Conservative OCer: X2 3800 or Opteron 165

[Salamero]

I have decided I will purchase either the Athlon X2 +3800 or Opteron 165. I plan to OC ONLY up to 2.2Ghz because a crash in my line of work is unacceptable. I'm a Mechanical/aerospace engineer and work with several resource intensive programs and I'm a HUGE multitasker.
I'm leaning towards the 165 'cause its designed/tested to be very reliable but once I OC it I worry that it looses that. Also, It seems that there are few reliable Motherboard for the 165 (I looked at a bunch of sigs of people in this form and they all have one of two Mb's). The X2 3800 is well liked and reputable but the 165 seems to be beloved!

So, what would you get, both are available at the same price, if you where me? I plan on getting a good video card (~$150-$200) and 2x1Mb RAM of memory.

S

 

[spacelord]

I just purchased a X2 3800+ mainly because it was readily available and cheaper than any 165 I could find these days.
I too plan on doing some mild overclocking and it seems that the 3800+'s tend to overclock decently. Plus I thought the x10 multiplier might give me more options to stabilize my system with my memory when overclocked.

 

[Salamero]

Originally posted by: spacelord
....Plus I thought the x10 multiplier might give me more options to stabilize my system with my memory when overclocked.

Pardon my ignorance but, could you explain that to me?

 

[Somniferum]

No offense, but if a crash on a particular machine is "unacceptable" then you really should not be overclocking that machine AT ALL. Overclocking mission-critical hardware is never a good idea. Just because it's overclocked "conservatively" does not mean the additional risk of failure is zero. You are increasing that risk with every MHz you clock above spec. If a crash is unacceptable then increasing the risk of a crash should be unacceptable as well.

Better to run the mission-critical machine at spec, and build another "enthusiast" machine to play with.

 

[RichUK]

http://forums.anandtech.com/messageview.aspx?catid=28&threadid=1760972

 

[allies]

Originally posted by: Salamero

Originally posted by: spacelord
....Plus I thought the x10 multiplier might give me more options to stabilize my system with my memory when overclocked.

Pardon my ignorance but, could you explain that to me?

With a higher multiplier, you have a little more flexibility when increasing fsb and running dividers. If your motherboard can handle say, 300 MHz FSB, and your memory can only run at 220, and your chip could run at 2.6 GHz, you would have an easier time reaching the 2.6 Ghz with a 10x multi (10x260) versus a 9x multi (9x289) because you could be easier on your ram, depending on what dividers your mobo has.

 

[RichUK]

Originally posted by: Salamero
I have decided I will purchase either the Athlon X2 +3800 or Opteron 165. I plan to OC ONLY up to 2.2Ghz because a crash in my line of work is unacceptable. I'm a Mechanical/aerospace engineer and work with several resource intensive programs and I'm a HUGE multitasker.
I'm leaning towards the 165 'cause its designed/tested to be very reliable but once I OC it I worry that it looses that. Also, It seems that there are few reliable Motherboard for the 165 (I looked at a bunch of sigs of people in this form and they all have one of two Mb's). The X2 3800 is well liked and reputable but the 165 seems to be beloved!

So, what would you get, both are available at the same price, if you where me? I plan on getting a good video card (~$150-$200) and 2x1Mb RAM of memory.

S

In that case overclocking to save yourself x amount of dollars is not worth the result of a system crash in your profession surely, is this a PC or a workstation? Overclocking is for fun home use, most definitely not professional use, especially if downtime is a concern. I wouldn?t dream of OCing my workstation (my work computer), although it probably can not be overclocked. OCing effects could be detrimental depending on the line of work you are in.

There is a reason why opterons have such a high yield, since it is designed for the high end business solutions and is therefore more versatile. Specifically when preventing such instances of downtime/corruption and or CPU/System failure, as this is paramount.

Just get a 175 or 4400+

 

[Fallengod]

I personaly dont see the point of overclocking to only 2.2ghz. Thats not much of a difference from 1.8 or 2.0ghz. Just get the X2 3800 then and run it at stock, than bother risking a crash.

But anyways, I posted a thread in hot deals. You can get the 165 opteron oem right now for $278 shipped. So that is probably cheaper than the x2. But anyways....

 

[mrkun]

Since reliability is an issue, get a retail (3-year warranty) Opty 165 and don't OC it.

 

[Salamero]

So, if I don't OC either which would be the best choice.
The 165 SHOULD be less prone to crashes than the 3800. Does the 165's extra 512 of L2 Cache cancel out the 3800's extra 200mhz?
Which is better for multitasking, more L2 or more clock speed?

 

[Salamero]

It'll be a workstation:
SOLIDWORKS
PRO/E
MATLAB
MICROSOF VISUAL STUDIO- VISUAL BASIC/C++
MATHCAD

That's for work.
The rest is video editing, music, web browsing,ect. Not a big gammer I'm afraid.

 

[RichUK]

Originally posted by: Salamero
So, if I don't OC either which would be the best choice.
The 165 SHOULD be less prone to crashes than the 3800. Does the 165's extra 512 of L2 Cache cancel out the 3800's extra 200mhz?
Which is better for multitasking, more L2 or more clock speed?

Neither will crash at stock as the manufacturer has already verified and binned that speed grade. When you run it out of spec (aka OCing) thats when you run into possible trouble, and then perceived stability is then in the hands of the user, not the manufacturer.

The moral of my statements, is it is not worth OCing on a platform that is relied on, especially not for a mere 200Mhz (X2 3800+ example). When OCing on a personal use computer, if something goes wrong what are the consequences, you cant play games you cant browse the net, you cant read emails etc (pretty insignificant would you not agree), and who losses out, the answer is you. If you do that on a work platform and depending on how significant that particular workstation or PC is, then the effects could be a lot more severe (obviously depending on the situation).

The point I was trying to get across with the Opteron processor, is that it is of a higher grade silicon, hence the higher average OC, and perhaps this is how you stumbled onto the subject of the capabilities of the opteron. These processors are engineered for upto ?mission critical? servers, and that?s the purpose for the better quality processor, even though it has the same architecture as the Athlon 64 equivalents.

But hey that is just my perception, if you are experienced in OCing and want to give it a try then fair play. I personally wouldn?t want to explain to my boss the cause of a problem if it was due to me OCing my workstation or work computer however. Then again I don?t understand your situation if this is a standalone computer then it could be ok as the consequences (if any) would only effect you right? You would have to make sure that it is quite stable and you do regular backups if choosing to OC.

I would personally (if budget conscious) go with the X2 3800+, as 200Mhz extra per core is much more beneficial than the added L2 cache with the 165. However if OCing then the choice of processor is a more debateable, and I would personally recommend a 165. This is if you have a decent motherboard and cooling solution in place.

If you end up going with a stock processor and decided you require a fair amount of processing power I would recommend going with a 4800+ if you can afford it, or possibly the 4400+ (both have 2Mb L2 cache) and run respectively at 2.2Ghz and 2.4Ghz stock, the 4800+ is the latter operating frequency.

Hope this helps.

 

[SickBeast]

Originally posted by: Somniferum
No offense, but if a crash on a particular machine is "unacceptable" then you really should not be overclocking that machine AT ALL. Overclocking mission-critical hardware is never a good idea. Just because it's overclocked "conservatively" does not mean the additional risk of failure is zero. You are increasing that risk with every MHz you clock above spec. If a crash is unacceptable then increasing the risk of a crash should be unacceptable as well.

Better to run the mission-critical machine at spec, and build another "enthusiast" machine to play with.

I disagree. If a machine is properly tested for stability it should be 100% stable when overclocked. I have had my 1833mhz chip overclocked to 2500mhz for 2 years now without a single crash due to overclocking. I've done my share of mission-critical CAD work on it, too.

 

[Somniferum]

Originally posted by: SickBeast
I disagree. If a machine is properly tested for stability it should be 100% stable when overclocked. I have had my 1833mhz chip overclocked to 2500mhz for 2 years now without a single crash due to overclocking. I've done my share of mission-critical CAD work on it, too.

It should be 100% stable, and in many or most cases it may well be, but I don't see how you can deny that statistically, the probability of the part failing early increases when you run it out of spec. Moreover, if it should fail, by OC'ing you have shifted the responsibility for that failure away from the manufacturer and onto yourself (as RichUK points out above).

Obviously the moral weight of that responsibility depends on the relative importance of the machine and the tasks it performs, but as long as there are people other than yourself dependent on the machine's stability, then at the very least I would think they deserve to be informed of the increased risks they are being exposed to. In other words, if you can tell your boss/coworkers/shareholders that you are OC'ing the mission-critical hardware, and they are fine with it, then hey, no harm no foul. But otherwise, you are putting other people at risk without their knowledge or consent -- which is morally problematic to say the least.

In 99.99% of cases it may never become an issue, but a moral principle must be true in all cases -- even that .01% where the OC'd mission-critical hardware fails due to electromigration at normal temperatures, causing disastrous consequences for people other than yourself who had no choice in the matter.

Just something to think about.

 

[Griswold]

What RichUK said is pretty much the bottom line. If you absolutely cant afford crashes, or worse, calculation errors due to overclocking, dont overclock. Choose wisely the parts you're going to use, but the CPU should be the last thing you need to worry about.

I also suggest a 4400+ or even 4800+ if your wallet permits that.

 

[themusgrat]

Look at pricewatch. Monarch has the 165 in stock now for 271, and I just bought one shipped, with the additional 3-yr. warranty, for 293. Plus, it has double the L2 cache, and is more stable at higher speeds due to its inherently better silicon. The lowest priced 3800 is 315, so which is the better value now? The 165 costs less and has more performance, since a stable OC at 2.0 is possible on any chip. Look at Shimmishim's thread of unofficial Opteron OCs. Pretty much no matter what, the 165 beats the 3800, as long as one's mobo can handle it. Get the 165, and you'll be happy.

 

[Salamero]

Originally posted by: Somniferum

Originally posted by: SickBeast
I disagree. If a machine is properly tested for stability it should be 100% stable when overclocked. I have had my 1833mhz chip overclocked to 2500mhz for 2 years now without a single crash due to overclocking. I've done my share of mission-critical CAD work on it, too.

It should be 100% stable, and in many or most cases it may well be, but I don't see how you can deny that statistically, the probability of the part failing early increases when you run it out of spec. Moreover, if it should fail, by OC'ing you have shifted the responsibility for that failure away from the manufacturer and onto yourself (as RichUK points out above).

Obviously the moral weight of that responsibility depends on the relative importance of the machine and the tasks it performs, but as long as there are people other than yourself dependent on the machine's stability, then at the very least I would think they deserve to be informed of the increased risks they are being exposed to. In other words, if you can tell your boss/coworkers/shareholders that you are OC'ing the mission-critical hardware, and they are fine with it, then hey, no harm no foul. But otherwise, you are putting other people at risk without their knowledge or consent -- which is morally problematic to say the least.

In 99.99% of cases it may never become an issue, but a moral principle must be true in all cases -- even that .01% where the OC'd mission-critical hardware fails due to electromigration at normal temperatures, causing disastrous consequences for people other than yourself who had no choice in the matter.

Just something to think about.

You're absolutley correct. The PC will be used to work at home but, ultimatly, work is work no matter where and on whos PC the task is performed. I will make my purchase based on Stock specs. I will however, do more homework into OC & testing it till its %99.9 stable!

 

[nealh]

I agree regarding need for 100% stability but unless you have a PC builder who will guarantee this...no computer is perfect even at stock speeds...we see all the time people posting with Dell's, HP, other vendors machine that are crashing at stock setup

I have an office PC from and a company that configures the PC to run there Digital Photo software for taking Ophthalmic angiograms..I have 2 offices..2PCs from Dell..6mo difference in age..no OCing and these things crash all the time..right now there seems to be an issue with the cd writer in one and dvd writer in another causing repeated crashes if a disc is in the writer(very strange issue they have replaced both writers)...go figure

So if this is a work machine avoid headaches of not knowing whether the OCing is an issue..get the fastest pc you can afford...Look into the Opteron 175 for about $500...also

 

[themusgrat]

Originally posted by: nealh
I agree regarding need for 100% stability but unless you have a PC builder who will guarantee this...no computer is perfect even at stock speeds...we see all the time people posting with Dell's, HP, other vendors machine that are crashing at stock setup

I have an office PC from and a company that configures the PC to run there Digital Photo software for taking Ophthalmic angiograms..I have 2 offices..2PCs from Dell..6mo difference in age..no OCing and these things crash all the time..right now there seems to be an issue with the cd writer in one and dvd writer in another causing repeated crashes if a disc is in the writer(very strange issue they have replaced both writers)...go figure

So if this is a work machine avoid headaches of not knowing whether the OCing is an issue..get the fastest pc you can afford...Look into the Opteron 175 for about $500...also

He's right. I'd like to see a sudy of what causes the most hardware crashes. I don't think that it would be the CPU, rather the motherboard or RAM or something else. If you can be prime95 stable for at least 24 hrs. and not go much past 20% of stock speeds, you'll be just fine. Anything that can be stressed for 24 hrs., unless it is a server, can be relied upon to do home or office work. And for those who say that by OCing, you are raising the chance for instability, when is the last time you saw an AMD be unstable with a .2 bump? Never, because of the way procs are binned, there is headroom in every AMD proc.

 

[Somniferum]

Originally posted by: themusgrat
And for those who say that by OCing, you are raising the chance for instability, when is the last time you saw an AMD be unstable with a .2 bump? Never, because of the way procs are binned, there is headroom in every AMD proc.

I can't answer your rhetorical question directly, because I don't have that much experience with AMD processors (just switched from Intel myself). But I don't think the general principle is dependant on experience one way or the other.

Look at it this way. If you buy a 2GHz CPU and run it at stock, then if it crashes it's clearly the fault of the manufacturer. The manufacturer knows this, so they implement burn-in procedures to eliminate those CPUs that are likely to fail.

Now take it to the other extreme. You overclock that 2GHz CPU to 5GHz out of the box, raising the VCORE to 3 volts (by modding the motherboard). More than likely it will fail instantly.

Every possible OC'ing scenario lies along an axis between those two points -- running at stock, vs. running at ridiculously high speeds that cause instant hardware failure. At stock settings, yes, there is still a chance of failure. But there are two important points to be noted here: (1) hardware failure at stock speeds is HIGHLY unlikely, due to manufacturing controls; and (2) if by some terribly bad luck your CPU does fail at stock speeds, it's clearly the fault of the manufacturer (and most likely covered by warranty).

Along that continuum there are several points. The farther away from stock speeds you go, the closer you get to the insanely high speed that fries the CPU. You can run all the stability tests you want, but the fact remains that statistically, at ANY speed above stock, you have increased the probability of hardware failure. You are moving in that direction, away from known stability and toward total system failure.

And if that machine is your enthusiast gaming rig, then more power to you. But the other part of my argument is based on the idea that people have a right to know the risks they are assuming. If other people are dependent on your machine's stability, and you are statistically increasing the likelihood of hardware failure, then you damn well better clue them in to that fact. It's a moral obilgation.

And if you feel you can't tell the people who might be aversely affected by a system crash about your overclocking adventures, then you have no business overclocking that machine. That's all I'm saying.

 

[Blain]

Originally posted by: SickBeast
I disagree. If a machine is properly tested for stability it should be 100% stable when overclocked. I have had my 1833mhz chip overclocked to 2500mhz for 2 years now without a single crash due to overclocking. I've done my share of mission-critical CAD work on it, too.

Exactly what do you mean by "properly tested for stability"?
Would your 1833mhz chip perform without crashing if it's OCd to say 2900mhz or maybe 3200mhz?

You make quite a bold statement...
"If a machine is properly tested for stability it should be 100% stable when overclocked"
Is there no point at which OC instability would occur on a machine that has been "properly tested"? :roll: