Wait for Intel's Haswell, or get AMD's Piledriver?

luckylachance

Honorable
Sep 15, 2012
1
0
10,510
Hi, Im looking to build a fast computer that does everyday tasks and A LITTLE gaming.

The gaming part isn't important, but I want to overclock a very tiny bit.
I'm going to be getting everything at the holidays so I wouldn't have to wait too long for the Haswell,

but Intel is very expensive. My budget is $600-$1000, or 457-761 Euros.
 
There is no real point in waiting for PileDriver.

If PD can achieve a 10% performance improvement over Phenom II / FX then that will mean PD will be about as powerful as the 1st generation of Core i3/i5/i7 CPUs (Clarkdale / Nehalem).

If you build a PC around an Intel Sandy Bridge / Ivy Bridge Core i3/i5/i7 it will be, by default, faster than PD.
 
According to AMD they stated that their goal is to achieve between 5% to 15% improvement in performance with the release of every new CPU.

5% is a bit weak, it's only slightly better than a "side-grade" in which case PD will be like another FX (but without all the hype).

15% is a bit extreme; not even Intel achieved that feat; although Sandy Bridge came close with about a 12% performance over Clarkdale / Nehalem Core i3/i5/i7 CPUs. Needless to day Intel has a larger R&D budget. Can AMD achieve a 15% performance increase? Yes, it is possible (anything is possible). But how probable is it?

10% represents a reasonable performance increase expectation for AMD. Given AMD's more limited R&D budget (about 1/6 of Intel's) it should be achievable. Having said that, Ivy Bridge CPUs are about 29% more powerful than Core 2 Duo / Quad, Phenom II and FX CPUs. Therefore, in order for AMD's PileDriver to become somewhat competitive to Intel's Ivy Bridge CPUs, PileDriver would need to have at least a 22% performance increase which would place it in realm of Sandy Bridge's performance. Close enough to be considered competitive against Ivy Bridge.
 

wavetrex

Distinguished
Jul 6, 2006
254
0
18,810
There's no such thing as a "bad" processor or a "bad" videocard.
It all comes down to price.

If the AMD cpu is 20% slower than Intel's but it costs 30% less, then it's a better buy. Not everyone wants to be on the bleeding edge of performance. I for example can use the already old Athlon X2 in my laptop very well for most common tasks ( browsing, music/video, a little web programming, sorting and processing photos, others )
My desktop is a highly overclocked quad core (even if it's not latest generation) ... Q6600 running at 3.4Ghz, and to be honest while it's benchmark scores are at least 3 times higher than the laptop, I do stuff at pretty much the same speed as on the laptop.

Unless you do HD encoding, Rendering or hardcore gaming, those top-end CPU's are severely underused.
So yea, it's down to price...
 
If you don't care about gaming very much then you would have no problem piecing together a setup with a 2500k and a 6770. A Z68 or Z77 board would have all the UEFI, 6GBPS, USB3, and whatever else you might want.

That way you don't have to place bets on any untested future products.
 
The disparity of AMD and Intel CPU performance allows for a wide range of prices for current generation CPU based on overall performance. Should everyone on this planet buy an Intel CPU? No. Sure, overall their CPUs have better performance, but they also cost more. There are less expensive Intel CPUs, but lower end AMD CPUs that costs just as much or less can provide similar performance levels.
 
Like 80% of all desktop/laptop processors sold are Intel.

Sure, its not 100%, but its close.

Sure they aren't for everybody, but their value proposition is good enough to take in the lion's share of the total.

TBH, AMD kinda really just sucks in performance per watt and that makes a pretty large difference in terms of the overall ownership cost of a given CPU.

The gap is set to widen a whole lot more soon too. Rumor is that for Ivy vs Haswell, the chips that are the same power level the Haswell will use about half the wattage as the Ivy Bridge uses.

That will put serious gamer CPUs at under 50w easily, maybe even under 40w. If history is any guide Piledriver will probably be more like 150w.

That is a lot of unnecessary heat and power bills avoided.

AMD chips have a steep markdown in the retailer, but when you add back to that all the indirect costs the Intel chips start to look better.

AMD needs to get their processor wattage down by like 2/3 to even compete.
 


Piledriver (Vishera) should be out by then, it sounds like a very good fit for you. If not, get an existing Llano (K10/Stars-based) APU. You want an inexpensive CPU for pretty untaxing work, you want to be able to overclock a little, and you want decent graphics. That is exactly what AMD's APUs offer. Intel's CPUs excel at CPU-heavy tasks but their graphics are significantly behind AMD's, most can't really overclock at all, and they are pricey as you mention. Intel is a company that prides itself on high CPU performance and a highly integrated laptop platform and charges large premiums for those. The preliminary data on Haswell pretty much says that it will be very much in line with this.
 

viridiancrystal

Distinguished
Jul 27, 2011
444
0
18,790
I would wait on PD. Trinity tests showed about a 15% increase in in performance clock-over-clock. If this holds true in the full PD, then your looking at some very competitive chips if priced right. A 4-core PD could show up around $130 or less. Might not be as good as Intel's i5's, but it will be much cheaper.
 


Llano is based on the Family 16(10h)/Stars/K10 architecture, which is is the same one used by the Athlon II and Phenom II. The Athlon II is generally slower than the Phenom II because the Phenom II is an Athlon II plus 6 MB of L3 cache. Llano has no L3 cache but is a little faster than the Athlon II because the memory controller and a couple other things in Llano got tweaked relative to the Athlon II/Phenom II.
 

InvalidError

Titan
Moderator

I agree that needing a 125W AMD chip to compete with a 80W (or sub-50W with Haswell) Intel chip is a little sad. With such a large performance-per-watt gap, the Intel chips can pay for themselves via lower power bills over as little as a single year.
 
I would go with Ivy Bridges if you want to build a computer now. Haswell isn't going to be out until next year and it doesnt seem like you want to wait that long. I wouldn't bother with AMD and I wouldn't wait for Pildriver either. Even IF it does have the performance increase AMD is claiming than it will put it on the level of the first generation I core processors, still way behind Intel. Even a low end SB or IB can beat out most of the AMD Bulldozer's and it will probably be the same for Piledriver once it's released..
 

egilbe

Distinguished
Nov 17, 2011
1,417
0
19,460


45 watts difference is enough to pay for the itself? I'm not sure what you are paying for electricity, but it would take over 10 years at that rate to pay for itself.
 


If you are going to make assumptions, you should say what they are. So should the person you quoted.

10 years of how much per day usage?

1 hour? If so then it would be possible to get 10 years shrunk down into 1 year just by using the PC 10 hours a day.

Also, its not important to recover the entire cost of the processor. It is more important to recover the gap between worse processor A and better processor B. Once that gap is recovered, you are in the black purely on the numbers. Thus to "pay for itself" you really only need to cover 1/2 to 1/3 of the total cost of the processor, ie the difference between most Intel chips and their cheaper higher wattage "equivalents".

Case in point the FX-8150 (at $190 on Newegg) usually comes up compared with a 2500k (and lacking btw, at $220). The difference in pricing is $30 out of $220 or about 1/7 the total cost. It is only that $30 that needs to be recovered for there to be zero downside to choosing the Intel chip, even in terms of the up front purchase price.

Rounding some numbers off for simplicity (and ignoring that OCing is heavily done with both processors and that skews the data even more towards getting the Intel chip), the usage of the Intel chip saves you about 1/20 of a KWH and at 10 cents per KWH, every hour the computer is used saves about 1/2 of 1 cent. 2 hours makes 1 cent. 200 hours makes 1 dollar. 6000 hours makes $30.

Where I live, electricity varies from 7.2 to 13.2 cents per KWH using plans from about 100 different providers with widely varying terms of service, so I went with a value in the middle both for ease of calculation and because values lower often have service terms that make up the difference anyway. Plenty of plans at 10 cents here are 100% renewable energy sources as well, for those environmentally aware.

Anyway, at 6 hours a day that is 1000 days to recover the gap which is about 3 years. At 12 hours a day the time shrinks by half down to a year and a half. Most hardcore gamers will probably fall somewhere in the difference between those. Some may even push the break even point down pretty close to 1 year.

It may really take 10 whole years to recover the entire $220, but that is an entirely flawed way to look at cost-benefit analyses anyway, regardless what you assume "pay for itself" means.

Going back to above where I said ignoring OCing, its pretty common for gamers not to ignore this and at +2% power gains +1% performance average rate the higher wattage FX would scale up in wattage much faster than the lower wattage SB chip would.

Saying the OCer were wanting a +50% performance in both cases (taking the 2500k to 5GHZ vs 8150's 5.4 GHZ) which is pretty aggressive, that would double the gap between the high and low and half the numbers above, making it quite easily to recover the difference in one year for hardcore gamers. Even if you tamed the OCs some down to a more normal +1/3 (2500k goes to about 4.4 GHZ and 8150 goes to more like 4.8 GHZ) it would still require less than 12 hours a day to recover the whole $30 in one year.

Anyway, that is all just a serious effort to look at the numbers for direct cost. They don't even begin to factor indirect costs like the fact that the CPU fan runs at a lower RPM with the SB as compared to the FX, the cost of added strain on the PSU, and the effects of higher temperatures on other hardware in the computer. Those figures serve to slant the results even more towards Intel instead of AMD.

- edit - clarity
 

InvalidError

Titan
Moderator

As Raidinn said, I was referring to the price gap between the AMD and nearest indisputably faster Intel CPU. If going with Intel costs $60 extra for the CPU+board, $60 is the up-front cost to recover over time.

Since I leave my PC on 24/7 most of the year (except summer), my break-even point would be around two years.
 

egilbe

Distinguished
Nov 17, 2011
1,417
0
19,460


45 watts is less than $5 a year. Are you really going to keep a pc that long?
 


This $5 figure means nothing. It could be randomly picked out of the air for all we know. Indeed, without any backing it appears so.

Two people have shown their calculations backing the opposing side. Feel free to show your calculations backing your own side.
 

InvalidError

Titan
Moderator

Depends on who you ask.

My main computer probably averages 16h/24/ For people running SETI, Folding or other things like that, this is indeed 100% load 100% of the time the computer is turned on.

And Intel's chips have lower idle power as well, so there are savings to be made there as well. With Intel's Haswell having an allegedly 20X better idle power, the idle savings may become quite significant as well.