Is there any reason to use FX CPUs right now?

[Abwx]

http://images.anandtech.com/graphs/graph7003/55330.png

X264 where AVX2 will help Haswell...

http://www.bit-tech.net/hardware/2013/06/12/intel-core-i5-4670k-haswell-cpu-review/6

Nice pro Intel review, since when the 2600K score with CB 11.5 is 7.49 at stock frequency when all others sites display 6.88-6.92..?..

http://www.xbitlabs.com/images/cpu/core-i7-3770k-i5-3570k/power-2.png

ROFL, Anton Shilov as exemple, the very guy that started massacring AMD once they stopped sending him gear because of his obvious Nvidia shilling..

http://www.hardwareluxx.de/images/s.../2013/haswell-launch/diagramme/strom-last.jpg

This one is a serious site, it was time, after your three first propositions, their power number measure efficency over the whole benches, ill have to dig the thing since i didnt read their reviews, not sure that they ll please you though since the 8320E is granted the same efficency as a 2600K or so but the dicrepancy between SB and HW seems to me too weird, as said i ll dig further the thing.

http://uk.hardware.info/reviews/443...est-energy-consumption-igpu-cinebench-115-max

This one i know, it s a correct site generaly but i dont check it that often.

33.6w idle for Haswell and 43-46W for SB and IB, not difficult to artificialy boost HW perf/watt thanks to unbalanced plateforms, what about using the power delta instead..?

This will show better perfs watt for the older chips in Cinebecnh, as i stated it and you just proved my point.

The move from 32nm to 22nm lowered load power consumption, lowered operating voltages and noticeably improved mobile battery life. My quad core i5-3570 is actually drawing less power under load than my old dual-core i3-530 with 300mv lower voltage at 4GHz. 32nm is not the magic panacea of CPU efficiency and hasn't been for 3 years. Now go find someone else to troll your chronic attention seeking & childish "viral marketing" paranoid conspiracy theories on, kid...

So there was three unworthy sites and two that were of interest, i gave you my opinion for the most professionals, i cant be accurate for Hardware.luxx since my habit is to first read the whole review before producing any opinion, i ll answer later for this one.

That said what would be your reaction if i proposed a site like Bit.tech and its magical 2600K with say a 8350 that would do better than the 4770K in CB..?.

For sure you would use this exemple as a definitive prove that i m no more credible, you can imagine that you re benefiting from quite a lot of doubt benefit that you surely be not willing to grant to other members if the tables were turned, unless they share your opinions of courses...

 

[Phynaz]

Unworthy sites, lol!

 

[TurtleCrusher]

Unworthy sites, lol!

Unless you can discuss otherwise, he explained in detail why they weren't. I understand if you've got no counter to bring to the table. Facts are facts!

 

[BSim500]

Unless you can discuss otherwise, he explained in detail why they weren't.

No he didn't. Example : "33.6w idle for Haswell and 43-46W for SB and IB" is due to Haswell's lower idle state (800MHz vs 1.6GHz plus C7) not "unbalanced platform cheating", AVX2 actually increases Haswell's power consumption not decreases it, etc. In short, just the usual baseless "every site on the net except my beloved hardware.fr is biased" comedy routine in an attempt to "attack" the owners of websites rather than actually deal with the facts. As mentioned, my Ivy quad is drawing less than my old Clarkdale i3 despite having double the cores. That's one hell of a "placebo" for a test that's easily repeatable on a variety of measuring devices. And how 'clueless' ARM, Apple, Samsung, etc, must all be to not see 32nm is the pinnacle of efficiency and "waste" all that money pursuing sub 30nm stuff... :whiste:

It's one thing to argue over the subjective ("an AMD FX chip is better for me personally because I run x"). But fanboys who run around with hallucinatory BS fantasies of "16-22nm has identical electrical power characteristics to 32nm and every site that says otherwise is corrupt & biased" are just making complete fools of themselves. Especially when AMD's own partners openly contradict them:-

http://www.tsmc.com/english/dedicatedFoundry/technology/20nm.htm
http://www.tsmc.com/english/dedicatedFoundry/technology/16nm.htm
http://globalfoundries.com/newsroom...ering-of-14nm-finfet-semiconductor-technology

If AMD were already on 14-22nm, we simply wouldn't be having this conversation...

 

[AtenRa]

No he didn't. Example : "33.6w idle for Haswell and 43-46W for SB and IB" is due to Haswell's lower idle state (800MHz vs 1.6GHz plus C7) not "unbalanced platform cheating"

Dont think so, its more about the Motherboards than the Idle state of the CPU.

AVX2 actually increases Haswell's power consumption not decreases it, etc.

Seams you dont understand the difference between Efficiency and power or energy usage. Haswell with AVX may use more power but it has higher efficiency (perf/watt) than without it. Otherwise Intel would not create AVX in the first place

 

[BSim500]

Dont think so, its more about the Motherboards than the Idle state of the CPU.

It's a combination of both. You can force idle state on Ivy Bridge down to 800MHz using Throttlestop, and the idle drop is noticeable.

Seams you dont understand the difference between Efficiency and power or energy usage. Haswell with AVX may use more power but it has higher efficiency (perf/watt) than without it. Otherwise Intel would not create AVX in the first place

I do understand it, you're simply conflating another issue. The recent claim (before Abwx moved the goalposts again) was effectively "22nm has no measurable efficiency gain over 32nm" based on yet another flaky hardware.fr "review" showing one single Haswell and Sandy at the same load power draw. In general though, this is pure BS as virtually every other major review site has shown (especially at higher frequencies). Plenty of other load power tests taken with non-AVX benchmarks out there. You can test it yourself by running Sandy vs Ivy / Haswell under non-AVX Prime v26.6.

 

[AtenRa]

Plenty of other load power tests taken with non-AVX benchmarks out there. You can test it yourself by running Sandy vs Ivy / Haswell under non-AVX Prime v26.6.

If you only run Prime in stress test only, you cannot measure efficiency since you dont get any results.

Also, prime is a FPU heavy application and Abwx specifically talking about Integer efficiency and not peak power usage.

 

[Andrew Chunis]

Wow this thread is awesome thanks for the info

 

[BSim500]

Also, prime is a FPU heavy application and Abwx specifically talking about Integer efficiency

Well of course he was. It's the main thing AMD has going for it due to the shared CMT module design. So 8-thread integer only tests obviously have to be "preferred" to obtain the "correct" data-set, with everything else rejected as "irrelevant" or "bad coding"... :whiste:

 

[AtenRa]

Well of course he was. It's the main thing AMD has going for it due to the shared CMT module design. So 8-thread integer only tests obviously have to be "preferred" to obtain the "correct" data-set, with everything else rejected as "irrelevant" or "bad coding"... :whiste:

When someone says Bananas are Yellow, you dont counter by saying that this is ridiculous because Apples are Red.

He said that in some applications(integer) the FX8320E is more efficient (perf/watt) than Haswell Core i5. Now if you dont like that or you believe that he is only doing it to show the FX in the best possible way, then that is irrelevant.
If you want to talk about FPU or anything else that is fine, compere Pov-Ray or any other FPU related application and post the perf/watt.

 

[Abwx]

No he didn't. Example : "33.6w idle for Haswell and 43-46W for SB and IB" is due to Haswell's lower idle state (800MHz vs 1.6GHz plus C7) not "unbalanced platform cheating", AVX2 actually increases Haswell's power consumption not decreases it, etc. In short, just the usual baseless "every site on the net except my beloved hardware.fr is biased" comedy routine in an attempt to "attack" the owners of websites rather than actually deal with the facts.

44W idling for both HW and IB including said GFX that add 13W comsumption at the main level, i wont even adress the rest of your post since you dont even bother questionning such sites that show a 2600K scoring 7.49 at stock with Cinebench.

A hint, there s no AVX in Cinebench, it s all SSE2 at a 70% rate.

http://www.hardware.fr/articles/913-9/cpu-consommation-overclocking.html

 

[Abwx]

When someone says Bananas are Yellow, you dont counter by saying that this is ridiculous because Apples are Red.

He said that in some applications(integer) the FX8320E is more efficient (perf/watt) than Haswell Core i5. Now if you dont like that or you believe that he is only doing it to show the FX in the best possible way, then that is irrelevant.
If you want to talk about FPU or anything else that is fine, compere Pov-Ray or any other FPU related application and post the perf/watt.

Vishera is also competitive in FP but in Integer it s much more powerfull than what the usual suspects think, thing is that some people are deliberatly using Cinebench as representative of Integer perf, wich is completely ridiculous, AT s Johan De Gelas said that Piledriver has good integer single thread performance and provided the numbers that say so.

The Opteron core is also better than most people think: at 2.4GHz it would deliver about 2481 MIPs. That is about 80% of Intel's best server core at the moment

There s a typo in the article, actualy it s 2723 Mips at 2.3GHz.

http://www.anandtech.com/show/7757/quad-ivy-brigde-ex-60-cores-120-threads/8

 

[frozentundra123456]

Of course the author himself says in the opening sentence that those two benchmarks are "almost irrelevant". Otherwise perhaps AMD would have more than a single digit share of the server market.

He also says it is better than people think, but "not good enough".

 

[Abwx]

I do understand it, you're simply conflating another issue. The recent claim (before Abwx moved the goalposts again) was effectively "22nm has no measurable efficiency gain over 32nm"

Talk of moving the goalposts, you re an expert on this within this very post that i m quoting.

Typical trick of badfaithers, to write forged quotes in italic to make people think that this lie, an hard habit of yours, is a genuine quote, here what i said :

Lol, official TDP to claim a perf/watt improvement, but since you know HFR let s check the perf/Watt of Haswell comparatively to Sandy Bridge :


http://www.hardware.fr/articles/913-9/cpu-consommation-overclocking.html

Oh, but the 2500K is at the level of the 4670K, and that s thanks to the plateform because at the CPU level the old SB is slightly better, and both are well below the 3570K wich has 20% better perf/watt, to calm down your eventual anger and eventual accusations of bias i ll add that it s only 10% in Cinebench...

The comparison was SB vs HW, not 32nm vs 22nm since i quoted IB, wich is 22nm isnt it, as better than both, anyway this at least has the advantage to debunk that you cant make any point, hence the continual reliance on lies and deliberate forgeries...

From now if you want to quote me use my posts or i ll report your obviously deffamatory behaviour.

 

[Abwx]

Of course the author himself says in the opening sentence that those two benchmarks are "almost irrelevant". Otherwise perhaps AMD would have more than a single digit share of the server market.

He also says it is better than people think, but "not good enough".

You re not quoting what he said accurately, here it is :

I admit, the following two benchmarks are almost irrelevant for anyone buying a Xeon E7 based machine.

That s not the same thing as saying that it s irrelevant, i guess that you edited his saying by putting a period in the meddle of his sentence, indeed he s implying that it hold for non server apps, wich is....i let you conclude..

But let s look further :

The results aren't totally meaningless either, as the profile of a compression algorithm is somewhat similar to many server workloads: hard to extract instruction level parallelism (ILP) and sensitive to memory parallelism and latency. The instruction mix is a bit different, but it's still somewhat similar to many server workloads. And as one more reason to test performance in this manner, the 7-zip source code is available under the GNU LGPL license. That allows us to recompile the source code on every machine with the -O2 optimization with gcc 4.8.1.

So from you claim of irrelevancy it s now "not meaningless", it s obvious that you didnt read the article, or only one line without even reading the whole sentence.

Also notice that they used GCC to compile 7 Zip....

AS for the "not enough" it s in respect of Intel CPUs and certainly not on absolute terms.

 

[BSim500]

From now if you want to quote me use my posts or i ll report your obviously deffamatory behaviour.

To whom, your mom? How old are you, 12? As pointed out previously, you're the one who's been banned before and are still throwing out almost daily BS false accusations against other people of "viral marketing" (ie, being paid to "bash" AMD) which ironically, genuinely is "defamation" - a word you love to throw around without seemingly understanding what it means. So you can come down off your high horse and drop the "purity enforcer" ego power trip, kid... :thumbsdown:

 

[Abwx]

To whom, your mom? How old are you, 12? As pointed out previously, you're the one who's been banned before and are still throwing out almost daily BS false accusations against other people of "viral marketing" (ie, being paid to "bash" AMD) which ironically, genuinely is "defamation" - a word you love to throw around without seemingly understanding what it means. So you can come down off your high horse and drop the "purity enforcer" ego power trip, kid... :thumbsdown:

What is the relevance of a 4 year old post in respect of this thread.?.

To the moderators since you are constantly using this trick that end only derailing the thread or forcing me to re post what i already posted to prove that you forged sayings.

This say that you have no argument and that you re deliberatly trying to shift the technical debate in a personal matter irrespective of the discussed datas, i already pointed that you re willfully creating false quotations, exemple above with the 32nm vs 22nm false quote of yours, its like you dont even bother being proved as lying.

That said your posts are, paradoxaly, not that useless since they are a convenient mean for me to burst all urban legends, myths and other hearsay based badmouthings that are used to trash AMD s products.

 

[monstercameron]

To whom, your mom? How old are you, 12? As pointed out previously, you're the one who's been banned before and are still throwing out almost daily BS false accusations against other people of "viral marketing" (ie, being paid to "bash" AMD) which ironically, genuinely is "defamation" - a word you love to throw around without seemingly understanding what it means. So you can come down off your high horse and drop the "purity enforcer" ego power trip, kid... :thumbsdown:

I do this all too often myself, after a seemingly pointless debate, I might do a little poking. Might want to discuss why piledriver can be more efficient in some workloads than attacking abwx. It's beneath you.

It would be interesting to see a new, in depth and professional review showing the strengths of the fx platform as a last hurrah.

 

[lord_emperor]

Over the holidays I got a new system on a really restrictive budget. AMD FX-8310 OEM + Hyper 212 EVO for a combined cost of $135CAD. OC'ed to 4.0/4.8GHz running cool.

Took my savings and put them toward a better video card, R9 290. I think I got a better overall system than had I spend $200+ on an i5 CPU or $280+ for any "K" version.

 

[SlowSpyder]

Over the holidays I got a new system on a really restrictive budget. AMD FX-8310 OEM + Hyper 212 EVO for a combined cost of $135CAD. OC'ed to 4.0/4.8GHz running cool.

Took my savings and put them toward a better video card, R9 290. I think I got a better overall system than had I spend $200+ on an i5 CPU or $280+ for any "K" version.

I'm a big proponent of big GPU + moderate CPU vs. big CPU + moderate GPU, if you are on a limited budget. At 4.8GHz the FX is quite snappy, even if power hungry.

 

[frozentundra123456]

I see it the opposite way. I would prefer to spend a bit more on the cpu even at the expense of some gpu, since

1. it is easier to upgrade the gpu than the cpu
2. gpus are advancing faster than cpus, so the future upgrade path for gpus is better
3. it is easier to adjust settings for lack of gpu power than cpu power.

I know this is at the sacrifice of some current performance, so not saying either of us is right or wrong, just my take.

 

[SlowSpyder]

I see it the opposite way. I would prefer to spend a bit more on the cpu even at the expense of some gpu, since

1. it is easier to upgrade the gpu than the cpu
2. gpus are advancing faster than cpus, so the future upgrade path for gpus is better
3. it is easier to adjust settings for lack of gpu power than cpu power.

I know this is at the sacrifice of some current performance, so not saying either of us is right or wrong, just my take.

My take:

1 - At least with AMD, upgrading a CPU is a five minute job and no drivers to mess with. Intel might be a bit more complicated, I don't know. Either way, it's an opinion, and I disagree. As long as I'm using air cooling and the factory cooler mounts, upgrading a CPU is pretty easy (not that upgrading a graphics card is difficult either, but driver changes can make it much hairier job in my opinion... I went from one 5870 to CF 5870's. Had to reinstall Windows to make it work :thumbsdown: )

2 - I agree, GPU's are advancing faster than CPU's advance. But I think games are built with that in mind. We saw from the last generation of consoles, computer gaming is a bit of an after-thought. We get the crumbs that are ported over. We've seen plenty of games where a lowly FX 4300 or i3 can run the game at more than acceptable frame rates.

3 - Agree. Some shadow and physics settings in games seem to be related to the CPU, but the vast majority of setting have to do with your graphics card. But, if your choice is to turn a bunch of settings down to get playable frame rates so the game looks 'blah' or deal with lower frame rates but can have lots of eye candy, I guess that's personal taste, though neither scenario is ideal. The handful of games I've seen that really struggle on FX CPU's didn't exactly fly on their much more expensive counterparts from Intel from what I've seen. Anyway, I agree with this one, but I just don't think it is that big of a deal, but the GPU settings do give you more flexibility for tweaking a game.

 

[MiddleOfTheRoad]

I am being serious. The FX can be quite close, maybe even surpasses the i5 in perforamnce/watt in a handful of multi-threaded benches it appears, depending on the model CPU.

Power use, the FX 8370E uses 37% more power than an i5 4670 when fully loaded going by this graph that was posted.. The FX 8370E has a base clock of 3.3GHz. An FX 8320E has a base clock of 3.2GHz, so the multi-threaded performance difference between the two shouldn't be very different.

+1
I'm really not sure why so many people have trouble understanding what you're saying. Maybe it is how you are explaining it -- Even if the FX 8370E uses 37% more power, it can process twice the threads (a 100% advantage) at once compared to the i5. Even with the slower AMD cores, you are still going to be able to process more simultaneously for heavily multithreaded workloads. It's really not a difficult concept to grasp.

 

[SlowSpyder]

I see it the opposite way. I would prefer to spend a bit more on the cpu even at the expense of some gpu, since

1. it is easier to upgrade the gpu than the cpu
2. gpus are advancing faster than cpus, so the future upgrade path for gpus is better
3. it is easier to adjust settings for lack of gpu power than cpu power.

I know this is at the sacrifice of some current performance, so not saying either of us is right or wrong, just my take.

Not to beat a dead horse, but a 4.0/4.8GHz FX with a cooler for $135 is hard to beat. A Pentium system with an overclock would be a good option too, and has a better upgrade path, I'll concede that. But for the performance now and how it'll run future games without putting another penny into it for an upgrade, I think that's a tough combo to beat. Of course no overclock is guaranteed. But, I'd be quite surprised to see an FX that couldn't reach 4GHz base.

 

[SlowSpyder]

+1
I'm really not sure why so many people have trouble understanding what you're saying. Maybe it is how you are explaining it -- Even if the FX 8370E uses 37% more power, it can process twice the threads (a 100% advantage) at once compared to the i5. Even with the slower AMD cores, you are still going to be able to process more simultaneously for heavily multithreaded workloads. It's really not a difficult concept to grasp.

Core for core, Intel get's more done. But, I could see a situation where a very well threaded piece of software that could take advantage of all eight integer cores of an FX could best an i5 in efficiency, or at least get very close. Obviously that's not going to be the norm, and I'm not trying to argue it is. Abwx made a comment that sometimes the FX can actually be more efficient, the numbers posted seem to support that. That's a niche case, but it looks like it isn't outside the realm of possibility in some situations. Pretty impressive for the FX which is generally panned here. If anything I think that shows how ho-hum CPU releases have been since Sandy Bridge on Intel's side. And AMD has had no real release other than 95 watt TDP FX's.

I miss the old days, the Pentium 3/4, the Athlons. We'd get a launch, and it seemed like every few months a new faster version would come. And then a new CPU architecture and tweaked parts from both companies. The good ol days when process technology was improving fast and there was real competition. I remember being so angry when the K6III launched, there weren't reviews right away. I'm aging myself here... haha. Things have changed...!