• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Speculation: Haswell upgrade for Folding

T

thesecond

[H]ard|Gawd
Joined
Jan 15, 2010
Messages
1,083
Sup gents!

My X58 rig is getting quite long in the tooth, and all of these tantalizing declarations of increased IPC and efficiency are rustling my jimmies!

I'm folding in windows, using HFM.net and F@H GPU tracker, got the spicy e5649 running at 3.8ghz with a gtx480 running stock, and averaging around 30k ppd a day.

Anybody think a transition to Haswell when it comes out could increase my PPD whilst decreasing my power consumption?
 
If you plan on OCing then Haswell might not be the right choice for you.
Since Ivy Bridge intel will be using TIM inside the IHS* so there is a good chance that Haswell will have the same cooling issues as IB, and Haswell also moves the VRMs off the mobo and onto the CPU so Haswells TDP is higher then IB so it is likely that Hasell will OC worse then UB.

Now if you get normal Xeons that can't be OC'd then Haswell will be a good choice.

Also Haswell-E and presumably the Xeons will use DDR4, which will be expensive and if it goes like the DDR3 launch may be a bit slower to begin with.

*This is because they have been having issues with Un-leaded solder cracking when used on 22nm to smaller dies.
 
Is 30k ppd combined between a 6 core CPU and the gtx 480? Seems low, my 980x gets 32k ppd at 4ghz.

as far as haswell, don't know
 
You are looking at cumulative IPC improvements from SB, IB, and Haswell which will estimate is a total 35 to 55% increase depending on the workload. Also you are probably going to run a higher clockspeed so even with the reduction in cores from 6 to 4 you should expect a modest improvement. Lots of variables here of course so I'd love to hear results if you make the upgrade. Definitely an improvement in efficiency though.

Have you considered a new video card? Your 480 is a model of inefficiency compared to newer cards.
 
Spazturtle said:
If you plan on OCing then Haswell might not be the right choice for you.
Since Ivy Bridge intel will be using TIM inside the IHS* so there is a good chance that Haswell will have the same cooling issues as IB, and Haswell also moves the VRMs off the mobo and onto the CPU so Haswells TDP is higher then IB so it is likely that Hasell will OC worse then UB.

*This is because they have been having issues with Un-leaded solder cracking when used on 22nm to smaller dies.
Click to expand...

Yes, i will wish to OC because thats half the fun, and i do realize the problems with intel's "Secret Sauce" so to speak. Guess i'll have to consider delidding. But hey, even if it is limited by temps, people were breaking 4.5 ghz, which in addition to IPC improvements should mean better PPD than my current xeon amirite?

Ultima99 said:
You are looking at cumulative IPC improvements from SB, IB, and Haswell which will estimate is a total 35 to 55% increase depending on the workload. Also you are probably going to run a higher clockspeed so even with the reduction in cores from 6 to 4 you should expect a modest improvement. Lots of variables here of course so I'd love to hear results if you make the upgrade. Definitely an improvement in efficiency though.

Have you considered a new video card? Your 480 is a model of inefficiency compared to newer cards.
Click to expand...

And yes, i have considered a new GPU for sure. But I will probably be making the purchase all at once for convenience sake. I have my eyes on a Titan for their mad compute power, but we will see.
 
Oh hell yes even haswell at stock should give you much better PDD.

A 3770k has a TDP of 77W, a 4770k has a TDP of 84W (due to the VRMs being on chip), so the CPU will run hotter, runs hotter = less oc.
So expect a de-lided 4770k to OC lower to what a de-lided 3770k OC'd to.

For comparison my 3770k OC'd to 5GHz and got 50K PPD.

Also de-liding haswell may be hard, as there are components mounted on the CPU.
intel-haswell.jpg
 
Last edited:
Spazturtle said:
Oh hell yes even haswell at stock should give you much better PDD.

A 3770k has a TDP of 77W, a 4770k has a TDP of 84W (due to the VRMs being on chip), so the CPU will run hotter, runs hotter = less oc.
So expect a de-lided 4770k to OC lower to what a de-lided 3770k OC'd to.

For comparison my 3770k OC'd to 5GHz and got 50K PPD.

Also de-liding haswell may be hard, as there are components mounted on the CPU.
Click to expand...

I don't mean to call you out in a negative way, but unless you are willing to show some screenshots as proof, I highly doubt you had a 3770K @ 5Ghz stable getting 50K PPD my friend..

I have a 3700K de-lidded @ 4.8Ghz with DDR 2133 ram and only see 25~32K, with my 480 adding another ~30K PPD..
 
ccityinstaller said:
I don't mean to call you out in a negative way, but unless you are willing to show some screenshots as proof, I highly doubt you had a 3770K @ 5Ghz stable getting 50K PPD my friend..

I have a 3700K de-lidded @ 4.8Ghz with DDR 2133 ram and only see 25~32K, with my 480 adding another ~30K PPD..
Click to expand...

I sold the system a while back and got a 2P g34 system.
25-32K seams very low, are you running on windows or linux?
And I had no GPU folding in the system so no driver to use CPU time.

EDIT: What anti virus do you have? I have heard some slow down FAH CPU folding.
 
Last edited:
I never thought about the possibility you were running in Linux..I am running the current V7 client in Windows..With the new Core 17 GPU client, the cpu usage is 2~3% at most..I was surprised it has gotten so low..

The only AV I use is pretty much what most everyone here uses, which is MSE..I have never had any issues with it slowing things down in the past..
 
I have a 3770K (with it's lid still on) that will fold at 5GHz,
but best efficiency is at 4.5 w/stock volts so that's where it's at.

This was my PPD for Project 7003 (from February), running V7.2.9 on Win7-64 at 4.5GHz, with a GTX 650 also folding.

Code: 
 Project ID: 7003
 Core: GRO-A4
 Credit: 600
 Frames: 100


 Name: Evoa Slot 01
 Path: 127.0.0.1-36330
 Number of Frames Observed: 300

 Min. Time / Frame : 00:01:37 - 45,813.25 PPD
 Avg. Time / Frame : 00:01:41 - 43,118.82 PPD

Depending on WU mix, 50K at 5GHz would definitely be possible.

ccityinstaller said:
The only AV I use is pretty much what most everyone here uses, which is MSE...
Click to expand...
Evoa is also running MSE.
 
Amaruk said:
I have a 3770K (with it's lid still on) that will fold at 5GHz,
but best efficiency is at 4.5 w/stock volts so that's where it's at.

This was my PPD for Project 7003 (from February), running V7.2.9 on Win7-64 at 4.5GHz, with a GTX 650 also folding.

Code: 
 Project ID: 7003
 Core: GRO-A4
 Credit: 600
 Frames: 100


 Name: Evoa Slot 01
 Path: 127.0.0.1-36330
 Number of Frames Observed: 300

 Min. Time / Frame : 00:01:37 - 45,813.25 PPD
 Avg. Time / Frame : 00:01:41 - 43,118.82 PPD

Depending on WU mix, 50K at 5GHz would definitely be possible.


Evoa is also running MSE.
Click to expand...

Thanks for posting this info..I am curious what it going on with this rig..I might need to uninstall everything and do a fresh install...The 480 is coming out (hopefully) tomorrow so I can quit staring @ my 7950 propped up next to my LCD:p..

Do you remember what your temps/voltage were @ 5Ghz? My cpu uses a very low Vcore @ 4.8Ghz compared to most I have seen, which is why I was hesitant to crank the voltage for another 200-300mhz..

I have plenty of thermal headroom assuming I don't hit a voltage wall, as the hottest core runs 64C, with the rest in the 58-62C range with my AP-15s running @ 7V..I am a noise freak, but since I have gravitated to using headphones for everything, noise really doesn't matter..The system is quiet @ 12V anyway so I can get the temps down another 2-3C if needed..

EDIT:

Are you guys running any special flags for the SMP client? Since -BigAdv went away for us little guys, I have only been running the client type -adv flag..Is there something I forgot to add?
 
I have a 2600k @ 4.2 and never see over 20k ppd for it. Looking at some of your numbers I am thinking I have something wrong.I am running V 7.3.6 on Win64. I keep the CPU at 100% folding too. (while afk).

I do have Bitdefender running.
 
It all depends on the OS, WU, and other processes running on the CPU.

Linux will get higher PPD than Windows running the same WU.

Processes fighting the core for resources (this includes any GPU folding) will cause load imbalance which requires dynamic load balancing or other threads to wait.

In spite of PG's benchmarking effort, there is still a good bit of PPD variation project to project and even some variation between WUs in the same project.

As a comparison, my 3770k at 4GHz in Linux (with minimal non-folding loads) runs in the 30-34k PPD range (average - there are some major outliers).
 
You must log in or register to reply here.
Back
Top