My impressions of Marks P-D 820

[Duvie]

First off I have to say now that I have seen it first hand I am not impressed at all....I had access to both systems simultaneously with a switch of the KVM monitor switch. IMO the AMD X2 system was noticeably faster with multiple programs (and/or any program that held 100% load) running...The Intel chip felt lethargic. So I decided to investigate and what I found was a bit surprising...

Let me also state right off hand, that the X2 system is at the disadvantage when it comes to cooling...It has only 1 case fan (not including the PSU) and a smaller XP90 (aluminum) with smaller 92mm fan. At full load running S&M 1.73 the heatsink was at 54c load on average an almost unnoticeable to the notch for heat. It also has a disadvantage in that it is running on a 380watt antec while the Intel system uses a Fortron 450 watter....

On the other hand the INtel system has 2 case fans, one frekkin behemoth of a heatsink (typhoon) and a massive 120mm fan....The heat around it as soon as you stuck your hand near it was noticeably verm warm. The heatpipes actually were burning hot.....


Initially Mark had the beast at 3.5ghz, w/ 1.45v...At that speed it had ran apps fine for 2 days including dual instances of F&H....As soon as I started messing with it I loaded it with 2 instance of FFT Prime95 and it froze within 5 minutes though heat only reached like 58-60c. Temp shutoff is 70c and we verified that to work later.....

The board after reboot clears out all settings so I set it to same vcore to lower speed of 245fsb for 3430mhz....It subsequently ran 2 instances of prime for 37minutes so power definitely seemed like an issue at 3.5ghz. Especially since I saw a max temp of 64c doing it and it did not freeze....

Now for the interesting part....

I decided to run S&M 1.73...It is the program that spanks prime95's arse when it comes to stress of heat and power....especially when you focus on the cpu (FPU) portion of the test. Now S&M is dual core setup so 1 instances is all it takes to load both cores. It also has tabs for each core....It was during running this that I noticed that...


THE P-D was THROTTLING!!!!!

At 55c the cpu being reported by the S&M for core1 starting dropping from 3430mhz.....By the time we hit 60c the speed was listed about 100mhz lower then core 0 and set speed.....When we hit 64c the speed was over 400 mhz and actually dipped below 3ghz. To verify this was temp related we removed the side of the case and subsequently the temps fell to around 56c and the cpu mhzz of core 1 climbed back to within 20mhz of core 0....Place the side back on and the temps rise and cpu1 drops again back to 400+ mhz lower.....

We at the same time ran this on the X2 3800+ oc'd to 2.55ghz at 1.475v and noiticed max temp of 56c and no differnece greater then 1-3mhz of reported speeds between core 0 and core 1....It is clearly just a P-D thing here...

Ultimately we decided to see how high it would go. Using the long test, 100% load and using the cpu (fpu) test about 29 minutes into the test we hit 69c and the chip had throttled a whopping 800mhz....that is right!!! we then watched slowly as it creep to 69.2, 69.3, 69.4, etc....Finally the system rebooted but this time did not clear the settings...it was strictly the thermal safety setting in the bios that shut it down....

I then remember the stuff Stevty was talking about, the higher the vcore the lower the throttle point. I decided to prove that theory....and I did....Setting the cpu volts to 1.4v in the bios and dropping the cpu speed now to 240 for 3360mhz I then restarted S&M and the same cpu (fpu) test....Once it rose to 55c (rather quickly too) I checked the cores and no throttling...checked at 57c....checked at 59c and was starting to get iffy, and then finally at 60c the cpu speed of core 1 dropped more then 10mhz below the other....The temps slowly rised over time until it reach 64c and it was about 70mhz lower clock speed then core 0. Bottom line was the throttle temp point defintiely had changed.


At 64-65c and removing the side of case the heatpipes we so damn hot holding your figure there too long was uncomfortable....

THIS THING IS A BEAST...


My best guesses is that 3.5ghz will not be stable. 1 cause it needs more power and 2 unless mark wants to run with case side off he will likley throttle quicker then current 55c. Even with case side off I was able to get it to throttle with 2 instance of prime95 running as it hit 58c....

3.43ghz likely can work with case side off and folding at home but it may just crosss into slight throttling evry now and then when the room gets a bit warmer....room temp was 76f but he did have a window open and it is like 48-50f at the time outside.

Most likely to insure no throttling 1.4v might be the max vcore to use and a speed of about 3360mhz....If you can keep vcore low you can OC these chips but temps will be likely still high....


The sad thing is, that HSF is the best damn air cooling you can buy and it doesn't muster it at these rather conservative levels....IMHO...




PS.

Guys run S&M and if anything just launch the app and use its monitoring of temps and core speeds...you dont have to run the app but can use it to do those 2 things and is far more effective then cpu_z was...


EDIT 1:

let me clarify 1.45v was what was set in the bios...It actually was undervolted a bit from that but the surprising thing may be more related to this mobo but the droop under load was tremendous. I have heard reportes from other boards for these chips similar stories of undervolting a drooping...In this case it didn't affect ocing and did't seem to help the temps either...

 

[mechBgon]

Yikes! :Q

Someone remind me again where Intel goes from here (roadmap-wise, I mean).

 

[n7]

I love S&M myself

Err, i mean the program :laugh:

I was waiting for this review, or should i say scolding.

Pardon me for saying the obvious again, but what a joke these P-Ds are...

The sooner they kill Netburst, the better!

 

[n7]

Originally posted by: mechBgon
Yikes! :Q

Someone remind me again where Intel goes from here (roadmap-wise, I mean).

They go towards a variant of the Pentium-M, thank god.
That's a couple yrs. away though, sadly.

 

[Duvie]

N7 what kind of temps do you see with the big typhoon on that X2???

 

[Markfw]

Duvie and I spent quite a while on this.(thanks Duvie !)

As he said, its hot, it throttles, and its not that fast. Also if you have more than a couple apps going, it "feels" slow, even with a gig of ram. The 3800+ system we were comparing to has only 512 meg, and was still faster and more responsive. I will be leaving the case cover off the 820 to not loose performance, but I won't get another one !!!!!

 

[Duvie]

The intersting future test will be to run it at stock vcore to the highest stable speed and see if we can get it to mid 60's and show no signs of throttling. A while back keys talked about his works P-D 830's and the high temp they had. If the were stock vcore perhaps even 68c would not throtlle then.

It would be interesting to map the different throttle points from default climbing upwards...

 

[DrMrLordX]

Originally posted by: mechBgon
Yikes! :Q

Someone remind me again where Intel goes from here (roadmap-wise, I mean).

This THG article lays it out pretty well:

http://www.tomshardware.com/cpu/20051010/index.html

Intel will be releasing Cedar Mill(Pentium 4 replacement) and Pressler(Pentium D replacement). In short, they're nothing more than 65 nm versions of Prescott and Smithfield, with some extra l2 cache and maybe 200 mhz higher clock speeds at the top end.

Word is that they run cooler clock per clock than Prescott/Smithfield, but performance per mhz doesn't look much better. Intels next-gen can't really beat AMD's current best chips, which is kinda sad.

As many have said before, wait for Conroe.

 

[dmens]

Since we're big on "personal experiences" and "feelings" here... The PD 840 demo machine at the employee break room has been running for 4 months. To be specific, it has rebooted once since it was set up. And that is pre-PRQ stepping silicon. I got myself a freebie 830 and it has not crashed since I set it up. But I did not bother overclocking.

When people go crazy with the frequency, all bets are off. Whining about stability and thermals when you OC is absurd, especially when you compare the stability with some other processor while overclocked.

 

[Lonyo]

Where can you get S&M from? Google (obviously) isn't hugely helpful, because it gives "alternative" results

 

[bunnyfubbles]

Originally posted by: dmens
Since we're big on "personal experiences" and "feelings" here... The PD 840 demo machine at the employee break room has been running for 4 months. To be specific, it has rebooted once since it was set up. And that is pre-PRQ stepping silicon. I got myself a freebie 830 and it has not crashed since I set it up. But I did not bother overclocking.

When people go crazy with the frequency, all bets are off. Whining about stability and thermals when you OC is absurd, especially when you compare the stability with some other processor while overclocked.

I guess you don't get it, if you want a slow dual core system, then a P-D is fine. Overclocking was an unlikely path of redemption, but when it runs so hot before doing so expecting much from an OC is unreasonable. This just confirms it as so.

We already know that the X2 is faster than the P-D at stock speeds, what this post confirms is that not even overclocking can save the P-D and the X2 becomes even more attractive.

I'll say it again, the P-D is a CPU that can only fit in a very small niche due to its low price, and this only includes the 820. If for whatever reason you need dualcore power and can settle for the well under par single core performance, and you absolutely cannot pony up for a 3800+ system, then an 820 might be for you. If you can get a P-D for free then there's no complaining there. I'd take a Celeron for free, because free is free.

 

[n7]

Originally posted by: Lonyo
Where can you get S&M from? Google (obviously) isn't hugely helpful, because it gives "alternative" results

http://www.benchmarkhq.ru/english.html?/be_cpu.html

Scroll down toward the bottom of the page.

There's a number of kewl appz on there

 

[Markfw]

One thing I didn't see mentioned ? This was by an open window, and outside it was 56-58f, so there was an abundance of cool air to feed the beast as far as abient temp goes. The actual abient temp right by the intake probably close to 65-68f as it was about 18 inches away from the window, but still very cool air. Again, no way this sucker could every hit 4 ghz on air unless it was in the middle of winter and the outside temp was 15 below with a fan sucking that cool air right to the intake !

And dmens, sure at even 3.2 you could probably use a normal HSF and it would be stable. But cool ? no way! And speed compared to a stock 3800+ with the 820 at 3.2 (or an 840) ? As the benches show, the 3800+ still wins virtually every benchmark. But at 2.8 ? its miserable. Point is you get what you pay for. At $250 its a dual core cpu, but the slowest that can be had, and on single core apps, its 3-4 years ago speed or more.

I will probably leave the side cover off, and let it be a F@H box untill I can find a sucker to buy it.

 

[DAPUNISHER]

Originally posted by: bunnyfubbles

Originally posted by: dmens
Since we're big on "personal experiences" and "feelings" here... The PD 840 demo machine at the employee break room has been running for 4 months. To be specific, it has rebooted once since it was set up. And that is pre-PRQ stepping silicon. I got myself a freebie 830 and it has not crashed since I set it up. But I did not bother overclocking.

When people go crazy with the frequency, all bets are off. Whining about stability and thermals when you OC is absurd, especially when you compare the stability with some other processor while overclocked.

I guess you don't get it, if you want a slow dual core system, then a P-D is fine. Overclocking was an unlikely path of redemption, but when it runs so hot before doing so expecting much from an OC is unreasonable. This just confirms it as so.

We already know that the X2 is faster than the P-D at stock speeds, what this post confirms is that not even overclocking can save the P-D and the X2 becomes even more attractive.

:thumbsup:

 

[carlosd]

The P-D sucks,sucks, sucks...

 

[Duvie]

Originally posted by: dmens
Since we're big on "personal experiences" and "feelings" here... The PD 840 demo machine at the employee break room has been running for 4 months. To be specific, it has rebooted once since it was set up. And that is pre-PRQ stepping silicon. I got myself a freebie 830 and it has not crashed since I set it up. But I did not bother overclocking.

When people go crazy with the frequency, all bets are off. Whining about stability and thermals when you OC is absurd, especially when you compare the stability with some other processor while overclocked.

Hey this is an overclocking forum as well so go thread crap elsewhere....


I wasn't trying to test it at stock but I can tell you a stock 4400+ on my system runs a load of 40-42c with a lesser HSF then the one Marks P-D had.....

At 3360mhz the last speed we tried the OC for the P-D was 560mhz and the X2 was 550mhz. It is clear the INtel system was not running without throttling. It likely will throttle before it reboots and had we had that feature off in the bios I gues we could have been happy with mediocrity too...It likley would have never rebooted but just kept slowing the one core down.

I have no doubt...DUH!!!! That this cpu would have not throttled even with side case on if I had it at stock or perhaps even 3.0ghz. The sad thing is this setup is so much better then what an average person may even use at stock speed. Hence why we have seen threads about 820's at stock even being very hot.

I can only cringe to think of what we wouod have seen with this in a smaller case with stock cooler.....



<<I got myself a freebie 830 and it has not crashed since I set it up.>>

It likely will just throttle before that happens anways at stock....

<<Whining about stability and thermals when you OC is absurd, especially when you compare the stability with some other processor while overclocked.>>

Who is whinnning? I am lucky I dont own the POS!!! Once again for the clinically slow...THIS IS A CPU/OVERCLOCKING FORUM where we talk about these things. In terms of ability, it is a less percenatge of an OC versus the AMD, with more expensive cooling given to it, and the temps are still 10c more.....

It is all relative...


I guess we will be fortunate not to see you comment in this thread anymore, right???

 

[Pabster]

Not to mention, Duvie, that X2 4400+ even at stock would literally wipe the floor with the D 820. Then again, you are comparing a $525 chip to a $215 one.

This forum does cover overclocking discussion and I'm sure a lot of people were interested in how well the D 820 would do, seeing as how we don't get to hear from Pentium D owners / users very much here.

 

[mechBgon]

I'm still interested to hear how the P-D and 3800+ compare in some Adobe Premiere Elements encoding. Actually, I'd like to see the benchmarks run at stock clockspeed if possible. If the results are good for the 3800+, then this would be ammo against "Them" ever trying to give me a pre-built P-D at work and expecting me to give up my SCSI-equipped A64, which is going to become a SCSI-equipped A64 X2 at approximately 2PM on Monday

 

[Markbnj]

[I guess we will be fortunate not to see you comment in this thread anymore, right???]

Hahahaha :thumbsup:

 

[coomar]

nice work duvie and mark

 

[Duvie]

Originally posted by: mechBgon
I'm still interested to hear how the P-D and 3800+ compare in some Adobe Premiere Elements encoding. Actually, I'd like to see the benchmarks run at stock clockspeed if possible. If the results are good for the 3800+, then this would be ammo against "Them" ever trying to give me a pre-built P-D at work and expecting me to give up my SCSI-equipped A64, which is going to become a SCSI-equipped A64 X2 at approximately 2PM on Monday

I even tried it and something was not right as it would never use more then like 5-6% cpu usage when I tried to render the clip to a movie.....I messed around and got it to 23% but never to where it should have been....

We decided to test it with AutoGK and Divx and it was performing where it should based on my 4400+'s testing at 2.6ghz....

 

[Dadofamunky]

Originally posted by: dmens
Since we're big on "personal experiences" and "feelings" here... The PD 840 demo machine at the employee break room has been running for 4 months. To be specific, it has rebooted once since it was set up. And that is pre-PRQ stepping silicon. I got myself a freebie 830 and it has not crashed since I set it up. But I did not bother overclocking.

When people go crazy with the frequency, all bets are off. Whining about stability and thermals when you OC is absurd, especially when you compare the stability with some other processor while overclocked.

I think he's got a point, just because most office systems with dualies won't ever be tweaked and are probably fine in that context.... but they still run hotter and eat up more electricity and are probably not suitable for the home environment where AC is not as commonplace. I sure wouldn't want one of those things in my cube, blasting hot air at my legs!

 

[Dadofamunky]

Originally posted by: Duvie

Hey this is an overclocking forum as well so go thread crap elsewhere....


Who is whinnning? I am lucky I dont own the POS!!! Once again for the clinically slow...THIS IS A CPU/OVERCLOCKING FORUM where we talk about these things. In terms of ability, it is a less percenatge of an OC versus the AMD, with more expensive cooling given to it, and the temps are still 10c more.....

It is all relative...

I guess we will be fortunate not to see you comment in this thread anymore, right???

I think I'm too nice a guy, Duvie's posts are so much more entertaining...

 

[aka1nas]

The Pentium D machines we set up for some clients weren't reliable until we replaced the stock cooling with XP90-Cs and some AS5. This chip is ridiculous to keep cool. What is going to happen to all those Dells with them in 2 to 3 years of accumulating dust? Regular office workers never clean out their omputer cases. I bet there will be a big string of heat-related failures on these boxes over the next fewyears.

 

[anandtechrocks]

I've put together a PD 820 (for someone who wanted a $600 dual core system) and I thought it was plenty fast (games, multitasking, encoding) and regardless of what others have said it did not throttle once during its 24 hour Prime 95 session.

However, the heat is the main thing that bothered me. My A64 idles 28-30C and loads 38-40C and I just can't get used to the PD that idles 49-50C and loads 59-64C.

According to Anandtech's Intel road maps hopefully we'll see Conroe 2nd half of next year! Untill then AMD is the only path I'll take!