ppd.
Click to expand...

Compairing A2 PPD vs A3 PPD does not work well.

Two different projects and totally different benchmark systems.

That new Core i5 benchmark system would also drop PPD on A2 if they went back and redid those units.

musky said:
It just seems like kind of a crummy move on Stanford's part to cripple AMD owners even more than they already are with F@H.
Click to expand...

I would not place to much blame on stanford for the compiler.

From my undersading, the agreement between AMD & Intel a few months ago should help this situation once the newer compilers are released. This is one factor that has hurt AMD.

Not to bash AMP products, but many benchmarks have shown them lagging when compaired to Intel products of the same class.
 
EvilAlchemist said:
Compairing A2 PPD vs A3 PPD does not work well.

Two different projects and totally different benchmark systems.

That new Core i5 benchmark system would also drop PPD on A2 if they went back and redid those units..
Click to expand...

The 5200 was from A3 units, not A2s. This system did well over 6K on A2's. I took another 1K+ ppd drop going from A2s to A3s.

I would not place to much blame on stanford for the compiler.

From my undersading, the agreement between AMD & Intel a few months ago should help this situation once the newer compilers are released. This is one factor that has hurt AMD.

Not to bash AMD products, but many benchmarks have shown them lagging when compaired to Intel products of the same class.
Click to expand...

I blame Stanford for making a scoring system that is making AMD PCs hardly worth the electricity to run them strictly for F@H anymore. The last two "upgrades" to the SMP units (A3s and this p6040 A3) have both cut the production of this AMD system that is only 3 months old. That seems ridiculous to me. If all I had was the two AMD systems like I did when I started with the Horde back in December and saw my ppd numbers drop by over 30%, I would really be pissed. If my numbers stayed the same through these last two changes, I would be fine with that. It is the lose of production I am seeing with relatively new hardware that ticks me off. F@H is a completely voluntary thing. You really shouldn't screw over your volunteers in such a project. Go ahead and tell me it is about the science and finding a cure for cancer or whatever. In reality, as soon as you have a "dedicated folding box", you are in it for the points, with the side effect of helping with the project. Don't kid yourself into thinking otherwise. Look at it this way, how many people would have dedicated boxen running right now if they didn't keep score? It is a brilliant move on Stanford's part. Points don't cost them much of anything, yet having them increases the amount of work put into the project by a significant amount.

OK, I am done ranting now...back to getting my latest I7 stable to replace my two AMD boxen that are going to go off-line today... :)

 
musky said:
I blame Stanford for making a scoring system that is making AMD PCs hardly worth the electricity to run them strictly for F@H anymore.
Click to expand...
It has nothing to do with the scoring system. A3 units perform worse on AMD CPUs than Intel CPUs, period. The same scoring system is used for both Intel and AMD systems; the only determining factor is how fast each CPU processes the workunits, and in that respect, AMD CPUs can't match up at the moment (in part due to the lack of optimization for AMD CPUs in the Intel compiler that is used for F@H, as EvilAlchemist mentioned).
 
The compiler only has so much to do with it tho.

The fact is that AMD cpus just arent up to snuff.

Anandtech did a performance comparison to AMD Magny Cours 12-core vs Intels Westmere 6 core and the magny cours underperforms. Plain and simple.
 
Back to the original topic, these units move us one step closer to getting -bigadv units to A3.

I hope the units keep getting bigger as time goes on to really push my dual xeon setups
 
nitrobass24 said:
The compiler only has so much to do with it tho.

The fact is that AMD cpus just arent up to snuff.

Anandtech did a performance comparison to AMD Magny Cours 12-core vs Intels Westmere 6 core and the magny cours underperforms. Plain and simple.
Click to expand...

Keep in mind that the current Phenom 2 architecture has about the same IPC as the Kentsfield based Intel quad cores. Match a Phenom 2 and something like a Q6600 with the same clockspeeds and you should get similar performance in most things. You can assume that any lower performance you see from Phenom 2s compared to something like a Q6600 at the same clockspeeds has to do with the compiler.

 
It has nothing to do with the scoring system. A3 units perform worse on AMD CPUs than Intel CPUs, period. The same scoring system is used for both Intel and AMD systems; the only determining factor is how fast each CPU processes the workunits, and in that respect, AMD CPUs can't match up at the moment (in part due to the lack of optimization for AMD CPUs in the Intel compiler that is used for F@H, as EvilAlchemist mentioned).
Click to expand...
It's not just F@H, it's across the board. AMD needs a better architecture, and they need it fast. Athlon was awesome, Phenom was a huge flop. Adding L3 cache to Phenom makes it better, but that doesn't automatically make it magical. AMD need to start working to improve their architectures and improving their performance in other ways or else they'll fall too far behind.
 
It seems I finally got one of these. They definitely do cause a bit of a hit in PPD on "older" systems. I have one running on a [email protected] and it's only doing 6000PPD whereas the system will usually do around 7000PPD with the normal A3 work units. I'm hoping I don't see many more of these on any of my systems. My Linux folder already doesn't do much better than 6200PPD on A3s running a [email protected]. One of these would murder the PPD on that machine. I also don't see my main machine running a [email protected] doing too well with these.

It seems that you're probably going to lose PPD on these work units unless your running i5 or better Intel architecture.
 
Interesting. My Opteron system finally broke the 2000 PPD mark in SMP2 with a P6041. Before it was hovering around the 1500 PPD mark and now seeing ~700 PPD better as long as I'm not doing anything with the computer that's CPU-intensive. Interesting.
 
EvilAlchemist said:
Is that with Hyper-Threading On or Off? Is this a dedicated system or one you use daily?
Click to expand...

its a soo called dedicated. i post and look at stats. no gaming? yes hyperthreading is on. i run this with a gpu client. i get around 17,000 with a 470 and 500's pointers 16,0000 with 353's and 768's 18,000 without gpu client.these get 14,000 with the gpu client running 470's . this should show up on my 4pm stats. http://folding.extremeoverclocking.com/user_summary.php?s=&u=45203
 
Last edited:
It looks like your GPU is killing the 6040 / 6041 performance.

Remember that this is a mid-sized work unit so bonus points really affect your PPD.

Adding a few seconds on the 6012-6025 (small WUs) don't hurt as bad as these because 6040's are worth so many more points when that bonus is factored in.

The larger the units get, the more important Time per frame comes in.
 
EvilAlchemist said:
It looks like your GPU is killing the 6040 / 6041 performance.

Remember that this is a mid-sized work unit so bonus points really affect your PPD.

Adding a few seconds on the 6012-6025 (small WUs) don't hurt as bad as these because 6040's are worth so many more points when that bonus is factored in.

The larger the units get, the more important Time per frame comes in.
Click to expand...

I didn't know that. Since I have an AMD processor, I guess i'll have to be glad that I keep getting assigned smaller WU(mostly 6012, 6013's and on 6023) once I finally pass 10 SMP work units folded with my passkey so that I get a better bonus.
 
I blame Stanford for making a scoring system that is making AMD PCs hardly worth the electricity to run them strictly for F@H anymore. The last two "upgrades" to the SMP units (A3s and this p6040 A3) have both cut the production of this AMD system that is only 3 months old. That seems ridiculous to me. If all I had was the two AMD systems like I did when I started with the Horde back in December and saw my ppd numbers drop by over 30%, I would really be pissed. If my numbers stayed the same through these last two changes, I would be fine with that. It is the lose of production I am seeing with relatively new hardware that ticks me off. F@H is a completely voluntary thing. You really shouldn't screw over your volunteers in such a project. Go ahead and tell me it is about the science and finding a cure for cancer or whatever. In reality, as soon as you have a "dedicated folding box", you are in it for the points, with the side effect of helping with the project. Don't kid yourself into thinking otherwise. Look at it this way, how many people would have dedicated boxen running right now if they didn't keep score? It is a brilliant move on Stanford's part. Points don't cost them much of anything, yet having them increases the amount of work put into the project by a significant amount.

OK, I am done ranting now...back to getting my latest I7 stable to replace my two AMD boxen that are going to go off-line today...
Click to expand...
It's the same as with AMD GPUs. The chips have different architectures and it's only natural that they have different performance. And it's also fairly predictable that Stanford would optimize the code for Intel, because they hold such a huge proportion of the market of consumers that are likely to run SMP that it just seems crazy to waste time optimizing for AMD as well.

As for points, that's what I've been arguing for a very long time now. It is not fair, and they probably know it's not fair.
 
EvilAlchemist said:
It looks like your GPU is killing the 6040 / 6041 performance.

Remember that this is a mid-sized work unit so bonus points really affect your PPD.

Adding a few seconds on the 6012-6025 (small WUs) don't hurt as bad as these because 6040's are worth so many more points when that bonus is factored in.

The larger the units get, the more important Time per frame comes in.
Click to expand...

I stopped my gpu client for a few hours. they just dont make the points the smaller wu's do
 
lehmann said:
I stopped my gpu client for a few hours. they just dont make the points the smaller wu's do
Click to expand...

Stopping the GPU for a few hours will help some, but you have already dropped your bonus on that work unit due to the delay. Wait for a new one to start and keep the GPU off.

Then get a PPD reading on the next unit.

Every Core i7 based system I have got a ppd boost on Project 6040 vs the smaller 6012-6025 projects. (All 5 of them)
Only my Core 2 Based systems showed a drop in PPD since they are below the benchmark system.
 
Last edited:
EvilAlchemist said:
Stopping the GPU for a few hours will help some, but you have already dropped your bonus on that work unit due to the delay. Wait for a new one to start and keep the GPU off.

Then get a PPD reading on the next unit.
Click to expand...

i used the times per fram calculator web page. thought i will double check myself next time i see one.
 
P6041, I7 @ 3.57, smp 8, 10:13 frame times, 14140 ppd. This is around 700 ppd less than the same machine on smaller A3's. No GPU client, Win7 64, doing nothing but folding, 20% through the unit. Moral to the story, I am still seeing a ppd drop on these bigger A3's, even with a decent I7 system.
 
musky said:
Moral to the story, I am still seeing a ppd drop on these bigger A3's, even with a decent I7 system.
Click to expand...

6040 / 6041 are medium size units.
If you are seeing a drop in ppd now, you will not be happy when the Large units or the A3 -bigadv units come out later.

I am seeing the exact oppoiste on my Dual Xeon systems. The larger the units, the more my PPD goes up.
Guess having those extra cores are paying off.

I have not seen those PPD drop on my single X5550 system so I dont' have any advice to give
 
My E2140 is currently doing about 1850PPD on a 6040 compared to 1400-1500PPD on smaller A3 units.
 
EvilAlchemist said:
6040 / 6041 are medium size units.
If you are seeing a drop in ppd now, you will not be happy when the Large units or the A3 -bigadv units come out later.
Click to expand...
I'm a bit concerned about A3 -bigadv. How well will it run on older hardware? Will I see similar performance as the A2 WUs or will it be poorer like I'm seeing with the current SMP2 WUs?
 
You must log in or register to reply here.
Back
Top