Folding CPU?

[freakcrap]

Starting to research into getting the best folding cpu for about $200, but not really sure where to go, any takes? Here is what I have come up with so far...

AMD Athlon 64 3400+ Newcastle 2.4ghz $205
intel Pentium 4 540J Prescott 3.2ghz $210

I really dont understand the difference between the 2 when it comes to folding, so I'm hoping that somebody out there can tell me

Going only by speed, the P4 is the clear winner, but I'm not sure if I need to be looking for something else...L1 cache? HT? or maby there is a different processor I should be looking at?

 

[Bill Clo]

AMDs are better at some varieties of the folding work units, Intels are better at others. Most people use AMDs, by and large. But if all you have is an Intel, by all means, contribute anyways.

If you look at the link I posted earlier, in the post your hot deals, you'll see where you can get that AMD 2.4ghz Newcastle for $173... here: http://www.pcmicrostore.com/PartDetail.aspx?q=c:36124;pl:1;b:270;p:10502244 oops, the price jumped up to $188. Got mine for the $173 price.

With AMD vs Intel, you can't look at the ghz rating alone, as they operate in very different ways. Not that long ago, an AMD XP2700 was pretty much equivalent to an Intel 2.7 ghz. Things have changed with the A64 series, so it's harder to tell. Much easier to compare them with benchmarks. Gaming or whatever software you plan to use on it. It's harder to benchmark Folding, as there is no consistent way to do so.

I'd get the AMD, personally. I have 5 machines (6 shortly), and 5 of the 6 are AMDs (Intel was a gift)...

 

[freakcrap]

Thanks for the 411 on it. I swear by AMD but heard that the Itel might be better, was just checking on it. All the machines I build have the AMD. Havent touched an Intel since my old, *cough* gateway *cough*

 

[SmokeRngs]

You will find that most times, AMD is a better folder overall. It doesn't choke on Tinker core proteins and isn't that much slower, albeit somewhat slower, than Intels in most cases on Gromacs. I haven't seen my P4 at work (first one I have ever played around with) crunch on an Amber core protein although my AMDs seem to like them a lot. Right now, AMDs aren't allowed to crunch on QMD cores so no one really knows how they will perform on those and they would only be able to be run on A64s since that core needs SSE2.

I would personally go with the AMD. I prefer them over Intels and they seem to be better folders when it's all said and done I think.

 

[BillR]

Actually for your $200 I’d look at the 64/939 chips. The Winchester or Venice cores can be had for that kind of budget. The Venice is newer with better mem controller and in the 3200 flavor will do 2.5 gig or better on air cooling.

Keep in mind the 939 boards will also accept the new dual core CPU’s that are on the way so you have an upgrade path to a “Dually” when your ready.

Just some thoughts.

 

[SmokeRngs]

Actually for your $200 I’d look at the 64/939 chips. The Winchester or Venice cores can be had for that kind of budget. The Venice is newer with better mem controller and in the 3200 flavor will do 2.5 gig or better on air cooling.

Keep in mind the 939 boards will also accept the new dual core CPU’s that are on the way so you have an upgrade path to a “Dually” when your ready.

Just some thoughts.

I concur with this thought also. Even if I had the money, I don't know if I would go A64 right now since what I want is dual core. If I knew I could afford a dual core CPU later I'd probably go with Venice for now. But then again, I would also need money right now to do upgrades so this is not a worry of mine.

 

[Hito Bahadur]

I'm saving for a X2 for the next box right now. For people paying for electricity it makes sense. Especially if it ever goes down in price.

 

[gnewbury]

Although all of my homebuilts are now AMD (best bang for the $), when you compare an A64 and a P4 you are comparing a 64bit system runnning 32 bit software versus 32bit running 32bit. The extra power of the A64 is pretty much wasted.
Like I've posted in the past - my AMD's are putting out about 500PpGhz per week. My p4 is almost at 1,000PpGh per week, on QMD's.
Stanford tunes first for Intel, then for AMD, then for Mac's.

Looking at Price Watch:
A P4 3.0 costs about $150
An A64 3000+ costs about $150
An Athlon 3000+/Sempron 3000+ costs about $100

 

[Hito Bahadur]

gnewbury,

You are lucky to be getting QMDs all the time.

My Athlon 64 computer is getting around 780 PpGhpW (Newcastle)

My Celeron J gets around 500 PpGhpW

My Sempron S754 gets around 700 PpGhpW

My Athlon XPs get around 650 PpGhpW

My Pentium M gets anywhere from 431 PpGhpW for Tinkers to 1200 PpGhpW for 300+ point Gromacs

 

[EinsteiN]

My main machine

AMD Athlon XP 2500+ Barton @ 2.205 GHz with 1 GB RAM

here's a snap from qd from this week(pretty average week btw ):

Index 2: finished 149.00 pts (4.620 pt/hr) 32 X min speed
Index 3: finished 600.00 pts (9.485 pt/hr) 31.1 X min speed
Index 4: finished 364.00 pts (5.170 pt/hr) 17.7 X min speed
Index 5: finished 241.00 pts (7.750 pt/hr) 52.5 X min speed
Index 6: finished 241.00 pts (7.729 pt/hr) 52.3 X min speed
Index 7: finished 241.00 pts (6.468 pt/hr) 43.8 X min speed
Index 8: finished 241.00 pts (6.813 pt/hr) 46.1 X min speed
Index 9: finished 241.00 pts (6.564 pt/hr) 44.5 X min speed
Index 0: finished 600.00 pts (8.921 pt/hr) 29.3 X min speed
Index 1: folding now 210.00 pts (4.592 pt/hr) 31 X min speed; 62% complete

Average pph: 7.037, ppd: 168.89, ppw: 1182.2, ppy: 61686

-E

 

[OSUguy98]

Here is some info on the CPUs I'm running right now... Take a look at the 2 P4 2.66s in the "moving average"... my "Work PC" is running big packets and getting a steady diet of QMDs.... while the other PC here at work (identical), doesn't have big packets enabled... HUGE margin there...



Keep on Folding!! For the [H]orde!!

 

[gnewbury]

Here's a quick qd analysis of my 2.8Ghz Dell

Index 9: finished 450.00 pts (14.200 pt/hr) 32.6 X min speed
Index 8: finished 450.00 pts (14.742 pt/hr) 33.8 X min speed
Index 7: finished 450.00 pts (14.477 pt/hr) 33.2 X min speed
Index 6: finished 450.00 pts (12.382 pt/hr) 28.4 X min speed
Index 5: finished 450.00 pts (14.931 pt/hr) 34.2 X min speed
Index 4: finished 450.00 pts (14.872 pt/hr) 34.1 X min speed
Index 3: finished 450.00 pts (14.822 pt/hr) 34 X min speed
Index 2: finished 450.00 pts (14.601 pt/hr) 33.5 X min speed
Index 1: folding now 450.00 pts
Index 0: finished 450.00 pts (14.024 pt/hr) 32.2 X min speed

The few low points were when someone shut it down.
/edit -
and as far as power usage - the PC is sucking 120W at full fold.

 

[Bill Clo]

Points are very well and good, but please concentrate on which gets the work units done, and the science advanced first.

 

[SmokeRngs]

Points are very well and good, but please concentrate on which gets the work units done, and the science advanced first.

The reason some of the WUs have higher points compared to how long it takes to crunch them is because of two things 1. Stanford wants these proteins back as quickly as possible and as many done as possible and so they award higher points value to achieve that goal. 2. Many of the big point value proteins also use a lot more resources than older proteins. Therefore, not everyone can run them and Stanford awards extra points to those who can to give an incentive to get people to grab them that can run them.

It's not completely about greed for points that we look for the best folders on certain proteins. Stanford does not award higher points for certain proteins just because they want to. They have a reason for it and we are just taking advantage of that by looking at what will crunch those proteins better.

On that note, I don't believe that these big point proteins will be around forever or they may come out with a different type that will not work as well on systems that before sliced through the big proteins. I look for the best overall folder that seems to get everything done in a decent time. That is why when I'm looking at a processor for folding I go AMD. For the most part they go through the majority of proteins rather well. They do choke on some of the smaller gromacs it seems and can't do QMDs right now, but they slice through Tinkers about twice as fast as a P4 does. From what proteins I have chewed through, that seems to make up any difference that AMDs get penalized on anything else. These are just my personal observations and opinions.