Jump to content


 


Register a free account to unlock additional features at BleepingComputer.com
Welcome to BleepingComputer, a free community where people like yourself come together to discuss and learn how to use their computers. Using the site is easy and fun. As a guest, you can browse and view the various discussions in the forums, but can not create a new topic or reply to an existing one unless you are logged in. Other benefits of registering an account are subscribing to topics and forums, creating a blog, and having no ads shown anywhere on the site.


Click here to Register a free account now! or read our Welcome Guide to learn how to use this site.

Photo

AMD vs Intel: MDD1963 vs Drillingmachine


  • Please log in to reply
114 replies to this topic

#31 MDD1963

MDD1963

  • Members
  • 663 posts
  • OFFLINE
  •  
  • Local time:04:40 AM

Posted 03 September 2017 - 07:06 AM

 

"There might have been GPU bottleneck, still BF4 supports 8 cores and FX-8350 is keeping up with i7 CPU's:"

 

The fact that the FX8350 kept up absolutely confirms the GPU bottleneck. (I knew there had to be some reason you favored the game for comparing CPUs; it's total unsuitability for the comparisons was only mildly surprising.)

 

It's like testing CPUs at 4k Ultra....the R3 is not really the R7's equal, is it?

 

 

Considering we have integer heavy software with "equal" optimizations for AMD and Intel and software uses 8 cores/threads. Then FX-8350 is faster than i7-4770K (both at stock speeds). So basically if any i3 is faster than FX-8350, then software does not use 8 cores/threads properly, speed is depending on single thread or software is Intel optimized or non-AMD optimized.

 

http://gamegpu.com/action-/-fps-/-tps/battlefield-4-test-gpu.html

 

bf4_proz_2.jpg

 

As you can see, dual cores have no chance against quad cores. This is why:

 

 

 

The graph you posted does not seem to show the FX8350 faster than the 4770K, as the latter appears to be 18.5% faster in the graph shown....

 

 

Given the low placement of the 2600K in the results shown, which is hyperthreaded, it'd be difficult to make any conclusions about any relative 'strong core potential' or utility of BF4 in core scaling based solely on semi-respectable placement of the assorted FX processors, given their near total decimation in pretty much every other game in existence. (If we can find some graphs showing R5 and R7 orr even the 5820K, 6900K, etc, scaling upward with cores in BF4, lets check them out...with GOOD GPUs.

 

The rest of  the graphs showing CPU utilization I have little use for, other than confirmation of various cores receiving semi-even tasking thru Windows Scheduler, as I've seen some processor fans declare some sort of semi-mutated semi-victories over moderate processor utilization while still taking a stomping on all measures of delivered framerates. (as though 45% utilization and 60 fps is somehow 'better' than 95% utilization and 90 fps...)


Edited by MDD1963, 03 September 2017 - 07:28 AM.

Asus Z270A Prime/7700K/32 GB DDR4-3200/GTX1060


BC AdBot (Login to Remove)

 


m

#32 Drillingmachine

Drillingmachine
  • Topic Starter

  • Members
  • 2,090 posts
  • ONLINE
  •  
  • Gender:Male
  • Local time:09:40 PM

Posted 03 September 2017 - 08:50 AM

The graph you posted does not seem to show the FX8350 faster than the 4770K, as the latter appears to be 18.5% faster in the graph shown....

FX-8350 CPU usage is nowhere near 100%.

Considering we have integer heavy software with "equal" optimizations for AMD and Intel and software uses 8 cores/threads. Then FX-8350 is faster than i7-4770K (both at stock speeds).

That software does not use 8 threads fully.
 

Given the low placement of the 2600K in the results shown, which is hyperthreaded, it'd be difficult to make any conclusions about any relative 'strong core potential' or utility of BF4 in core scaling based solely on semi-respectable placement of the assorted FX processors, given their near total decimation in pretty much every other game in existence. (If we can find some graphs showing R5 and R7 orr even the 5820K, 6900K, etc, scaling upward with cores in BF4, lets check them out...with GOOD GPUs.

Comparing almost equal CPU's 2500K and 2600K, with only major difference being hyper threading, that game makes at least some use of more than 4 threads. FX-8350 does good in that game as it uses more than 4 threads and that's also main reason why all dual cores simply suck. i3-2100 cannot even beat FX-4100...
 
[quote name="MDD1963" post="4326180" timestamp="1504440385"The rest of  the graphs showing CPU utilization I have little use for, other than confirmation of various cores receiving semi-even tasking thru Windows Scheduler, as I've seen some processor fans declare some sort of semi-mutated semi-victories over moderate processor utilization while still taking a stomping on all measures of delivered framerates. (as though 45% utilization and 60 fps is somehow 'better' than 95% utilization and 90 fps...)Simply counting CPU utilizations to achieve "total CPU utilization" is not good way but it tells something about how processor cores are used:

Phenom II x2 550: 97%

FX-4300: 90%

FX-6300: 73.5%

FX-8350: 64.75%

Both FX-6300 and FX-8350 have almost same clock speeds so based on those, FX-6300 has 4.41 cores fully utilized and FX-8350 has 5.18 cores fully utilized. That's 15% difference. FPS difference is 12%.Those calculations show that BF4 is able to take advantage from more than six cores and while theoretical scaling from 6 to 8 cores is 25%, in game that translates to 12%.

CPU utilization figures also prove why dual cores are so slow, they simply haven't got enough processing power and better GPU would not change that. Comparing to Battlefield 1 where dual cores dominate 8 core CPU's, I stand my opinion about Battlefield 1 bad multi core use.

#33 MDD1963

MDD1963

  • Members
  • 663 posts
  • OFFLINE
  •  
  • Local time:04:40 AM

Posted 03 September 2017 - 07:58 PM

 

The graph you posted does not seem to show the FX8350 faster than the 4770K, as the latter appears to be 18.5% faster in the graph shown....

FX-8350 CPU usage is nowhere near 100%.

 

:)

 

Didn't think it would occur quite so quickly, I was thinking you'd perhaps typo'd in your claim the FX8350 was 'faster' than the 4770K, but...

 

Perhaps this is one ever-so-prime example of the aforementioned scenarios I spoke of, where, some try to claim a given processor is 'faster' or 'better' solely because of lower core usage, with such ever so slight details such as your 'faster processor' actually delivering a not-so-trivial 18.5% less actual measured delivered gaming performance (just a barely spoken of 'side-note', almost a benefit of the improved performance, in fact). Results do matter, whether it be sustained throughput, GB/sec, delivered frames per second (average, .1/1% lows), etc...

 

If you've convinced yourself low processor utilization matters more than delivered performance, please open a business around selling economical NAS 4 bay servers with 5400 rpm drive-equipped (to save power and cost) servers equipped with 12/16 core Threadrippers equipped only with 100 MB/sec internal NICs to offset cost. Good luck explaining the 10 MB/ sec transfers to your customers.

 

If you were doing large network transfers via a 10GBe connection, would you be happy one of your 16 core Threadripper servers, despite a teamed/bonded 10Gbe connection with only 300Mb/sec performance from one server as long as it's core usage was a lowly 20% across all cores? Would it still be considered faster than an R3-1200 delivering sustained 1 Gigabyte/sec throughput tin delivered network performance, ....to you?

 

Based on this line of dubious/questionable line of thinking, a 16 core 1950X Threadripper is indeed a pretty good choice of CPU for a competitive gaming system for 144 Hz monitors, as although it might deliver slightly less (20%  or so) framerates than a $1000-less R5-1600 based system,  at least it's core usage is only 10-12% or so....


Edited by MDD1963, 03 September 2017 - 08:05 PM.

Asus Z270A Prime/7700K/32 GB DDR4-3200/GTX1060


#34 MDD1963

MDD1963

  • Members
  • 663 posts
  • OFFLINE
  •  
  • Local time:04:40 AM

Posted 03 September 2017 - 10:27 PM

I would not say the i3 'dominates' the FX8350 in BF1 (although 22% is not a tiny lead, either), I'd just say the generational differences/improvements in IPC and clock speed from your chosen 5 year old example of the unchanged FX8350 leading an old i3-2100 in one game example no longer directly applies when establishing a processor 'hierarchy of performance'.... save for in in that particular game with those outdated processors using some possibly GPU limited scenario.....

 

It's not an insult to AMD that one of it's older processors is outperformed by a higher clock speed and/or much later generation/IPC-improved i3, as the same situation exists with the modern i3-7350K often now outperforming the venerable i7-2600K. Often, better IPC/throughput might prevail even at lower clock speeds. Same reason an old Athlon 64/3500+ (at an 2200 MHz) outperformed the Prescott P4/3000 MHz (400 MHz FSB). We very well might see newer i5 processors (6c/6t) outperforming 7700K in some games, depending on the former's clock speed...; it could happen. The result will tell the tale in each particular game, as we certainly don't want to reach false conclusions from one set of what could be anomalous/peculiar results at best.

 

I will see if I can talk Hardware Unboxed into throwing in some BF4 1080P results with a good GPU (1080Ti) using modern R3/1200, R3-1500, and R5-1600, and R7-1700 just so we can test the theories of core vs. thread scaling...(I see little utility in dragging out obsolete 6-7 year old processors of either brand for this exercise, or attempting to draw scaling conclusions of assorted core/thread counts and usefulness of BF4 as 'core friendly' with hardware now best served as a younger nephew's/ Grandmother's e-mail, Youtube, and coupon printing machine...


Asus Z270A Prime/7700K/32 GB DDR4-3200/GTX1060


#35 Drillingmachine

Drillingmachine
  • Topic Starter

  • Members
  • 2,090 posts
  • ONLINE
  •  
  • Gender:Male
  • Local time:09:40 PM

Posted 04 September 2017 - 05:12 AM

Didn't think it would occur quite so quickly, I was thinking you'd perhaps typo'd in your claim the FX8350 was 'faster' than the 4770K, but...
 
Perhaps this is one ever-so-prime example of the aforementioned scenarios I spoke of, where, some try to claim a given processor is 'faster' or 'better' solely because of lower core usage, with such ever so slight details such as your 'faster processor' actually delivering a not-so-trivial 18.5% less actual measured delivered gaming performance (just a barely spoken of 'side-note', almost a benefit of the improved performance, in fact). Results do matter, whether it be sustained throughput, GB/sec, delivered frames per second (average, .1/1% lows), etc...
 
If you've convinced yourself low processor utilization matters more than delivered performance, please open a business around selling economical NAS 4 bay servers with 5400 rpm drive-equipped (to save power and cost) servers equipped with 12/16 core Threadrippers equipped only with 100 MB/sec internal NICs to offset cost. Good luck explaining the 10 MB/ sec transfers to your customers.
 
If you were doing large network transfers via a 10GBe connection, would you be happy one of your 16 core Threadripper servers, despite a teamed/bonded 10Gbe connection with only 300Mb/sec performance from one server as long as it's core usage was a lowly 20% across all cores? Would it still be considered faster than an R3-1200 delivering sustained 1 Gigabyte/sec throughput tin delivered network performance, ....to you?
 
Based on this line of dubious/questionable line of thinking, a 16 core 1950X Threadripper is indeed a pretty good choice of CPU for a competitive gaming system for 144 Hz monitors, as although it might deliver slightly less (20%  or so) framerates than a $1000-less R5-1600 based system,  at least it's core usage is only 10-12% or so....


It's very logical. More cores = better if cores are fully or at least partially utilized. In BF4 around 2/3 of FX-8350 potential power is used excluding CPU specific optimizations. There are many benchmarks where FX-8350 is faster than Intel quad core i7 and that essentially proves FX has more throughput when all cores are used.

About your Threadripper example. Threadripper is meant for scenarios where many cores are used at same time. If CPU usage on Threadripper is around 20%, then it's clearly wrong CPU for that scenario.

I would not say the i3 'dominates' the FX8350 in BF1 (although 22% is not a tiny lead, either), I'd just say the generational differences/improvements in IPC and clock speed from your chosen 5 year old example of the unchanged FX8350 leading an old i3-2100 in one game example no longer directly applies when establishing a processor 'hierarchy of performance'.... save for in in that particular game with those outdated processors using some possibly GPU limited scenario.....
 
It's not an insult to AMD that one of it's older processors is outperformed by a higher clock speed and/or much later generation/IPC-improved i3, as the same situation exists with the modern i3-7350K often now outperforming the venerable i7-2600K. Often, better IPC/throughput might prevail even at lower clock speeds. Same reason an old Athlon 64/3500+ (at an 2200 MHz) outperformed the Prescott P4/3000 MHz (400 MHz FSB). We very well might see newer i5 processors (6c/6t) outperforming 7700K in some games, depending on the former's clock speed...; it could happen. The result will tell the tale in each particular game, as we certainly don't want to reach false conclusions from one set of what could be anomalous/peculiar results at best.
 
I will see if I can talk Hardware Unboxed into throwing in some BF4 1080P results with a good GPU (1080Ti) using modern R3/1200, R3-1500, and R5-1600, and R7-1700 just so we can test the theories of core vs. thread scaling...(I see little utility in dragging out obsolete 6-7 year old processors of either brand for this exercise, or attempting to draw scaling conclusions of assorted core/thread counts and usefulness of BF4 as 'core friendly' with hardware now best served as a younger nephew's/ Grandmother's e-mail, Youtube, and coupon printing machine...


For modern i3 being 22% faster than FX-8350, it would mean that it should receive minimum frame rate of 91.5 on that test meaning modern i3 is whopping 2.34 times faster than i3-2100. Easy to say, it's not that much faster. Point is still that BF4 makes much better use of more cores than BF1.

IPC and clock speeds have not improved nearly enough so that i3-7350K have any chance against i7-2600K. That
can happen in cases i3 has support for instruction sets i7-2600K does no support or core usage is limited. i5
6/6 with right clock speeds can easily beat i7 4/8, no doubt about that one but that's not really fair comparison abyway.

That test would be fine. Also adding BF1 results into mix. I'm pretty sure BF1 would suck badly in that comparison, as current evidence show without any doubt that BF4 uses more cores much better.

#36 MDD1963

MDD1963

  • Members
  • 663 posts
  • OFFLINE
  •  
  • Local time:04:40 AM

Posted 04 September 2017 - 07:27 PM

 

Didn't think it would occur quite so quickly, I was thinking you'd perhaps typo'd in your claim the FX8350 was 'faster' than the 4770K, but...
 
Perhaps this is one ever-so-prime example of the aforementioned scenarios I spoke of, where, some try to claim a given processor is 'faster' or 'better' solely because of lower core usage, with such ever so slight details such as your 'faster processor' actually delivering a not-so-trivial 18.5% less actual measured delivered gaming performance (just a barely spoken of 'side-note', almost a benefit of the improved performance, in fact). Results do matter, whether it be sustained throughput, GB/sec, delivered frames per second (average, .1/1% lows), etc...
 
If you've convinced yourself low processor utilization matters more than delivered performance, please open a business around selling economical NAS 4 bay servers with 5400 rpm drive-equipped (to save power and cost) servers equipped with 12/16 core Threadrippers equipped only with 100 MB/sec internal NICs to offset cost. Good luck explaining the 10 MB/ sec transfers to your customers.
 
If you were doing large network transfers via a 10GBe connection, would you be happy one of your 16 core Threadripper servers, despite a teamed/bonded 10Gbe connection with only 300Mb/sec performance from one server as long as it's core usage was a lowly 20% across all cores? Would it still be considered faster than an R3-1200 delivering sustained 1 Gigabyte/sec throughput tin delivered network performance, ....to you?
 
Based on this line of dubious/questionable line of thinking, a 16 core 1950X Threadripper is indeed a pretty good choice of CPU for a competitive gaming system for 144 Hz monitors, as although it might deliver slightly less (20%  or so) framerates than a $1000-less R5-1600 based system,  at least it's core usage is only 10-12% or so....


It's very logical. More cores = better if cores are fully or at least partially utilized. In BF4 around 2/3 of FX-8350 potential power is used excluding CPU specific optimizations. There are many benchmarks where FX-8350 is faster than Intel quad core i7 and that essentially proves FX has more throughput when all cores are used.

About your Threadripper example. Threadripper is meant for scenarios where many cores are used at same time. If CPU usage on Threadripper is around 20%, then it's clearly wrong CPU for that scenario.

 

 

Not sure what quad core(s) you compared FX to (presumably the very oldest/lowest clocked of Core2Quads?), but, the FX8350 has been pretty much been dominated /decimated in almost every single game by nearly every Intel processor since the FX8350 delivered. (occasionally, the FX would catch up with an i7 from 5-6 years earlier, if these are the i7 victories you speak of)

 

If the FX8350's 'victory' resides on having a lower cpu utilization, you are welcome to those 'victories', but, alas, others might expect an actual victory to be measured in something more tangible, such as average framerates, minimum framerates, etc...i.e., actual performance.

 

If you are talking about FX cpus perhaps defeating i7-2600 and up in Cinebench, Handbrake, and things that might be able to utilize the cores, that is entirely possible. Six and 8 core cpus can defeat quad cores of course....

 

But when a processor is significantly slower in 98% of all games in existence at 1080P at even 1 GHz higher clock speed 'advantage', it's rather difficult to argue that it is 'better' or 'faster'. If folks bought FX CPUs for movie editing, Cinebench and solely encoding/decoding, the lower per core performance might have been a reasonable tradeoff to many with high 975X/980X prices...

 

And of course more cores are better when and if the cores can be and are utilized...; if not utilized, they are useless at best, and might end up an actual hindrance to performance occasionally, due to added inter-core/cache latency, etc..

 

If you want to discuss multi-threaded performance in measurable tasks/ benchmarks, please at least show the perhaps 1 out of 20 where an FX 6 or 8 core defeats an Intel 6 or 8 core....

 

http://www.tomshardware.com/reviews/core-i7-4960x-ivy-bridge-e-benchmark,3557-8.html

 

Here is an MPEG2-H.264 conversion comparison, where the FX8350 lags the 4c/8t 2700K only slightly, but where the 4960X completes the task in 75% of the time required...


Asus Z270A Prime/7700K/32 GB DDR4-3200/GTX1060


#37 MDD1963

MDD1963

  • Members
  • 663 posts
  • OFFLINE
  •  
  • Local time:04:40 AM

Posted 04 September 2017 - 07:37 PM

 

 


For modern i3 being 22% faster than FX-8350, it would mean that it should receive minimum frame rate of 91.5 on that test meaning modern i3 is whopping 2.34 times faster than i3-2100. Easy to say, it's not that much faster. Point is still that BF4 makes much better use of more cores than BF1.

IPC and clock speeds have not improved nearly enough so that i3-7350K have any chance against i7-2600K. That
can happen in cases i3 has support for instruction sets i7-2600K does no support or core usage is limited. i5
6/6 with right clock speeds can easily beat i7 4/8, no doubt about that one but that's not really fair comparison abyway.

That test would be fine. Also adding BF1 results into mix. I'm pretty sure BF1 would suck badly in that comparison, as current evidence show without any doubt that BF4 uses more cores much better.

 

Except for the games that it almost equals, equals, or defeats the 2600K in...

 

To wit...

http://www.anandtech.com/show/11083/the-intel-core-i3-7350k-60w-review/7


Edited by MDD1963, 04 September 2017 - 07:45 PM.

Asus Z270A Prime/7700K/32 GB DDR4-3200/GTX1060


#38 MDD1963

MDD1963

  • Members
  • 663 posts
  • OFFLINE
  •  
  • Local time:04:40 AM

Posted 04 September 2017 - 08:53 PM

How much better could the BF4 really be in your proclaimed 'better use of 8 cores' scenario when the FX is still outperformed by Intel quad cores? (If you have found one possibly GPU- bound game to partially minimize the FX's IPC weaknesses and proclaim it to be the gaming benchmark of all time, have at it...) Most of the BF4 comparisons seemed heavily GPU bound even with the GTX980 at the time. It's be nice to see some more modern cpu comparisons equipped with a 1080Ti comparisons in BF4.

 

At least BF1 scales upward nicely jumping from 4 to 6 cores, or, increasing clockspeed, or both. In fact, even the 10 core highly clocked systems do quite nicely as well...even the 1920X/1950X do pretty well in it, although lagging the 1800X slightly.

 

Battlefield 4 CPU scaling...

 

 

Fast forward to 11 seconds to see all CPUs: once you rule our the too-slow AMDs at the bottom, most of the pack is bunched up with a narrow spread. Look just a tad GPU-limited to you?


Edited by MDD1963, 04 September 2017 - 08:58 PM.

Asus Z270A Prime/7700K/32 GB DDR4-3200/GTX1060


#39 Drillingmachine

Drillingmachine
  • Topic Starter

  • Members
  • 2,090 posts
  • ONLINE
  •  
  • Gender:Male
  • Local time:09:40 PM

Posted 05 September 2017 - 08:38 AM

Not sure what quad core(s) you compared FX to (presumably the very oldest/lowest clocked of Core2Quads?), but, the FX8350 has been pretty much been dominated /decimated in almost every single game by nearly every Intel processor since the FX8350 delivered. (occasionally, the FX would catch up with an i7 from 5-6 years earlier, if these are the i7 victories you speak of)
 
If the FX8350's 'victory' resides on having a lower cpu utilization, you are welcome to those 'victories', but, alas, others might expect an actual victory to be measured in something more tangible, such as average framerates, minimum framerates, etc...i.e., actual performance.
 
If you are talking about FX cpus perhaps defeating i7-2600 and up in Cinebench, Handbrake, and things that might be able to utilize the cores, that is entirely possible. Six and 8 core cpus can defeat quad cores of course....
 
But when a processor is significantly slower in 98% of all games in existence at 1080P at even 1 GHz higher clock speed 'advantage', it's rather difficult to argue that it is 'better' or 'faster'. If folks bought FX CPUs for movie editing, Cinebench and solely encoding/decoding, the lower per core performance might have been a reasonable tradeoff to many with high 975X/980X prices...


There are very few games that actually use 8 cores so no wonder there are few victories too in games. FX-8350 was excellent CPU for certain cases, like virtualization. It had 8 physical cores and it was cheap. So it's not that bad some people say, they just didn't realize that was not made for SuperPi like obsolete software or single threaded games.

We also know now that AMD never truly fixed FX-series. It had some design flaws AMD never bothered to fix as they concentrated on Zen. It would have been nice to see fixed Bulldozer but now it seems we never see it.
 

And of course more cores are better when and if the cores can be and are utilized...; if not utilized, they are useless at best, and might end up an actual hindrance to performance occasionally, due to added inter-core/cache latency, etc..
 
If you want to discuss multi-threaded performance in measurable tasks/ benchmarks, please at least show the perhaps 1 out of 20 where an FX 6 or 8 core defeats an Intel 6 or 8 core....
 
http://www.tomshardware.com/reviews/core-i7-4960x-ivy-bridge-e-benchmark,3557-8.html
 
Here is an MPEG2-H.264 conversion comparison, where the FX8350 lags the 4c/8t 2700K only slightly, but where the 4960X completes the task in 75% of the time required...


Better to look for Linux benchmarks so that software is ran through compiler with equally good settings, not "if VendorID=Authentic AMD, then lower speed 50%" kind of stuff.

Those tests make quite bad use for many cores. LAME is known top have very poor multi-core support for that kind of software.

#40 Drillingmachine

Drillingmachine
  • Topic Starter

  • Members
  • 2,090 posts
  • ONLINE
  •  
  • Gender:Male
  • Local time:09:40 PM

Posted 05 September 2017 - 08:45 AM


For modern i3 being 22% faster than FX-8350, it would mean that it should receive minimum frame rate of 91.5 on that test meaning modern i3 is whopping 2.34 times faster than i3-2100. Easy to say, it's not that much faster. Point is still that BF4 makes much better use of more cores than BF1.

IPC and clock speeds have not improved nearly enough so that i3-7350K have any chance against i7-2600K. That
can happen in cases i3 has support for instruction sets i7-2600K does no support or core usage is limited.
i5
6/6 with right clock speeds can easily beat i7 4/8, no doubt about that one but that's not really fair comparison abyway.

That test would be fine. Also adding BF1 results into mix. I'm pretty sure BF1 would suck badly in that comparison, as current evidence show without any doubt that BF4 uses more cores much better.

Except for the games that it almost equals, equals, or defeats the 2600K in...
 
To wit...
http://www.anandtech.com/show/11083/the-intel-core-i3-7350k-60w-review/7


You bolded wrong part.

How much better could the BF4 really be in your proclaimed 'better use of 8 cores' scenario when the FX is still outperformed by Intel quad cores? (If you have found one possibly GPU- bound game to partially minimize the FX's IPC weaknesses and proclaim it to be the gaming benchmark of all time, have at it...) Most of the BF4 comparisons seemed heavily GPU bound even with the GTX980 at the time. It's be nice to see some more modern cpu comparisons equipped with a 1080Ti comparisons in BF4.
 
At least BF1 scales upward nicely jumping from 4 to 6 cores, or, increasing clockspeed, or both. In fact, even the 10 core highly clocked systems do quite nicely as well...even the 1920X/1950X do pretty well in it, although lagging the 1800X slightly.


I said better use, as I previously calculated, FX 8-core usage was around 60%. That is much better than, say, 25%.

BF1 does not scale well because even dual cores can easily beat 8-cores while on BF4 dual cores have no chance. Also BF1 DirectX 12 implementation simply sucks.
 

Battlefield 4 CPU scaling...
 

 
Fast forward to 11 seconds to see all CPUs: once you rule our the too-slow AMDs at the bottom, most of the pack is bunched up with a narrow spread. Look just a tad GPU-limited to you?


That is single player scenario from BF4, it really doesn't tell anything about CPU scaling in multiplayer scenario. So those results have no value.

#41 MDD1963

MDD1963

  • Members
  • 663 posts
  • OFFLINE
  •  
  • Local time:04:40 AM

Posted 05 September 2017 - 09:25 PM

 

 

 

 


For modern i3 being 22% faster than FX-8350, it would mean that it should receive minimum frame rate of 91.5 on that test meaning modern i3 is whopping 2.34 times faster than i3-2100. Easy to say, it's not that much faster. Point is still that BF4 makes much better use of more cores than BF1.

IPC and clock speeds have not improved nearly enough so that i3-7350K have any chance against i7-2600K. That
can happen in cases i3 has support for instruction sets i7-2600K does no support or core usage is limited.
i5
6/6 with right clock speeds can easily beat i7 4/8, no doubt about that one but that's not really fair comparison abyway.

That test would be fine. Also adding BF1 results into mix. I'm pretty sure BF1 would suck badly in that comparison, as current evidence show without any doubt that BF4 uses more cores much better.

 

Except for the games that it almost equals, equals, or defeats the 2600K in...
 
To wit...
http://www.anandtech.com/show/11083/the-intel-core-i3-7350k-60w-review/7

 


You bolded wrong part.

How much better could the BF4 really be in your proclaimed 'better use of 8 cores' scenario when the FX is still outperformed by Intel quad cores? (If you have found one possibly GPU- bound game to partially minimize the FX's IPC weaknesses and proclaim it to be the gaming benchmark of all time, have at it...) Most of the BF4 comparisons seemed heavily GPU bound even with the GTX980 at the time. It's be nice to see some more modern cpu comparisons equipped with a 1080Ti comparisons in BF4.
 
At least BF1 scales upward nicely jumping from 4 to 6 cores, or, increasing clockspeed, or both. In fact, even the 10 core highly clocked systems do quite nicely as well...even the 1920X/1950X do pretty well in it, although lagging the 1800X slightly.


I said better use, as I previously calculated, FX 8-core usage was around 60%. That is much better than, say, 25%.

BF1 does not scale well because even dual cores can easily beat 8-cores while on BF4 dual cores have no chance. Also BF1 DirectX 12 implementation simply sucks.
 

Battlefield 4 CPU scaling...
 

 
Fast forward to 11 seconds to see all CPUs: once you rule our the too-slow AMDs at the bottom, most of the pack is bunched up with a narrow spread. Look just a tad GPU-limited to you?


That is single player scenario from BF4, it really doesn't tell anything about CPU scaling in multiplayer scenario. So those results have no value.

 

"I said better use, as I previously calculated, FX 8-core usage was around 60%. That is much better than, say, 25%"

 

So now you actually like high percentage of core utilization, I see? :)

 

You did not seem to address the severe symptoms of GPU-limited performance in the above new core-friendly BF4 1080P benchmark as seen in the above video at 11 seconds onward...

 

Yes, the game scaled well with more FX cores up to at least or even at a maximum of 8, that was/is known....; as AMD processors were not exactly optimal , hitting the GPU -bound point did not occur until much higher in both core count and clock speed. Before saturating the GPU at even 1080P res, they continue to scale with more cores and/or more clock speed somewhat at 3 GHz, 4 GHz, 4.5 GHz, up until hitting the 98% saturated/ GPU-bound point. 

 

If you look at the graph, you will see that all of the CPUs near the top of the graph are bunched up, all near the same framerates. Is that because all of the CPUs are suddenly equal in this one game, or, is it possibly more likely that they are now GPU-limited, do you think?

 

I'd say you concluded that BF4 is/was 'core friendly' only because the weaker AMD FX processors, far from being able to saturate the GPU at lower core counts and clockspeed, still continued to scale with more cores and/or more clock speed, until FINALLY nearing saturating the GPU with 8 core/threads at 4.5 GHz or so... That the game did not continue to scale with more Intel cores and faster clockspeeds should make it clear it was GPU-limited at that point.

 

I'm not here to bash or denigrate AMD's 7-10 year old FX processor designs; that their performance trailed i5's and i7's in gaming since 2006 onward was/is pretty well undisputed. They were reasonably priced multicore alternatives that filled many folks' multitasking needs in many applications during the era of $1000+ dollar 975X, etc....; and now the R7 does well with $1000 7900X...

 

Will the Ryzen2 scale to 5 GHz, and with better IPC than current Ryzen processors? I for one hope so. Let AMD/Intel both release their best so that one day 3 years from now, the R7-1800X and 7700K  are out-framed by 50-60% or so by newer CPUs...and at the $250 price point, and with 10-12 cores, or, at lest 6-8 cores at 5 GHz or so...


Edited by MDD1963, 05 September 2017 - 09:32 PM.

Asus Z270A Prime/7700K/32 GB DDR4-3200/GTX1060


#42 MDD1963

MDD1963

  • Members
  • 663 posts
  • OFFLINE
  •  
  • Local time:04:40 AM

Posted 05 September 2017 - 09:41 PM

 

The graph you posted does not seem to show the FX8350 faster than the 4770K, as the latter appears to be 18.5% faster in the graph shown....

FX-8350 CPU usage is nowhere near 100%.

Considering we have integer heavy software with "equal" optimizations for AMD and Intel and software uses 8 cores/threads. Then FX-8350 is faster than i7-4770K (both at stock speeds).

That software does not use 8 threads fully.
 

 

I apologize, but, I lost your point here, as the quote /statement you seem to have been addressing or perhaps rebutting.....was yours. Are you now debating the accuracy of your own quotes/claims? :)

 

Let's give the stunning complete lack of processor scaling in BF4 above the GPU saturation point a brief reprieve.

 

When is Ryzen 2 arriving? And what will it's performance be...and at what clock speed for the flagship? 


Asus Z270A Prime/7700K/32 GB DDR4-3200/GTX1060


#43 Drillingmachine

Drillingmachine
  • Topic Starter

  • Members
  • 2,090 posts
  • ONLINE
  •  
  • Gender:Male
  • Local time:09:40 PM

Posted 06 September 2017 - 11:52 AM

For modern i3 being 22% faster than FX-8350, it would mean that it should receive minimum frame rate of 91.5 on that test meaning modern i3 is whopping 2.34 times faster than i3-2100. Easy to say, it's not that much faster. Point is still that BF4 makes much better use of more cores than BF1.

IPC and clock speeds have not improved nearly enough so that i3-7350K have any chance against i7-2600K. That
can happen in cases i3 has support for instruction sets i7-2600K does no support or core usage is limited.
i5
6/6 with right clock speeds can easily beat i7 4/8, no doubt about that one but that's not really fair comparison abyway.

That test would be fine. Also adding BF1 results into mix. I'm pretty sure BF1 would suck badly in that comparison, as current evidence show without any doubt that BF4 uses more cores much better.

Except for the games that it almost equals, equals, or defeats the 2600K in...

To wit...
http://www.anandtech.com/show/11083/the-intel-core-i3-7350k-60w-review/7

You bolded wrong part.

How much better could the BF4 really be in your proclaimed 'better use of 8 cores' scenario when the FX is still outperformed by Intel quad cores? (If you have found one possibly GPU- bound game to partially minimize the FX's IPC weaknesses and proclaim it to be the gaming benchmark of all time, have at it...) Most of the BF4 comparisons seemed heavily GPU bound even with the GTX980 at the time. It's be nice to see some more modern cpu comparisons equipped with a 1080Ti comparisons in BF4.

At least BF1 scales upward nicely jumping from 4 to 6 cores, or, increasing clockspeed, or both. In fact, even the 10 core highly clocked systems do quite nicely as well...even the 1920X/1950X do pretty well in it, although lagging the 1800X slightly.

I said better use, as I previously calculated, FX 8-core usage was around 60%. That is much better than, say, 25%.

BF1 does not scale well because even dual cores can easily beat 8-cores while on BF4 dual cores have no chance. Also BF1 DirectX 12 implementation simply sucks.

Battlefield 4 CPU scaling...



Fast forward to 11 seconds to see all CPUs: once you rule our the too-slow AMDs at the bottom, most of the pack is bunched up with a narrow spread. Look just a tad GPU-limited to you?

That is single player scenario from BF4, it really doesn't tell anything about CPU scaling in multiplayer scenario. So those results have no value.

"I said better use, as I previously calculated, FX 8-core usage was around 60%. That is much better than, say, 25%"

So now you actually like high percentage of core utilization, I see? :)

You did not seem to address the severe symptoms of GPU-limited performance in the above new core-friendly BF4 1080P benchmark as seen in the above video at 11 seconds onward...

Yes, the game scaled well with more FX cores up to at least or even at a maximum of 8, that was/is known....; as AMD processors were not exactly optimal , hitting the GPU -bound point did not occur until much higher in both core count and clock speed. Before saturating the GPU at even 1080P res, they continue to scale with more cores and/or more clock speed somewhat at 3 GHz, 4 GHz, 4.5 GHz, up until hitting the 98% saturated/ GPU-bound point.

If you look at the graph, you will see that all of the CPUs near the top of the graph are bunched up, all near the same framerates. Is that because all of the CPUs are suddenly equal in this one game, or, is it possibly more likely that they are now GPU-limited, do you think?

I'd say you concluded that BF4 is/was 'core friendly' only because the weaker AMD FX processors, far from being able to saturate the GPU at lower core counts and clockspeed, still continued to scale with more cores and/or more clock speed, until FINALLY nearing saturating the GPU with 8 core/threads at 4.5 GHz or so... That the game did not continue to scale with more Intel cores and faster clockspeeds should make it clear it was GPU-limited at that point.

I'm not here to bash or denigrate AMD's 7-10 year old FX processor designs; that their performance trailed i5's and i7's in gaming since 2006 onward was/is pretty well undisputed. They were reasonably priced multicore alternatives that filled many folks' multitasking needs in many applications during the era of $1000+ dollar 975X, etc....; and now the R7 does well with $1000 7900X...

Will the Ryzen2 scale to 5 GHz, and with better IPC than current Ryzen processors? I for one hope so. Let AMD/Intel both release their best so that one day 3 years from now, the R7-1800X and 7700K are out-framed by 50-60% or so by newer CPUs...and at the $250 price point, and with 10-12 cores, or, at lest 6-8 cores at 5 GHz or so...

Of course I like high CPU utilization.

That video was from single player BF4, so results are different than on multiplayer. No wonder single player is much less demanding than multiplayer. I was commenting about multiplayer results. And so putting "single player benchmarks say this and that" just won't work.

There were no i5 or i7 at 2006.

Ryzen 2 will have better IPC. Clock speeds depends on GlobalFoundries process, their track record is not very good.

Edited by Drillingmachine, 06 September 2017 - 11:56 AM.


#44 Drillingmachine

Drillingmachine
  • Topic Starter

  • Members
  • 2,090 posts
  • ONLINE
  •  
  • Gender:Male
  • Local time:09:40 PM

Posted 06 September 2017 - 11:56 AM

The graph you posted does not seem to show the FX8350 faster than the 4770K, as the latter appears to be 18.5% faster in the graph shown....

FX-8350 CPU usage is nowhere near 100%.

Considering we have integer heavy software with "equal" optimizations for AMD and Intel and software uses 8 cores/threads. Then FX-8350 is faster than i7-4770K (both at stock speeds).

That software does not use 8 threads fully.
 

I apologize, but, I lost your point here, as the quote /statement you seem to have been addressing or perhaps rebutting.....was yours. Are you now debating the accuracy of your own quotes/claims? :)
 
Let's give the stunning complete lack of processor scaling in BF4 above the GPU saturation point a brief reprieve.
 
When is Ryzen 2 arriving? And what will it's performance be...and at what clock speed for the flagship?



There's nothing wrong what I said. As I clearly stated software must use 8 threads, be integer heavy and have "equal" optimizations. BF4 does not meet at least first criteria.

And once again, you are pointing to single player BF4 benchmarks where CPU utilization is much lower than on multiplayer.

Ryzen 2 is expected to arrive Q4/2018 or Q1/2019. Clock speeds are all about process and how it works.

#45 MDD1963

MDD1963

  • Members
  • 663 posts
  • OFFLINE
  •  
  • Local time:04:40 AM

Posted 06 September 2017 - 09:21 PM

Are we honestly still (attempting) convincing someone, anyone, that a measured 18.5% deficit in average delivered frames per second or a 10-15% deficit in minimum frames per second from the FX processor in most games is/was... 'faster'?

 

Good luck in this what must prove to be an undoubtedly quite challenging endeavor. :)

 

(That same approximate gap exists today in many games, only now instead of FX8350 and i7-3770, we are at Ryzen 5-1600/R7-1700 and 7700K...)

 

 

stock clocked 7700K has 35% higher .1% minimum low framerate than the R7-1700 in Overwatch at 1080P, a 67% lead in minimum framerate in DOTA2 at 1080P, a 19% lead at Doom 1440P minimum framerates, 65% higher in Rocket League minimum framerates, and a 35% lead in miinimum framerates in BF1... I did not choose the games here, as I personally only play BF1. Disclaimer: would the measured differences here matter on a 60 Hz monitor? No. Might they matter on a 144 Hz 1080P monitor? You bet.

 

The above percentages are calculated at stock clock speeds, using .1% lows/minimums; as OC'd results are shown, folks can calculate them as they see fit as well, but, as both processor can be overclocked, I doubt the spread will narrow much, if at all.

 

You can examine core usage,dividing it's total cpu utilization by numbers of cores,etc., all you want, but, *in gaming without streaming*, Ryzen is 'good' but, not as fast as the 7700K. Does the R7-1700 when OC'd pull ahead in gaming while streaming? Yes. is that something that folks who want to stream while gaming should consider? Absolutely.

 

Different use cases might or perhaps *should* result in different systems chosen/recommended...

 

(As a side note, I did notice that Ryzen on BF1 did not scale above 6 cores/12 threads, giving basically no improvement with more cores/threads of the R7 and up beyond the performance achieved with the R5-1600 at 4.0 GHz. No shock there, as more cores are not yet the universal answer in all workloads, as we already knew.)


Edited by MDD1963, 06 September 2017 - 09:22 PM.

Asus Z270A Prime/7700K/32 GB DDR4-3200/GTX1060





0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users