Sign in with
Sign up | Sign in
Your question

AMD FX 8350 or Intel I7 3770K ?

Last response: in CPUs
Share
February 1, 2013 6:59:43 AM

i just want to know which one will be better for a gaming computer.. A Core i7 3770k with sabertooth Z77 mobo or AMD FX 8350 with sabertooth 990FX ?



More about : amd 8350 intel 3770k

a b à CPUs
February 1, 2013 8:49:43 AM

3770k w/o question. If anyone tells you otherwise they are lying to you and themselves. The 3770k is the better CPU for EVERYTHING. I don't care what you're doing with your PC>

m
0
l
Related resources
a c 101 à CPUs
a b À AMD
February 1, 2013 8:56:20 AM

The Core i7 is faster but costs a lot more. The FX-8350 delivers quite good value in this comparison.

Thing is, for a gaming system you should be comparing it to the Core i5-3570k. That's still faster than the FX-8350, and the price gap more or less disappears.
m
0
l
February 1, 2013 4:45:53 PM

I like AMD and competition, but reality is without question 3570K or 3770K.

m
0
l
a b à CPUs
February 1, 2013 4:55:58 PM

If your looking to save money the amd 8350 is a fine cpu however the amd 8320 is cheaper and can be overclocked to the same level. If your going intel gamin a i5 3570k foo sure
m
0
l
May 20, 2013 10:11:45 AM

With recent price drops the FX8350 is a huge value. In the benchmarks the i7 has and edge but the quad core i7s are significantly higher priced than the FX. I recently built a rig and the mobo, mem (bundle deal from newegg), and proc costs were $230. You can't buy a i7 quad core proc for that low of a price.

The Fx consumes more power but I NEVER consider power usage since its a "sunk cost" (does anyone really monitor their power usage?).

Given the price/perfomance value I can't see how you couldn't consider using this proc.
m
0
l
a c 101 à CPUs
a b À AMD
May 20, 2013 10:15:37 AM

maccrory said:
With recent price drops the FX8350 is a huge value.

The FX-8350 costs the same $190 now that it cost back in February.
m
0
l
May 20, 2013 10:54:15 AM

Sakkura said:
maccrory said:
With recent price drops the FX8350 is a huge value.

The FX-8350 costs the same $190 now that it cost back in February.


I picked one up on sale for $160 - and $66 for the mobo/memory from newegg (still on sale now). THATS VALUE!!!
m
0
l
a c 101 à CPUs
a b À AMD
May 20, 2013 11:06:37 AM

maccrory said:
I picked one up on sale for $160 - and $66 for the mobo/memory from newegg (still on sale now). THATS VALUE!!!

You can always be lucky and stumble on a good sale. That goes for Intel too. For example, you can get the Core i5-3570K for $169 at Microcenter right now, so it's actually $21 cheaper than the FX-8350.
m
0
l
May 20, 2013 11:33:54 AM

Sakkura said:
maccrory said:
I picked one up on sale for $160 - and $66 for the mobo/memory from newegg (still on sale now). THATS VALUE!!!

You can always be lucky and stumble on a good sale. That goes for Intel too. For example, you can get the Core i5-3570K for $169 at Microcenter right now, so it's actually $21 cheaper than the FX-8350.


Agreed - cheapest i5 i see on newegg (benchmark) right now is $180. I'm all about value/performance - don't necessarily need the highest end. Given 8 cores (granted not = performance) i would still give the edge to the FX on a value purchase. AMD will NEVER = the performance of Intel but if its close and significantly cheaper (vs i7 and most i5s) i will always buy value.
m
0
l
a b à CPUs
May 20, 2013 11:46:21 AM

maccrory said:
The Fx consumes more power but I NEVER consider power usage since its a "sunk cost" (does anyone really monitor their power usage?).


People who are trying to cut back on their electric bill as much as possible?
m
0
l
a c 803 à CPUs
a c 153 À AMD
May 20, 2013 12:31:26 PM

Depends on your budget. You can get a better graphics card going with an AMD rig than you can an Intel one for the same amount of money. Most games rely on graphics more so than CPU. If I were to build a gaming rig today, it would be 8320/8350 based so I could spend more on GPU. Too often I have had issues where the CPU was up to the task, but the GPU was not. GPU's tend to get outdated faster than CPU it seems. Even after all this time, Core 2 Quads are capable of handling even modern games except the extremely CPU intensive ones. A GPU bought at the same time, probably wouldn't game so well today. That is why in a gaming rig, I always try to get the fastest GPU I can without sacrificing too much on the CPU side. The FX 8320/8350 meets that requirement.
m
0
l
May 31, 2013 4:08:22 AM

amd is a respectfull company but between these two no doubt the 3770k!its way more powerful and a better overclocker!
m
0
l
a c 210 à CPUs
a b À AMD
June 4, 2013 4:47:21 PM

The FX8350 is a far better overclocking CPU...I have seen as high as 5.559 GHz daily overclock on one. He was using external evaporative cooling to keep it cool, but it is LIGHTNING fast.

Performance at lower clocks favors the i7-3770k...but if you can get 5.0+ GHz out of the 8350, it will close the gap pretty well. The areas where it already excels, it will blow the doors off anything else. The biggest issue with the 8350 is the single thread performance...once AMD closes that gap, they're already ahead in multi threaded applications.
m
0
l
June 6, 2013 2:22:01 AM

i think the 8350 has the world record of overclocking for a few mhz from one 3770k but the second one had better celsius(even with chemicals)
m
0
l
June 6, 2013 3:48:40 AM

3770k for sure.
m
0
l
September 8, 2013 3:18:55 AM

Now that 8 core cpu might become a standard platform for gaming thanks to amd winning the next gen console battle. Upcoming titles might run smoother on 8 core standard cpu. But as of now i7 is your ferrari.
m
0
l
September 8, 2013 3:56:33 AM

Depends on the budget but if you have enough for both then go with the intel i7 3770k , with the amd fx-8350 all 8 cores wont be used because games only use up to 2 - 4 cores, So that wont be use full but other than that both CPU's are great but the i7 would be better to go with :D .
m
0
l
November 1, 2013 12:49:04 AM

bigj1985 said:
3770k w/o question. If anyone tells you otherwise they are lying to you and themselves. The 3770k is the better CPU for EVERYTHING. I don't care what you're doing with your PC>



Well, you obviously don't write MPI (message passing interface) parallel processing code in Fortran or C. I write finite element codes for solving large couple PDEs in 2 and 3 dimensions. You can solve bigger problems faster using all 8 cores of the AMD, compared to only 4 of the Intel. You can then port the codes to super-computers later to use hundreds of cores (or nodes).

Basically you can get greater throughput with more cores, even if each core runs at a lower frequency. I just got the AMD 8 core and it turns rounds problems ~1.9 times faster than my quad core i3 PC. I'd expect about a ~1.4 to 1.5 increase on a quad core i7. When run times are many hours, that's still a huge increase.

At the lab I work at, we just bought a new super computer with 20*more processors/nodes(i.e cores) (4500 processors in total) even though each processor was 6 times slower than our previous super computer. The new machine was cheaper per processor (i.e. no where near 20 times more exensive!), and throughput is greater.

OK if your games are limited to running on 4 cores, then go for the fastest quad core processor you can get.

But when I run my codes I choose the number of cores I want: mpirun -np 8 "code" (8 being the number of cores).
m
0
l
a b à CPUs
November 1, 2013 1:58:54 AM

Paul Z said:
bigj1985 said:
3770k w/o question. If anyone tells you otherwise they are lying to you and themselves. The 3770k is the better CPU for EVERYTHING. I don't care what you're doing with your PC>



Well, you obviously don't write MPI (message passing interface) parallel processing code in Fortran or C. I write finite element codes for solving large couple PDEs in 2 and 3 dimensions. You can solve bigger problems faster using all 8 cores of the AMD, compared to only 4 of the Intel. You can then port the codes to super-computers later to use hundreds of cores (or nodes).

Basically you can get greater throughput with more cores, even if each core runs at a lower frequency. I just got the AMD 8 core and it turns rounds problems ~1.9 times faster than my quad core i3 PC. I'd expect about a ~1.4 to 1.5 increase on a quad core i7. When run times are many hours, that's still a huge increase.

At the lab I work at, we just bought a new super computer with 20*more processors/nodes(i.e cores) (4500 processors in total) even though each processor was 6 times slower than our previous super computer. The new machine was cheaper per processor (i.e. no where near 20 times more exensive!), and throughput is greater.

OK if your games are limited to running on 4 cores, then go for the fastest quad core processor you can get.

But when I run my codes I choose the number of cores I want: mpirun -np 8 "code" (8 being the number of cores).


I stopped reading when you said quad core i3...
m
0
l
November 1, 2013 5:21:28 AM

maccrory said:
10835559,0,551379 said:
10835500,0,1323692 said:
"AMD will NEVER = the performance of Intel ....".
said:
said:


Be careful, never is a strong word! In fact, this is simply not true! Remember something called the Athlon series when Intel was producing the Pentium III series. The competition was neck to neck, sometimes with AMD leading in performace. Many enthusiasts felt the Athlon was a superior series..and later entries in the Athlon lineup were very overclockable too!

m
0
l
a c 101 à CPUs
a b À AMD
November 1, 2013 6:21:33 AM

Paul Z said:

Basically you can get greater throughput with more cores, even if each core runs at a lower frequency. I just got the AMD 8 core and it turns rounds problems ~1.9 times faster than my quad core i3 PC. I'd expect about a ~1.4 to 1.5 increase on a quad core i7. When run times are many hours, that's still a huge increase.

The Core i3 has 2 cores, so your data indicates a Core i7 would outperform the 8-core AMD CPU.
m
0
l
a c 803 à CPUs
a c 153 À AMD
November 1, 2013 6:37:12 AM

NotSoCivil said:
Depends on the budget but if you have enough for both then go with the intel i7 3770k , with the amd fx-8350 all 8 cores wont be used because games only use up to 2 - 4 cores, So that wont be use full but other than that both CPU's are great but the i7 would be better to go with :D .


I would like to introduce you to Crysis 3 and BF4 and probably future titles going forward with the next gen consoles running 8 core AMD chips. Intel will still probably be faster, but AMD will be more competitive than it has been.
m
0
l
a c 101 à CPUs
a b À AMD
November 1, 2013 7:02:43 AM

logainofhades said:
I would like to introduce you to Crysis 3 and BF4 and probably future titles going forward with the next gen consoles running 8 core AMD chips. Intel will still probably be faster, but AMD will be more competitive than it has been.


m
0
l
a c 803 à CPUs
a c 153 À AMD
November 1, 2013 7:09:29 AM

Sakkura said:
logainofhades said:
I would like to introduce you to Crysis 3 and BF4 and probably future titles going forward with the next gen consoles running 8 core AMD chips. Intel will still probably be faster, but AMD will be more competitive than it has been.




Not sure what to think of the THG benchmarks as they don't match others I have seen.
http://www.bf4blog.com/battlefield-4-retail-gpu-cpu-ben...
m
0
l
a c 101 à CPUs
a b À AMD
November 1, 2013 7:11:43 AM

So a blog(!) referencing some Russian testing is more reliable than Tom's. Right.
m
0
l
a c 803 à CPUs
a c 153 À AMD
November 1, 2013 7:23:32 AM

THG isn't the god of reviewing it used to be. I look at other sources too. I have the same attitude towards Anandtech and power consumption numbers. Not sure about now, but awhile back theirs were always in left field of everyone else. Hell, look at the complaints on the CPU and GPU of the month lists with regards to errors. You have reviews that contradict the lists and yet the lists rarely get updated properly to show it. It took forever for the FX 6300 to even get an honorable mention, despite the fact people already knew it was a better overall gaming CPU compared to an i3. Then there was the FX 4130 making the list being called a Vishera/Piledriver part. :lol:  Face it, the quality has dropped. I am skeptical of everything. Not to mention the Russian site showed more effort and tested a lot more CPU's. 20 vs 6. Why would someone go through that much effort just to lie?
m
0
l
a b à CPUs
November 1, 2013 7:30:55 AM

logainofhades said:
THG isn't the god of reviewing it used to be. I look at other sources too. I have the same attitude towards Anandtech and power consumption numbers. Not sure about now, but awhile back theirs were always in left field of everyone else. Hell, look at the complaints on the CPU and GPU of the month lists with regards to errors. You have reviews that contradict the lists and yet the lists rarely get updated properly to show it. It took forever for the FX 6300 to even get an honorable mention, despite the fact people already knew it was a better overall gaming CPU compared to an i3. Then there was the FX 4130 making the list being called a Vishera/Piledriver part. :lol:  Face it, the quality has dropped. I am skeptical of everything. Not to mention the Russian site showed more effort and tested a lot more CPU's. 20 vs 6. Why would someone go through that much effort just to lie?


As a native russian speaker I would like to point out that at this time those are single player benchmarks on that site.

They intend to do multiplayer one at a later when servers become stable and some bugs get fixed.

Tom's results are irrelevant as well, as those are from beta and there was some massive performance fix after those benches.

There are some actual very detailed BF4 release multiplayer CPU benchmarks by PClab, but I won't link them, since then I will be accused of a crime against humanity by a certain parties...
m
0
l
a c 803 à CPUs
a c 153 À AMD
November 1, 2013 7:34:01 AM

Thanks for input there. Didn't they do multiplayer for the beta benches too? Work limits how much of that kind of browsing I can do, so I cannot go back and look right now.
m
0
l
a c 101 à CPUs
a b À AMD
November 1, 2013 8:10:33 AM

logainofhades said:
THG isn't the god of reviewing it used to be. I look at other sources too. I have the same attitude towards Anandtech and power consumption numbers. Not sure about now, but awhile back theirs were always in left field of everyone else. Hell, look at the complaints on the CPU and GPU of the month lists with regards to errors. You have reviews that contradict the lists and yet the lists rarely get updated properly to show it. It took forever for the FX 6300 to even get an honorable mention, despite the fact people already knew it was a better overall gaming CPU compared to an i3. Then there was the FX 4130 making the list being called a Vishera/Piledriver part. :lol:  Face it, the quality has dropped. I am skeptical of everything. Not to mention the Russian site showed more effort and tested a lot more CPU's. 20 vs 6. Why would someone go through that much effort just to lie?

The list is not a review. It specifically says that it's not ever going to be perfectly accurate, because it's not possible (both because a one size fits all performance metric is impossible, and because they don't devote many resources to those articles). Pointing to that to discredit their actual reviews is a red herring.

As for the Russian site having more CPUs tested, that's an indicator of sloppy testing. The less time you spend on getting the most reliable data, the more hardware configurations you can squeeze into your review.
m
0
l
a c 803 à CPUs
a c 153 À AMD
November 1, 2013 8:17:33 AM

I see it as not being lazy and giving people a better idea of what to expect. I would rather see a wide range of options, giving me a better idea how a particular setup will run instead of seeing just a crappy faildozer, an ivy i3, Vishera FX, SNB quad and SNB-E hex. It isn't perfect, but still I find it better. I am equally annoyed by the GPU portion of both reviews that use a i7 extreme to test a low to midrange GPU, that shouldn't be paired with anything more than a FX 6300 or a low end i5. Yea it might give a user a better idea of pure GPU performance difference, but who is going to pair a $1k cpu with a $150 GPU? :lol:  The less informed are going to go and buy an FX 6300 or an i5 3350p and pair it up with an HD 7850 and then sit and wonder why hey aren't getting close to the same results as the GPU reviews. Balanced approach is preached on here quite frequently, yet rarely does any review display any kind of balance. Sadly, most review sites have this problem.
m
0
l
a c 101 à CPUs
a b À AMD
November 1, 2013 9:37:05 AM

So in short, you'd prefer unreliable reviews over reliable ones.
m
0
l
a c 803 à CPUs
a c 153 À AMD
November 1, 2013 10:07:01 AM

Sakkura said:
So in short, you'd prefer unreliable reviews over reliable ones.


No I prefer realistic reviews. A $500-1000+ CPU with, say, a GTX 650, is not in any shape or form a realistic scenario. Why preach balance buy never show a balanced system for review? Yea it might show which GPU is better over another. It does little to help an uninformed person running, say an i5 750, for them to know realistically what kind of FPS they will get. People will go out and choose the card that got 60fps in the review to find out that they are only getting 45 at best. Using a more realistic and balance approach would do far better to the uniformed user. At least add a realistic option or two for a review. I don't expect every review to show every combo out there, but an FX 6300 with a low to midrange spread, and something like an i5 4670k with midrange to high end, would be better for people that don't know any better. That would cover far more people than showing the $1k CPU with a $150 gpu.
m
0
l
a c 210 à CPUs
a b À AMD
November 1, 2013 11:12:39 AM

Sakkura said:
So in short, you'd prefer unreliable reviews over reliable ones.


You make the false assumption that they did not have enough time to test them all thoroughly. Which is a critical error on your part. That same Russian site was also the first site to notice that HTT on i7 CPUs in Crysis 3 actually performed worse than without HTT. Crytek then later patched the game to improve HTT performance and prevent it from hindering...now the results are merely the same.

So, don't make assumptions...you know what they say about the word "assume" right?
m
0
l
a c 101 à CPUs
a b À AMD
November 1, 2013 11:42:21 AM

logainofhades said:
Sakkura said:
So in short, you'd prefer unreliable reviews over reliable ones.


No I prefer realistic reviews. A $500-1000+ CPU with, say, a GTX 650, is not in any shape or form a realistic scenario. Why preach balance buy never show a balanced system for review? Yea it might show which GPU is better over another. It does little to help an uninformed person running, say an i5 750, for them to know realistically what kind of FPS they will get. People will go out and choose the card that got 60fps in the review to find out that they are only getting 45 at best. Using a more realistic and balance approach would do far better to the uniformed user. At least add a realistic option or two for a review. I don't expect every review to show every combo out there, but an FX 6300 with a low to midrange spread, and something like an i5 4670k with midrange to high end, would be better for people that don't know any better. That would cover far more people than showing the $1k CPU with a $150 gpu.

But adding confounding factors to a test is ridiculously stupid and makes the review much less reliable. So you effectively want reviews that are completely useless.
m
0
l
a c 101 à CPUs
a b À AMD
November 1, 2013 11:43:20 AM

8350rocks said:
Sakkura said:
So in short, you'd prefer unreliable reviews over reliable ones.


You make the false assumption that they did not have enough time to test them all thoroughly. Which is a critical error on your part. That same Russian site was also the first site to notice that HTT on i7 CPUs in Crysis 3 actually performed worse than without HTT. Crytek then later patched the game to improve HTT performance and prevent it from hindering...now the results are merely the same.

So, don't make assumptions...you know what they say about the word "assume" right?

You're making assumptions about what I think. So you might want to take your own advice there.
m
0
l
a c 210 à CPUs
a b À AMD
November 1, 2013 12:04:30 PM

Sakkura said:
8350rocks said:
Sakkura said:
So in short, you'd prefer unreliable reviews over reliable ones.


You make the false assumption that they did not have enough time to test them all thoroughly. Which is a critical error on your part. That same Russian site was also the first site to notice that HTT on i7 CPUs in Crysis 3 actually performed worse than without HTT. Crytek then later patched the game to improve HTT performance and prevent it from hindering...now the results are merely the same.

So, don't make assumptions...you know what they say about the word "assume" right?

You're making assumptions about what I think. So you might want to take your own advice there.


You called them "unreliable"...I pointed out your logical fallacy. Who assumed what?
m
0
l
a c 101 à CPUs
a b À AMD
November 1, 2013 3:18:53 PM

8350rocks said:
You called them "unreliable"...I pointed out your logical fallacy. Who assumed what?

There was no logical fallacy on my part. On your part, though, there was an assumption - that I thought they didn't have enough time to test all of the hardware thoroughly. But it's just as likely that they didn't have the skill to do so. Or the connections.
m
0
l
a b à CPUs
November 1, 2013 3:43:03 PM

Sakkura said:
The Core i7 is faster but costs a lot more. The FX-8350 delivers quite good value in this comparison.


The 8350 has poor value. One can just get the 8320 and make up the $40 saved with an good heatsink which can at the very least match the 8350's turbo core of 4.2GHz.
m
0
l
a c 101 à CPUs
a b À AMD
November 1, 2013 3:49:24 PM

Lessthannil said:
Sakkura said:
The Core i7 is faster but costs a lot more. The FX-8350 delivers quite good value in this comparison.


The 8350 has poor value. One can just get the 8320 and make up the $40 saved with an good heatsink which can at the very least match the 8350's turbo core of 4.2GHz.

That's why I said "in this comparison". If you compare the FX-8350 and the Core i7 and ignore all other CPUs, the 8350 is good value.
m
0
l
a c 210 à CPUs
a b À AMD
November 1, 2013 4:06:02 PM

Sakkura said:
8350rocks said:
You called them "unreliable"...I pointed out your logical fallacy. Who assumed what?

There was no logical fallacy on my part. On your part, though, there was an assumption - that I thought they didn't have enough time to test all of the hardware thoroughly. But it's just as likely that they didn't have the skill to do so. Or the connections.


No, you assumed the reviews were bad, whatever reason for it is your own assumption.

I pointed that out, and said that it is entirely possible given the history of the site, and it's technical attention to detail, that they are good reviews.

You made a subjective statement, with an assumptive slant toward the fact that you thought the reviews couldn't be valid for one reason or another.

/argument closed.
m
0
l
a c 101 à CPUs
a b À AMD
November 1, 2013 4:16:53 PM

8350rocks said:
No, you assumed the reviews were bad, whatever reason for it is your own assumption.

I pointed that out, and said that it is entirely possible given the history of the site, and it's technical attention to detail, that they are good reviews.

You made a subjective statement, with an assumptive slant toward the fact that you thought the reviews couldn't be valid for one reason or another.

/argument closed.

I did not assume they were bad, I provided evidence that they were.

And you respond with a childish argument rather than actual evidence.
m
0
l
a c 210 à CPUs
a b À AMD
November 1, 2013 4:31:38 PM

Really...and what evidence was that? Your speculative post above, that you mentioned their disclaimer as a detractor?

I think that's refreshingly honest for someone to post the fact that their benchmarks are only the best possible testing that could be done, and are not 100% indicative of the possible performance expectations.

Benchmarks should be treated as a "guideline" and not a bible. They are the result of someone building a program to test whatever it is they want to test, and generating a score, or result.

No code is ever perfect, no hardware is 100% exactly the same, and no software can give a 100% reproducible result every time when it is testing hardware.

As with anything, when viewing benchmarks, YMMV should be a standard disclaimer with any reputable website.

Additionally, you've evidently never read the many review sites you "trust" because their fine print say things like..."Margin for error is expected to be X-Y% for these tests", and, "These tests are representative of the results of several tests, and reflect the average performance of those test runs".

I suppose you would point to that to detract from your benchmarks as well? If not, then that would be hypocritical...

m
0
l
November 3, 2013 11:42:48 PM

Gaidax said:
Paul Z said:
bigj1985 said:
3770k w/o question. If anyone tells you otherwise they are lying to you and themselves. The 3770k is the better CPU for EVERYTHING. I don't care what you're doing with your PC>



Well, you obviously don't write MPI (message passing interface) parallel processing code in Fortran or C. I write finite element codes for solving large couple PDEs in 2 and 3 dimensions. You can solve bigger problems faster using all 8 cores of the AMD, compared to only 4 of the Intel. You can then port the codes to super-computers later to use hundreds of cores (or nodes).

Basically you can get greater throughput with more cores, even if each core runs at a lower frequency. I just got the AMD 8 core and it turns rounds problems ~1.9 times faster than my quad core i3 PC. I'd expect about a ~1.4 to 1.5 increase on a quad core i7. When run times are many hours, that's still a huge increase.

At the lab I work at, we just bought a new super computer with 20*more processors/nodes(i.e cores) (4500 processors in total) even though each processor was 6 times slower than our previous super computer. The new machine was cheaper per processor (i.e. no where near 20 times more exensive!), and throughput is greater.

OK if your games are limited to running on 4 cores, then go for the fastest quad core processor you can get.

But when I run my codes I choose the number of cores I want: mpirun -np 8 "code" (8 being the number of cores).


I stopped reading when you said quad core i3...


Well you shouldn't have. It was a typo, my old desktop is actually a quad core i5, not at i3. My mistake. A simple 'top' command in my linux system shows me exactly how many 'cores' are running.

Here's an example of the MPI code I run in my codes:

INCLUDE "mpif.h"

! MPI variables

INTEGER(KIND=4) :: myid,numprocs,ierr
!---------------------------------------------------------------------------------------------------------
! Initialise MPI

call MPI_INIT(ierr)
call MPI_COMM_RANK(MPI_COMM_WORLD,myid,ierr)
call MPI_COMM_SIZE(MPI_COMM_WORLD,numprocs,ierr)

CALL MPI_FINALIZE(ierr)

The AMD 8-core has greater throughput than the i5 quad. I can run mpirun -np 8 <code> on 8 core the AMD, I cannot with the i5 or a quad core i7 (if an i7 is quad core).

So there.
m
0
l
November 3, 2013 11:48:59 PM

Sakkura said:
Paul Z said:

Basically you can get greater throughput with more cores, even if each core runs at a lower frequency. I just got the AMD 8 core and it turns rounds problems ~1.9 times faster than my quad core i3 PC. I'd expect about a ~1.4 to 1.5 increase on a quad core i7. When run times are many hours, that's still a huge increase.

The Core i3 has 2 cores, so your data indicates a Core i7 would outperform the 8-core AMD CPU.



Sorry, that was a damn typo on my part. My old desktop is a quad core i5, not an i3. And EVERYTHING I said is still true. An AMD 8 core offers greater throughput for FEM/finite difference code simulations than a faster quad core processor. This is generally reflected in the super computer market where there is a move towards greater bytes/flop ratios. Slower processors (but more of them) are cheaper to buy and run (i.e cool). See my other post.
m
0
l
November 4, 2013 12:16:51 AM

Gaidax said:
11835661,0,1465027 said:

I stopped reading when you said quad core i3...
said:


And just to add, an i3 is ctually hyper threaded, so it CAN run 4 separate processes when running MPI code. Typing "mpirun -np 4 <code>" will launch four separate processes on my i3 laptop, each running at 100%, as shown by a top command. How they grab memory relative to four 'true' cores of a non hyper-threaded i5 is a different issue. Plus two hyper threaded cores will not be as fast as the four true cores of the i5, about 30% slower for the same clock speed from what I read (and from comapring code runs on my i3 laptop v old i5 desktop).

My points about throughput are still completely valid, whether it be a hyperthreaded 'pseudo' 4 core i3 or a 'true' 4 core i5 that can run four separate processes (or threads) versus an 8 core (or quad module-dual core!) that can run eight.

Clearly the 4 thread dula core i3 will have the slowest run time for a given simulation, the AMD the fastest. And I can prove it, because I write the codes and run such simulations all the time.
m
0
l
a b à CPUs
November 4, 2013 1:22:44 AM

Real numbers would be nice, because all what you say so far is a load of assumptions...

Besides, I think we are done with I3? Why bring it back now?
m
0
l
November 5, 2013 11:02:06 PM

Gaidax said:
Real numbers would be nice, because all what you say so far is a load of assumptions...

Besides, I think we are done with I3? Why bring it back now?


Of course i3 is outdated now. The point I am making is that the i3 and i5 quad thread/core chips will both be outperformed by an 8 thread/core chip running at a similar clock speed when running a code that can use all 8 threads. It is pretty obvious, and if you spend any time writing and running parallel codes using Message Passing Interface you would appreciate this.

Now the 6 core i7-970 3.3Ghz chip with its 12Mb L3 cache might complete a given simulation quicker (using all 6 threads) than the 8 core AMD FX8350 4.0Ghz chip with its 8Mb L3 (using all 8 threads), but only if you structure your code to take advantage of the larger cache. This can be done, for example, by structuring the solution arrays of a DG FE mesh such that the sequential solution of each block of elements accesses only the part of the solution array held in the L3 cache. However, this is really very difficult to do, and you often end up losing in other areas.

Please understand, I am not a hardware expert, but I have a lot of experience in parallel processor coding on supercomputers. I run codes on Sequoia which is ranked the 3rd fastest super computer in the world, and Zin which is ranked 34th. Sequoia has 1,572,864 threads from 16 thread 1.6Ghz processors while Zin has 46,208 threads from 8 thread Xeon E5 2.6Ghz processors. A given simulation using 16 threads completes slightly quicker on Sequoia than when using 8 threads on Zin, despite the far faster Xeon processors on Zin. The clock speeds of the AMD FX8350 and Intel i5 are much closer, so it's pretty clear the AMD would outperform the i5 when running a given simulation (albeit on a relatively tiny scale) on a domestic PC by taking advantage of its 4 extra threads.

Give me a little time to get my Linux system set up on my new AMD (having trouble with UEFI secure boot!) and I will run some simulations and post numbers. There's a handy little subroutine in FORTRAN called CPU_TIME for benchmarking exercises.
m
0
l
!