• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

Pholding on a Phenom

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

AlabamaCajun

Member
Joined
Apr 22, 2005
Location
A Labyrinth
edit said:
(I originally thought I had this at 2.4 but sometime before folding I rebooted and BIOS had returned to stock 2.2. I discovered it after the unit checked in and I rebooted.)

I just wanted to pass along a sneak peak of what a typical Phenom 2.2G fold looks like.
This is AMD's latest arch with 4 Cores.

Project POPC #2653.

phah.JPG


Overclocked 200Mhz the most I can run stable on this board until a good BIOS update gets out or my new ram plays nice. It's a green platform but so far it's almost as fast as my 2.8G 5600 running some apps. I just let it run 4 cores flat out. I know some people are using 2 instances but I think with this architecture 4 might work better unless I reconfigure the ram into split channels which basically sets 2 cores per stick. I haven't tested that yet so it's an unknown.
 
Last edited:
Is this a dual or quad?

My e6400 @ 3.4 gets 14/mins a frame...seems kind of slow. :shrug:
 
I wonder if your gonna get to be able to crank one out in less then a day....
 
Is this a dual or quad?

My e6400 @ 3.4 gets 14/mins a frame...seems kind of slow. :shrug:

Yes at this time a 3.4 Kent Dual does what a 2.4 Agena Quad does. This is the fastest rig I have at this time. I may get a Q9450, s'pose to kick some arse.

I wonder if your gonna get to be able to crank one out in less then a day....

It will be just about one day. Hopefully better with the new ram.
 
If that's a 1760pt WU you are getting just @ 1800ppd on a 200MHz OC. In any case that WU will finish in just under a day.

That's calc'd using 14:03 as a base frame time.

If for some reason that's a p2608/2609 then the ppd will be lower but there's hope that the 1760pt variety will also have a better showing.

Needs to give us the project AC... without that it's just tpfs without meaning. ;)
 
I normally keep an eye out on a certain number for just this reason:
14m 24s * 100 = 1 day

As a reference, my E6600 at stock speeds gets 18m 32s /frame in windows. It was a lot better at 3.6ghz in Linux :rolleyes:
 
Hmmm, I would have expected MUCH better, but I guess with the small cache the SMP WU's cause it to bog. Hopefully some better mempry timings and tweaks will get it down.

Is this native Linux?

I'm installing it right now on my box to see what native brings to the table. I am getting two ~8:40 second frames in each VM under linux now at 4.16 Ghz on the QX9650... I think that's like 5500 ppd.

I think I'll also be getting a Q9450 to replace my wife's "aging" QX6700... Even at 3 Ghz, that should be good for ~4k ppd.
 
Assuming we're seeing frame times from a p2653 (or 2604, p2605, or p2651) running native Linux, Phenom performance looks to be in the same range, clock for clock, as an E6420. Trouble is, it won't clock anywhere close. I expected better and I doubt better ram is going to provide the miracle cure that makes the Phenom anywhere close to a match for an Intel Quad. I've no doubt that the Phenom is a great chip for day to day use, but bang for the buck isn't there as far as folding goes.

@Jon
Running two instances will make less ppd if the cpu is cache bound, as is likely the case.
 
It is a 2653, POPC.

I just wish AMD had the capitol to have brought it but now that Intel is hammering them on all fronts it's less likely we will see a big difference. These beat the Conroe's but fell short of the Kent's. My post is not to prove any "points" but to show others what they can do with AMD hardware. I don't expect to fold much on this rig as it's windows unless I dual boot. For those wishing to Fold on AMD hardware, my suggestion. is to go with the Duals until the updated stepping.
 
When you look at it from the perspective of the machine being your daily use machine, doing what you need done and doing it well, the Phenom will be a great addition. Look at the points as gravy. The Phenom just isn't going to be viable as a farming implement. Have you thought about installing vmware, running two instances and setting affinity on it. That could provide a nice boost in production. Using VMware, affinity sticks to the vmware-vmx processes instead of the cores that restart and lose affinity on every new WU. Also set priority to low on the vmware-vmx processes so they back out pretty much like the native Windows client.

Had I thought about it, with the screen shot it has to be a Windows install and in that case, it's quite a bit faster than a E6420/E6600, clock for clock. 1800 ppd in Windows is nothing to sneeze at.
 
Thanks for all the info on your new Phenom, AC.

It will be great to see how much the L3 cache correction will speed up it up, and where AMD will go from here with their 45nm next generation chips.
 
keep in mind that the Phenom is a 4 core chip! so, comparing it to a 2 core E6420/6600 isn't fair. (I haven't keep track on their price tho....) one should compare it to a X3210/Q6600 at the very least, if not Penryn! @750ppd/Ghz (or so), it doesn't look good at all when compare to Intel's QuadCore chip.....
 
keep in mind that the Phenom is a 4 core chip! so, comparing it to a 2 core E6420/6600 isn't fair. (I haven't keep track on their price tho....) one should compare it to a X3210/Q6600 at the very least, if not Penryn! @750ppd/Ghz (or so), it doesn't look good at all when compare to Intel's QuadCore chip.....

You might have to look at history to see why. These were designed to Kill Conroe Duals as to why people are making those comparison. We knew since the iQ6600 hit the market that the playing field was going to tilt. If it weren't for something that Intel did right in there process, you would not be seeing these overclock that these iQ6600 are getting. Phenom does belong with the Kents. We just hope AMD can match results up to at least 3.2Ghz.

BTW, I have to retest, I shutdown today to change RAM and I found the CPU clock back on Auto that would drop me back to 2.2Ghz or 800ppd/Ghz. for these 1760 pointers.
 
These native Linux results were posted on the Temp FC Fourum:

I've been folding with my Phenom 9500 since Friday. Here is my setup, no OC'ing, all stock speeds.

Phenom 9500
Biostar TF560 A2+
Team DDRII 800,1gig

P3050-3,30,17 10min 10sec
P3059-0,89,92 8min 48sec
P3043-8,25,87 10min 30sec
P3050-9,31,12 10min 4sec

That's an average of 2023 ppd for the 1440 pointers and 1910 ppd for the p3059. p2653 is known to produce more ppd than any of these WUs.
 
I compared the Phenom to the Intel chip (that I have) closest to it in performance. What's not fair about that? Not fair to who? At 3.3 GHz (almost every one of them goes at least that fast), the E6420 makes 2300+ ppd in native Linux. Maybe I should have compared the Phenom to a lesser Intel chip to make it fair? :D

Point is, AMD has a long way to go to equal Intel's cpus as far as folding goes. It's going to be a great chip at everything else.
 
I'm not sure where thread is going with this but AMD may not have that far to go. The only hurdle they really have is getting to 45nm. When test came in clock for clock with similar ranked CPUs the results were not that far apart. People are jumping happy over the success of this monopoly that is purposely attempting to disrupt other business for it's game. Enjoy those chips, what we are doing here is great for the science and fun for us. Overclocking is our advantage that the average consumer does not. A iQ9450 or iQ9550 is of interest to me but the more I see big headedness and bad business practices from the blue one the less I want to invest money in the products. I'll spend more on an AMD just to avoid supporting these practices. I stopped frequenting the Folding are because I got turned off reading about Conroe and Kentsfield. AMD hardware moved a lot of science last year putting me on page 1 and is still doing so.
 
I was alluding to the L3 cache bug that was found in the Phenom. Once they get that ironed out, I think the Phenom should be reevaluated for it's folding prowess. It definitely should be reevaluated when it's brought out at 45nm! :clap:

I remember several years ago, when Intel couldn't even get their 64 bit extensions done right, and finally gave up and bought a license to use AMD's. :)

Intel's done a remarkable turn-around in the last few years, but AMD has also been on the top of the heap, when they first came out with the first (true) dual-core cpu's. I'm sure AMD will remain competitive, and sometimes exceed Intel's offerings, in the future.

That competition between these two is so good for us consumers, that I couldn't imagine what it would be like to just have one commercial cpu provider. I cringe at the very thought of that. :cry:
 
I also found this on the FC Forum:

I have the explanation : this is the reason why actual Phenom, with B2 steppings aren't aviable at 2.4 GHz (model 9700) ...

AMD is waiting for B3 stepping to get higher for a simple reason : they have an issue with the L3 controller at clock highers than 2.3 GHz. That also explains poor overclocking with actual steppings. This issue is a local overheating near this controller that lead to crash like you experienced

I'm pretty sure that if you try to overclock your CPU, things will be worse

If everything goes well, B3 stepping will comes with 2.4+ GHz chips and solve the issue (and overclock higher than 3 GHz).

I'm no Intel fanboy. When AMD builds the best folding cpu, I'll be on it like a fly on sh**. And I'll say if it weren't for AMDs constant pressure, Intel wouldn't be building cpus like Yorkfield and be on such a short product cycle. We all need AMD to be successful to keep Intel in line.
 
Back