should I abandon the AMD ship? (Updated)

[crashtech]

The purpose of a benchmark is to measure the same task on competing hardware. If the specs of the other components in these system builds are all wildly dissimilar why even bother running a benchmark? You could simply fall back to the old adage that "You Get What You Pay For" and not waste the time on benchmarking.

Accurate benchmarks rely on a good baseline for comparison purposes. These system builds fail immediately because a consistent baseline was never established. Apples to Oranges -- which is better? Exactly, totally murky.

Yeah, but if the calculation gives a slight advantage to the underdog, it is difficult to discern the basis for your earlier objection. Determining the best setup for a given workload is a murky decision making process. Benchmarks can't accurately simulate real-world usage patterns. So it's all basically a guesstimate, but total system cost at least takes a stab at coming up with a more realistic figure. Apples to Oranges? The bare CPUs are more so than the complete systems, which in the referenced example share many identical parts.

 

[MiddleOfTheRoad]

Well if you asked me i wouldnt use $70-80 boards for budget build systems with AMD A6-6400K and Intel Celeron G1850.

There are excellent FM2+ boards starting at $50 with or without AR and Intel 1150 B85 chipset boards starting at $45,99 having almost the same features.
You can even go lower to A58 and H81 chipsets.

But you get more features with $229 Socket 2011-3 boards and that is something that cannot been seen in the performance per price graphs.

Agreed -- the $45 FM2+ MSI motherboards like the A55M-E33 have a remarkable amount of overclocking settings for a budget board. For a cheap gaming build -- it's pretty hard to beat it.

 

[MiddleOfTheRoad]

The bare CPUs are more so than the complete systems, which in the referenced example share many identical parts.

Regardless of who gets the edge -- Underdog or the Market Leader.... That current benchmark is totally flawed.

If they wanted a meaningful comparison for value on their system builds -- they should have opted for identically priced CPU's, motherboards and other identical components -- then run the performance benchmarks at each respective price point. But their testing methodology was so erratic -- obtaining meaningful results is impossible. The system builds themselves were so random -- it really isn't clear what the results mean. I can appreciate what they intended to do with their System Build benchmark, they just didn't execute it properly.

Focusing on the CPU alone at stock clock under an easily repeatable synthetic is a hell of a lot more clear on its performance attributes and value at least so far. If someone actually does a system build comparison for value with a disciplined and logical testing methodology -- I'd love to see the results.

System builds do introduce a lot of variables based on the buyer's preferences -- focusing on an individual component's performance/value does paint a far more clear picture for potential buyers.

 

[Ramses]

I never pay any attention to that system build crap, they always use parts that are just not to my taste and that nobody actually buys. I have $2500 in this FX box easy and it's exactly how I want it, if i was going to do budget it'd be a laptop or a chromebox these days.

 

[crashtech]

Regardless of who gets the edge -- Underdog or the Market Leader.... That current benchmark is totally flawed.

If they wanted a meaningful comparison for value on their system builds -- they should have opted for identically priced CPU's, motherboards and other identical components -- then run the performance benchmarks at each respective price point. But their testing methodology was so erratic -- obtaining meaningful results is impossible. The system builds themselves were so random -- it really isn't clear what the results mean. I can appreciate what they intended to do with their System Build benchmark, they just didn't execute it properly.

Focusing on the CPU alone at stock clock under an easily repeatable synthetic is a hell of a lot more clear on its performance attributes and value at least so far. If someone actually does a system build comparison for value with a disciplined and logical testing methodology -- I'd love to see the results.

Your critique is illogical, for instance there do not exist identically priced motherboards for FM2 and X99, and if there did they would be parts that nobody buys, but that does not mean the two configurations shouldn't be compared to one another. Isolating one component cost and using that as a basis for purchasing decisions would be rightfully scoffed at in the business world. If a company asked for cost estimates for the replacement of 1000 PCs, and for cost/benefit analyses of various different models, nobody in their right mind would suggest that this be determined by one isolated component of the system. That would be idiotic.

 

[witeken]

Well, I call complete B.S. -- you are such a fanboy........

Just linked the chart you clearly loved ignoring (which was directly above the one you just cherrypicked from Anand's web site) -- This one focusing on the CPU itself...... which BTW is very similar to Passmark's ratings.

Backing up exactly what I just stated. Unless it is a Pentium or Celeron, Intel CPU's generally have pretty lame bang for the buck. If bang for the buck is the ultimate goal for the build -- AMD still remains the better option as soon as the CPU price exceeds a Pentium/Celeron.

BTW, we all know you have a Intel shrine in your bedroom, Witeken.... Your blatant bias is all too obvious in every post.

AMD's CPU's do have several shortcomings -- but value is never usually one of them.

Not so fast with your assumptions...

I'm not ignoring or cherry picking anything. The part I quoted from you stated that people who buy Intel have systems with poor value, while in fact the opposite is true, as shown in my reply.

The truth is: hardly any consumer cares about single parts in their desktop. What really matters is the total (price of the) system, which doesn't consist of only the CPU. So if you dedicate a larger portion of your system to the processor, the performance per price in fact becomes higher.

Sure, the system will be pricier, but it will be faster as well, so it all cancels out nicely (and if you don't need the fastest CPU, you can buy a cheaper one from Intel and you'll see about the same).

If you just want to compare the processor in a vacuum (in which it doesn't live), then sure, AMD has an advantage as shown in your graph.

 

[crashtech]

What I would like to see is this chart, arranged by system price from low to high. That would, I think, better illustrate what is being obtained for the dollar. The way the graph is arranged doesn't really illustrate where the real values for the dollar lie, at least not to those who have a particular budget range in mind. I believe some AMD offerings would show better in such a light.

 

[frozentundra123456]

If one is looking at the CPU only cost, it must also be considered that you will need to buy a discrete gpu or live with chipset graphics for the FX, increasing both the initial cost and power usage. I am not that farmiliar with motherboard graphics, but I would think pretty much any apu or has well igp would be far better.

 

[Ramses]

If you guys are really interested, go find manufacturers like Lenovo that offer both and see what's what.

 

[crashtech]

If you guys are really interested, go find manufacturers like Lenovo that offer both and see what's what.

Interesting concept. The i3-4130 system above receives 8.75 Passmark points per dollar, and the A4-5300 receives 4.77. But I would guess the AMD to do well in OpenCL and game benchmarks, although the A4 is not represented in the Anandtech Bench so there isn't a handy comparison available.

 

[Ramses]

I would guess the people buying that computer could care less as long as it plays netflix and opens email and streams music. They will care about a hundred bucks though.
Gotta be murder selling those things for a living, least with a hard drive size or such you can explain in simple terms what it'll do for them. 4gig of ram and a 500gig drive in a box that size should be a crime these days, even using the inflated parts cost on the customize screen, you could take the $130 and double the ram or hard drive.
What I would be slightly interested to know is which one Lenovo makes more money on but that isn't likely to be common knowledge. I'm sure the guys configuring these things are well aware of teh relative performance/cost and/or lack of in both cases, there has to be a reason and they have to be selling or they wouldn't bother. I notice dell has no cheap AMD desktops that I could find in a quick look. I think I read Lenovo outsells them now awhile ago..

 

[crashtech]

I'd bet that the Intel is more profitable in absolute dollars but less profitable on a percentage basis. Those systems seem about $80 or so more than they should be, which as a system builder I am frankly glad to see. Getting pretty far OT, though, sorry.

 

[Ramses]

Being non-gaming I wouldn't pay $300 for either, and not for lack of CPU. Which is probly part of the reason traditional PC sales are in the tank I keep reading.

 

[frozentundra123456]

If you guys are really interested, go find manufacturers like Lenovo that offer both and see what's what.

Both those are terrible deals. You could go to NewEgg and on the first page of the computer section is a Haswell pentium with Win7 pro and a 3 year warranty for 449.00 and an A10 AMD system for 519.00. Personally, I would not even consider a single module APU, but even if you do, either of these is a much better deal than the two you linked.

 

[dave_the_nerd]

If you guys are really interested, go find manufacturers like Lenovo that offer both and see what's what.

I would glady pay $130 to not have to use an A4.

 

[MiddleOfTheRoad]

Your critique is illogical, for instance there do not exist identically priced motherboards for FM2 and X99, and if there did they would be parts that nobody buys, but that does not mean the two configurations shouldn't be compared to one another. Isolating one component cost and using that as a basis for purchasing decisions would be rightfully scoffed at in the business world. If a company asked for cost estimates for the replacement of 1000 PCs, and for cost/benefit analyses of various different models, nobody in their right mind would suggest that this be determined by one isolated component of the system. That would be idiotic.

Seriously, you have no common sense.

It would be idiotic to compare an X99 to anything AMD currently builds -- AMD doesn't compete in that segment currently.

It's not rocket science. Build an AMD FX-6300 to an identically priced i3. A G3258 to an A6 or Athlon.... This is not a difficult concept to grasp. Comparing equivalent priced components at several price points, benchmark their respective performance and baseline the value at each level. Seriously, you've got to be pretty dense not to be able to comprehend the purpose of a proper baseline for each segment when the benchmark is quantifying value. Otherwise, there is no consistency in the hardware sourced.

The lack of a proper baseline makes the current System Build chart pointless.

 

[MiddleOfTheRoad]

If one is looking at the CPU only cost, it must also be considered that you will need to buy a discrete gpu or live with chipset graphics for the FX, increasing both the initial cost and power usage. I am not that farmiliar with motherboard graphics, but I would think pretty much any apu or has well igp would be far better.

I've never met anyone who built an Intel setup that didn't also buy a discrete video card -- so it's probably a wash in cost savings.

The only people that I know that ever ran IGP were laptop owners or APU buyers.

 

[mrmt]

If you guys are really interested, go find manufacturers like Lenovo that offer both and see what's what.

AMD a4? Single module SKUs? Those chips are good for nothing, you will be crying for Silvermont or Jaguar after 10 minutes with them. Those are bottom of the barrel, die-salvaged parts that wouldn't see the light of the day if AMD were in a better situation.

 

[Ramses]

Like I said, will it check gmail? Stream Netflix? Download some BS?
That's about all the person buying them will do, and they could do that with a Chromebox for a quarter of the money or thereabouts. Unless it's big(or tiny) and ballsy I have zero use for a standard PC anymore short of nostalgia, tablets, laptops, chromebooks, chromebox, smartphones, lions, tigers, bears oh my. I don't know the first thing about any of the APU's or integrated stuff from either camp, booooring. When they break 25% performance over my FX junk, across the board, I'll pay attention. Till then, yawn.

Those examples, trying to help you guys get out of your go-nowhere rut of a discussion of relative system value/performance with some real life examples, were the first thing I found via google. You gotta get past the charts and graphs and put hands to dirt eventually.

 

[crashtech]

Seriously, you have no common sense.

I'm going to ask you respectfully not to make this discussion personal. I do accept PMs for that purpose.

It would be idiotic to compare an X99 to anything AMD currently builds -- AMD doesn't compete in that segment currently.

I ought to have said X79, sorry for the mistake. The graph includes Ivy-E as well as many other models to show their value relative to other systems. I don't think that is necessarily idiotic, the graph is pretty inclusive and gives an interesting snapshot of the value proposition.

It's not rocket science. Build an AMD FX-6300 to an identically priced i3. A G3258 to an A6 or Athlon.... This is not a difficult concept to grasp. Comparing equivalent priced components at several price points, benchmark their respective performance and baseline the value at each level. Seriously, you've got to be pretty dense not to be able to comprehend the purpose of a proper baseline for each segment when the benchmark is quantifying value. Otherwise, there is no consistency in the hardware sourced.

The lack of a proper baseline makes the current System Build chart pointless.

No, it's not pointless except that it probably contradicts preconceived notions and therefore must be attacked. But I don't think most readers have any trouble with the concept of using a more complete system cost to evaluate price/performance. "It's not rocket science," indeed. The system components tallied were for the most part identical save the motherboards (by necessity), you can't get any more baseline than that. I want to re-quote the relevant portion of the article because it makes it very clear how the calculation was done:

Naturally, there's one other element that needs to be considered: the price of the parts. Intel holds the crown for the fastest CPU performance, but if we just look at the cost of the CPUs/APUs and how much you pay for a certain level of performance the story is quite different. Not surprisingly, the least expensive processors tend to look the best here. However, processors don't exist in a vacuum – they're only a portion of the total system cost – so we really need to add in a baseline price for the rest of a system.
If you purchase a motherboard, RAM, SSD, case, and PSU it will tack on $350-$510. I used the base set of components from the recent Budget PC Guide (minus the HDD), and then chose reasonable motherboards for the various platforms ($75 for FM2/FM2+, $80 for AM3+, $100 for LGA1155, $90 for LGA1150, and $225 for LGA2011 with $10 extra on RAM to get 4x2GB instead of 2x4GB). You could certainly spend a lot more on system components, but while less expensive motherboards exist for some of the platforms, I'm a bit wary of any motherboard priced under $70. The result is that the minimum cost for an entire system ends up being around $425-$500 for the less expensive CPUs/APUs, and $600 for something like the i5-4690K, $670-$705 for Core i7 LGA1150 systems, and as much as $1559 for the i7-4960X. Using the complete system prices (including a $50 discrete GPU on the AM3+ and LGA2011 systems), the performance to price charts look as follows:

http://www.anandtech.com/show/8312/state-of-the-part-cpus

 

[crashtech]

I still don't understand the system charts. I mean, assume you pick two sets of identically priced components, the ram, mobo, psu, etc. etc., and stick AMD or Intel chip in it, wouldn't the total system performance just be a reflection/factor of of the processor performance chart?

And then considering price, wouldn't the total system performance per price again be a reflection? I don't see how the different system values can fluctuate so much, assuming the non-processor components are similar (ram etc)? It never made sense to me???

A hypothetical:

We have CPU A that costs $50 and scores 1000 Bogomarks, and CPU B that costs $100 and scores 1500 Bogomarks:

CPU A scores 20 Bogomarks per dollar, and CPU B scores only 15 Bogomarks per dollar. Looking at this alone, CPU A gives 33% more performance per dollar.

But then we put them into a complete system that costs $500, aside from the CPUs. System A costs $550, and System B costs $600:

System A scores 1.82 Bogomarks per dollar, however System B scores 2.5 Bogomarks per dollar.

Even though CPU A was half the price of CPU B, it only made a 10% difference in the overall price of the system, which caused the 50% performance advantage of System B to completely overwhelm the 10% cost advantage of System A.

But suppose the associated parts for CPU A are sourced cheaper, and system A only costs $450. Then System A scores 2.38 Bogomarks per dollar, and comes close (but not quite) to achieving parity, but at the price of sacrificing quality.

 

[Cheesesticksandbeer]

Update: I am the proud owner of a i5 4690K and ASRock Z97 Extreme 4. Thanks for all the advice guys.

 

[frozentundra123456]

Congrats.

 

[Ramses]

Very cool, enjoy it!

 

[Ramses]

A hypothetical:

We have CPU A that costs $50 and scores 1000 Bogomarks, and CPU B that costs $100 and scores 1500 Bogomarks:

CPU A scores 20 Bogomarks per dollar, and CPU B scores only 15 Bogomarks per dollar. Looking at this alone, CPU A gives 33% more performance per dollar.

But then we put them into a complete system that costs $500, aside from the CPUs. System A costs $550, and System B costs $600:

System A scores 1.82 Bogomarks per dollar, however System B scores 2.5 Bogomarks per dollar.

Even though CPU A was half the price of CPU B, it only made a 10% difference in the overall price of the system, which caused the 50% performance advantage of System B to completely overwhelm the 10% cost advantage of System A.

But suppose the associated parts for CPU A are sourced cheaper, and system A only costs $450. Then System A scores 2.38 Bogomarks per dollar, and comes close (but not quite) to achieving parity, but at the price of sacrificing quality.

Great hypothetical. The next logical question from a smart but not overly tech savvy computer shopper (assuming we're working that angle) is...

How many Bogomarks does it take to have a pleasing computing experience for the next X years? I assume someone looking at $500 pre-built PC's is trying to spend as little as possible (and Google hasn't tried to sell them a Chromebox yet lol).