Haswell i3-4150 vs FX 8320/e for budget gaming rig?

[BSim500]

My source is here :

Right, so no "63w / 65w" but measuring power usage in a different way which results in different measurements to all other sites, and when you do measure it in the same "apples for apples" way you get similar numbers:-

So by your own site - it uses 37w more than an i5-4670K yet simultaneously is supposedly only using 10w more than an i3 which usually uses 25-30w less than a FX-4670k. And you don't see a problem with that?... Most sites measure 90w delta with some outliers up to 120w and down to 75w (hardware.fr) and you hold up the hardware.fr as the "one true figure" = no real concept of what an "average" is.

It s much more professional than Anand s phony measurements, because what they do not mention is that Intel CPUs are rapidly throttling under IBT while AMDs do not.

Nothing "phony" about them and there's nothing in the article said it throttled. That's yet another thing you've projected and falsely attributed to a "source" based on yet another over-extrapolated of a few people having issues with stock coolers to "all Haswell's throttle all the time" in general. And 240v PSU losses / efficiency "at the wall" also has the same effect on Intel CPU's as it does AMD's. It more accurately reflects total energy consumed / electricity costs, and its "problem" of amplifying losses at bigger wattage chips, is presumably why some people do everything they can to ignore them for +100w AMD rigs (but were quite happy to use them when the A64 was beating the P4...)

how did Anandtech get thoses figures while HFR got about the same result as Guru3D with P95.??.

Because as explained to you before, CPU's are binned differently due to the silicone lottery. One chip have a stock voltage of 1.1v to reach x frequency whilst another may need 1.2v, and another 1.05v or 1.16v, etc. The hardware.fr may well be a particularly well binned chip with a low VID, whilst Anand may have been given a poor chip. The average is probably somewhere in between given average 90w figures across a dozen review sites. It's also why it's pointless to "dictate" all x chips use exactly 63w. They don't - even when compared to similar models ignoring all competition.

 

[Abwx]

Fritz is a complete BS benchmark and you know it. Comparing Conroe and HW there is virtually no gain in performance.

One, this is a power bench, two , haswell i7/i5/i3 got 4, 8 and 8% better perf in Fritz than IB according to HFR :

http://www.hardware.fr/articles/897-25/gains-moyennes-cpu.html

Find a revew that did such comparison...

Cherry picking power numbers (No one cares about the 12V rail, they care about the power the CPU is pulling from the PSU). A crappy platform does not make a processor power efficient (original atom), nor can it be ignored.

This is the 12V rail that supply the PSU wich is measured, the number at the main correlate the number at the CPU 12 V rail.

Their results make absolute no sense. All FX cpus idle at the same power (no they do not). Haswell idles >2x higher than FX? No it does not.

This is at the CPU 12V rail level, all FXs idle at the same power, the i series idle at lower power actualy, it s available in other reviews, what matter is the max comsumption at this level, that said if the results dont suit you...

Logic is not: Hardware.fr gets X while sites A-M get Y. All other sites must be wrong, because HWFR must be right. But rather the opposite.

One has to look at the methodology and so far they were the most professional for power measurements before sites like THG did update their own protocols to stick with this one.

 

[Abwx]

Right, so no "63w / 65w" but measuring power usage in a different way which results in different measurements to all other sites, and when you do measure it in the same "apples for apples" way you get similar numbers:-

When you see 72W before the VRMs it s simple to deduct that it s 0.9 x 72 = 64.8 W CPU TDP.

So by your own site - it uses 37w more than an i5-4670K yet simultaneously is supposedly only using 10w more than an i3 which usually uses 30w less than a FX-4670k.

Where did i write that it use only 10W more than a i3 since i specified that the i3 use about 32-33W..?..

And you don't see a problem with that?... Most sites measure 90w delta with some outliers up to 120w and down to 75w (hardware.fr) and you hold up the hardware.fr as the "one true figure" = no real concept of what an "average" is.

A 90W delta at the main is about 0.8 x 90 = 72W at the CPU level assuming the CPU consume all the delta, the only problem i see is that most people dont know how to measure a power comsumption with a dynamic system like a PC wich doesnt consume a constant power, people who use killawatts will often look at the max peak they can observe while a real measurement mandate to integrate, that is to average, the curve over time to measure the RMS power accurately, THG got recently some nice tools to do thoses measurements.

And 240v PSU losses / efficiency "at the wall" also has the same effect on Intel CPU's as it does AMD's.

Of course, did i say otherwise....?

Because as explained to you before, CPU's are binned differently due to the silicone lottery. One chip have a stock voltage of 1.1v to reach x frequency whilst another may need 1.2v, and another 1.05v or 1.16v, etc. The hardware.fr may well be a particularly well binned chip with a low VID, whilst Anand may have been given a poor chip. The average is probably somewhere in between given average 90w figures across a dozen review sites. It's also why it's pointless to "dictate" all x chips use exactly 63w. They don't - even when compared to similar models ignoring all competition.

I can tell you that Hardware.fr rarely if ever got better chips than other sites, they are sent quite average chips by both Intel and AMD...

About binning it s simple, the chips that do not pass the E requirements are rejected as regular parts, should you understand more the issue of voltage that you wouldnt had pointed that a chip can use 1.1V and another 1.2V, if they are to be E series then they must be good at 1.1V, at 1.2V they would consume (1.2/1.1)^2 = 1.19 , that is 19% more for thoses 0.1V more, it s just not acceptable, you would had a 65W chip and another that is at 77W, it would be good on regular apps but it wouldnt pass the Prime 95 test within reasonable distance of the 95W official TDP, AMD did use a 20% margin for thoses CPUs because they know that users with 95W MBs will get them, so they stayed at respectable distance of said 95W, to give a clue a 4C/8T chip that do not pass the 1.0V test at Intel is rejected as an i5 and its voltage upped to 1.1V, i would be surprised if there s 1.1V 4770Ks..

 

[BSim500]

people who use killawatts will often look at the max peak they can observe

"Energy per task used" metrics have been around for ages. As Enigmoid accurately said - on top of the higher CPU draw is higher platform draw too, so it's not really surprising some will try and drift away from "total consumption" on the rigs that consume the most... And +12v only consumption reviews are highly variable outside of the Mini-ITX "power brick" builds (where +12v is the only input voltage and +5v often pulls "through" the 12v via DC-DC). Why? Because +12v only on multi-rail ATX PSU's excludes +5v/+3.3v rails which can vary hugely across motherboards of even the same socket. AC accounts for all power draw regardless of 12v vs 5v rail. How is this "professional" to measure only 1 out of 2-3 power inputs in a power consumption test leaving you with highly variable power draw results on the motherboard itself? LOL. If you use exactly the same high-efficiency 90-95% efficient Platinum / Titanium rated PSU for testing, then AC efficiency variance is simply not only a total non-issue but actually far more useful in practice to see how much the the PC actually costs to run.

Bottom line : You can convince yourself otherwise to the contrary to "talk down" high TDP chips, but you pay your bills based on AC consumption, and that's what most people find useful in the real world. Or do you phone up your electricity company and offer to pay only 2/3rds of your bill because you swear blind they "read it wrong".

 

[Abwx]

"Energy per task used" metrics have been around for ages. As Enigmoid accurately said - on top of the higher CPU draw is higher platform draw too, so it's not really surprising some will try and drift away from "total consumption" on the rigs that consume the most... And +12v only consumption reviews are highly variable outside of the Mini-ITX "power brick" builds (where +12v is the only input voltage and +5v often pulls "through" the 12v via DC-DC). Why? Because +12v only on multi-rail ATX PSU's excludes +5v/+3.3v rails which can vary hugely across motherboards of even the same socket. AC accounts for all power draw regardless of 12v vs 5v rail. How is this "professional" to measure only 1 out of 2-3 power inputs in a power consumption test leaving you with highly variable power draw results on the motherboard itself? LOL. If you use exactly the same high-efficiency Platinum rated PSU for testing, then AC efficiency variance is simply not only a non-issue but actually far more useful in practice to see how much the the PC actually costs to run.

Bottom line : You can convince yourself otherwise to the contrary to "talk down" high TDP chips, but you pay your bills based on AC consumption, and that's what most people find useful in the real world. Or do you phone up your electricity company and offer to pay only 2/3rds of your bill because you swear blind they "read it wrong".

The bottom line is that neither you nor Enigmoid understand anything to basic electric laws or you are voluntarly ignoring the datas posted, what you didnt notice, that makes a lot of unnoticed things really, is that they measure the CPU 12V rail power drain and the drain at the main simultaneously, let s check if the power delta at the 12V rail correlate with the delta at the main when accounting with their PSU that has about 90% efficency :

Power at the 12V CPU Rail :

Let s use the 8370E and 4670k numbers, respectively 72 and 60W at the rail level, this should correlate to about 80W and 66 W respectively at the main level, let s see what happens at this main :

The delta at the main is 75W for the 8370E and 66W for the 4670K, very well correlated with their measurement at the relevant secondary of the PSU, you see, you proved that you are just trolling, numbers are in front of yours eyes, use any other CPU in thoses graphs and it will yield the same results.

To end on a personal advice i think that you should take some courses before branding Hardware.fr as incompetent and trying to explain to the rest of us the use of the different voltages on a MB, for your insight 3.3V and 5V cant be used for the CPU, thoses voltages are too low and the VRMs would had to switch 2.4x more current, with a 5V rail, to get the same power than when using the 12V rail.
At 3.3V they would switch 3.63 more current, the losses would increase by much more than thoses ratios, actualy by the square of thoses ratios.

 

[BSim500]

I see zero value in circular childish nit-picking any further. Even if we all "pretend" that the "true" consumption of 8-core FX's is "63w", as you said yourself - do it for all other chips and little has changed in relative terms. I have an i5-3570 that does 4GHz @ 52w with 0.996v vcore (AC measured). It was bought for under $170 - less than a listed FX-8350 at the time, and at 4GHz is same or slightly faster than an i5-4670K @ stock 3.4GHz. Knock off your "20% AC loss" (which itself is a very poor PSU these days - 80 Plus Platinum's / Titanium's are up to 90-96% efficient (rendering 99% of your argument irrelevant unless you're stuck with a 2005 era sh*tty PSU) and power consumption is down to 41.5w. Knock off your "10% VRM loss" and it's down to 37w. Knock off another 6w for the iGPU being disabled from using a dGPU and it's down to 31w, blah, blah, etc, etc. Meanwhile, it's up to 30% faster in every single game posted here including AtenRa's usual "worst-case for Intel's, best case for AMD's" list vs an FX-8350, (and up to 70% faster single-threaded limited games / apps), whilst an 8320E is in turn 10% slower than an FX-8350 that in games and up to 20% in apps). Woohoo, I won the "power efficiency game", do I get a cookie?

Meanwhile, I still can't use my "31w" 4Ghz i5 without plugging it into a motherboard or PSU, so bottom line it's still as utterly pointless to play your funny little "creative accounting" games vs how most people measure them in reality as it was 4 hours ago. Appliances (eg, TV's, fridge/freezers, microwaves, PVR's, monitors, etc) listed power consumption are all measured at the wall for a reason - it gives the most accurate information about cost usage & environmental impact to the end user. By all means continue to play your game by yourself if that's what excites you - I've got some work to do on my "31w load" rig. :biggrin:

 

[AtenRa]

Meanwhile, it's up to 30% faster in every single game posted here including AtenRa's usual "worst-case for Intel's, best case for AMD's" list vs an FX-8350, (and up to 70% faster single-threaded limited games / apps), whilst an 8320E is in turn 10% slower than an FX-8350 that in games and up to 20% in apps).

The OP was asking about Core i3 vs FX8320E OCed, not Core i5. And it was you quoting a single game (AC unity) to make the FX look bad. But even in a worst case scenario for the AMD CPUs like AC Unity, an OCed 4.4GHz FX8320E will be close if not faster than Core i3 using 980 SLI.

So as i have said before, both of those CPUs will be fine for the next few years.

ps: somebody mentioned M2 socket. ASROCK Fatal1ty 990FX Killer has a socket M2 Gen2.0 2x. This is two times faster than Gen 2.0 1x solutions and a feature the majority of Intel socket 1150 motherboards lack in the same price point. Currently at $95 AR on newegg.

Edit: Hell the FX8320E is currently at $110 on newegg, the FX + ASROCK Fatal1ty motherboard = $205 total. That is a very nice budget gaming combo if you are willing to OC a little bit (4GHz to 4.4GHz).

 

[BSim500]

The OP was asking about Core i3 vs FX8320E OCed, not Core i5. And it was you quoting a single game (AC unity) to make the FX look bad. But even in a worst case scenario for the AMD CPUs like AC Unity, an OCed 4.4GHz FX8320E will be close if not faster than Core i3 using 980 SLI.

Sure but it was Abwx who was earlier wailing about "Fx-8xxx's 'destroying' i5's" (as he has trolled on other comments before, as several others have observed). And I originally posted 23x 2014 benchmarks on page 1 (mostly AAA's). If I really wanted to "cherry pick to make the FX look bad", I would have included Skyrim & Starcraft 2 (and many other still popular games) and left out Thief & Watch Dogs. After all, you yourself when benchmarking the A10-7850K choose to pick "APU friendly" games like DOTA2 based on popularity rather than "newness", and Skyrim's still more popular than WatchDogs, right?...

Hell the FX8320E is currently at $110 on newegg, the FX + ASROCK Fatal1ty motherboard = $205 total. That is a very nice budget gaming combo if you are willing to OC a little bit (4GHz to 4.4GHz).

I completely agree, there's some great AMD deals around. But then there's a good reason why AMD has had to slash prices for even some high-end FX-8xxx's to match the i3 rather than i5...

 

[N8Magic]

Wow, had no idea this would get so heated. Any AMD vs. Intel discussion could go that way I guess. My options as of now, stretching the budget a little bit:

AMD FX-8320e - $132
MSI 990FXA-GD65V2 - $69.99 AR
Total: $203 + tax

OR

Intel i5-4440 - $150
MSI H97 Guard-Pro - $79.99 AR
or ASRock Z97 Anniversary - $69.99 AR (I question the build quality of this mobo)
Total: $229 or $219 + tax

I know I could really crank up the clock on the 8320e combo, but the i5 setup would be plug, play and done. Decisions, decisions.

 

[ShintaiDK]

The i5 without the slightest doubt. Specially with so little price difference.

When you start to OC the FX, you may also need a new cooler.

With the FX you are also stuck with PCIe 2.0, no native USB3, much higher power draw/noise etc.

Not that it change much, but the i5 4440 have been replaced with the i5 4460. So check if its not similar price.

 

[BSim500]

If there were $100 in it, I'd say go for the FX-8320e, but for the sake of $16-26 (even that'll fall to zero if you need a cooler to OC), the i5 is a no-brainer ($20 premium is worth it purely for the better features, ie, native USB3, PCI-E 3.0, etc). Edit : As ShintaiDK said, they've been "refreshed" since then with another +100mhz for the same price. i5 4460's max turbo = i5-4670K stock with Turbo Boost disabled.

 

[N8Magic]

The i5 without the slightest doubt. Specially with so little price difference.

When you start to OC the FX, you may also need a new cooler.

With the FX you are also stuck with PCIe 2.0, no native USB3, much higher power draw/noise etc.

That's what i'm thinking as well.

As for the mobo, seems the only substantial difference between H97 and Z97 is with regards to overclocking. H97 can modify the multiplier only (on K-series CPUs) while Z97 can modify multiplier and clock speed. Am I missing anything else?

 

[Arachnotronic]

Wow, had no idea this would get so heated. Any AMD vs. Intel discussion could go that way I guess. My options as of now, stretching the budget a little bit:

AMD FX-8320e - $132
MSI 990FXA-GD65V2 - $69.99 AR
Total: $203 + tax

OR

Intel i5-4440 - $150
MSI H97 Guard-Pro - $79.99 AR
or ASRock Z97 Anniversary - $69.99 AR (I question the build quality of this mobo)
Total: $229 or $219 + tax

I know I could really crank up the clock on the 8320e combo, but the i5 setup would be plug, play and done. Decisions, decisions.

The i5 without a doubt. Great chip.

 

[N8Magic]

FYI, 4460 is 30 bucks more.

Not really worth it for 100MHz IMO. But thanks for mentioning!

 

[NTMBK]

Wow, had no idea this would get so heated. Any AMD vs. Intel discussion could go that way I guess. My options as of now, stretching the budget a little bit:

AMD FX-8320e - $132
MSI 990FXA-GD65V2 - $69.99 AR
Total: $203 + tax

OR

Intel i5-4440 - $150
MSI H97 Guard-Pro - $79.99 AR
or ASRock Z97 Anniversary - $69.99 AR (I question the build quality of this mobo)
Total: $229 or $219 + tax

I know I could really crank up the clock on the 8320e combo, but the i5 setup would be plug, play and done. Decisions, decisions.

The i5, no question. For $20 it's a far better gaming chip.

 

[N8Magic]

Just purchased the i5-4440.

Now to find a motherboard.

 

[Abwx]

I see zero value in circular childish nit-picking any further.

Your values are trolling and willfull misquotes :

You said :

So by your own site - it uses 37w more than an i5-4670K yet simultaneously is supposedly only using 10w more than an i3

To wich i answered :

Where did i write that it use only 10W more than a i3 since i specified that the i3 use about 32-33W..?..

Your post was a blatant lie, no wonder that i got no answer, the extent of all thoses misquotes is not hasardous, it just point that you are a kind of professional troll, unless you can point where i said what you (mis)quoted, wich you cant since you lied...

 

[Gikaseixas]

i5 > FX > i3
You did well OP

 

[Enigmoid]

i5 > FX > i3
You did well OP

I agree completely.

 

[Ken g6]

Just purchased the i5-4440.

Now to find a motherboard.

If you want to go a little cheaper, the B85 chipset is for Basic Budget Business Boards. (I don't actually know what the B stands for, but I think it's one of those.) Your 4440 is not a Haswell Refresh, so it should work with, say, this Gigabyte board.

 

[Phynaz]

Op made the right choice.

 

[2is]

I said that the i5 had 42% advantage last year and 25% currently, it s not a theorical thing it s a fact, the FX is gaining while the i5 is downgraded progressively.

Now, if you look at my post history you ll notice that it is only recently that i ended posting in threads where people are asking for advices, a few times ago i didnt bother much to debate such issues, thing is that a lot of people here are deliberatly giving bad advices because of brand loyalism if not fanatism, let s summarize this thread :

Your theory of FX8 surpassing i5 in the mid-term is not fact, it's your own theory. Similar ones have been thrown around for 8 years and I will repeat, NEVER came to fruition.... But you're different right? Funny guy. The sad part is you're cherry picking individual games that FX does well in and it's still falls short of an i5.

I could care less about your post history. You need to look in the mirror if you're worried about people giving bad advise. Don't even get me started on this new power consumption venture you're on, trying to pretend it's similar to an i5. That's just more crap advise on top of crap advise. Perhaps you should re-evaluate and go back to not giving any.

 

[crashtech]

Not just cherry-picking game benchmarks, but picking benches that have little to no relevance for the average user, like chess engine benches.

All that hot air and tilting at windmills, and the OP still could see the best path.

 

[Abwx]

Your theory of FX8 surpassing i5 in the mid-term is not fact, it's your own theory. Similar ones have been thrown around for 8 years and I will repeat, NEVER came to fruition.... But you're different right? Funny guy.

It s you that is making theories, i posted numbers, and you..?.

Still no numbers.??.

Here a few that i already posted :

http://www.hardware.fr/articles/924-19/indices-performance.html

The sad part is you're cherry picking individual games that FX does well in and it's still falls short of an i5.

Sad part is that you re of bad faith, i posted more lowly threaded games that fully threaded games, if i had cherry picked i would had also checked BF or any other game that is well threaded, i just took what was at disposal at HFR, the overall scores i posted use deliberatly lowly threaded games to get a better picture, i pointed that Hardware.fr removed BF to replace it by a single threaded game, that s what you call biaised..?.

The real reason is that you didnt expect the FX to be better than an i3, hence it s all biaised games benches, of course...

I could care less about your post history. You need to look in the mirror if you're worried about people giving bad advise. Don't even get me started on this new power consumption venture you're on, trying to pretend it's similar to an i5. That's just more crap advise on top of crap advise. Perhaps you should re-evaluate and go back to not giving any.

The power numbers are real, measured at Hardware.fr, it s just that you dont know nothing about the FX E series while i did check the electrical caracteristics, so you hold your ignorance as a better rule that my knowledge about thoses chips.?? a big lol , really..

Of course, i dont expect numbers that would contradict me, but ad hominems, as usual, actualy it s your only "argumenation".


Crap advise are advices that rely on words and never on numbers, i m still waiting for numbers that contradict my sayings, to end this debate, with you at least, i ll add that the i5 in the graphs i posted is 3.4-3.8, the advised i5 is about 10% slower, it will have a 15% lead in games over a FX8350 and this latter will be better by 26% in applications, i estimated two years for the 4670K, this one will be exhausted in one year or so...

 

[2is]

Your own cherry picked gaming charts show i5 > FX so why do I need to give you numbers? The rest of your nonsense is you trying to predict the future that FX will be faster than an i5 in 2015

So lets summarize. You're suggesting someone by an FX over an i5 even though it's slower, based on a prediction that it will be faster in 2015. Fanboyish predictions that have been made for nearly a decade now and have never come to fruition.

And you think that's sound advise? lol

Like I said, if people giving bad advise bothers you, look in the mirror.