Do AMD cpus at least give a smoother desktop experience w/more cores?

[MiddleOfTheRoad]

This thread was originally about smoother desktop experience...

I bounce back and forth between AMD and Intel machines. I can't notice a difference at all. Biggest bottleneck of desktop experience is usually the hard drive and having an SSD is the biggest factor.

On the contrary to the previous sardonic poster about lowering Intel prices... I rather like that people buy up Intel chips and think so lowly of AMD. Nabbed an FX-6300 Hex core + Motherboard for $130 at MicroCenter. Battlefield 3 is smooth as hell. Comparable Intel hardware starts at $200+.

And you all could use some economics classes. They are *both* keeping each other's prices down. That's how competition works.

I love Micro Center! I have never driven through Atlanta without stopping to pick up one of their CPU/Motherboard combos. I'll probably grab an A10-7700K with motherboard for my girlfriend on the next trip. Seriously -- Micro Center's Motherboard / CPU combo for the 7700K is 3 bucks less than what Newegg is selling the 7700k apu for by itself. I don't know how they do it.

And for the casual gamer -- let me see you beat the A6-7400K with Gigabyte motherboard for $68. Nobody touches Micro Center's deals -- I just wish they had more stores (or at least had one in Florida).

 

[frozentundra123456]

PClab openly admitted they look for the places where amd tanks and call it a day.

Must have been easy to find, huh?

 

[nerp]

You can take whatever at PClab the story will be the same, so ok for crysis but then :

Def is the same at 1080p ultra, and other scenes than the ones used by Hfr are not reproducible according to their reviewer.

And remember, anything Intel below a i7 wont provide those frame without stuttering if there s anything ese that the game activated, with a i3 you couldnt even download a youtube video using wifi while there s a tab open with some flash in the background, all things that are trivial for a FX6.

Nonsense. You're making stuff up and your hypothetical is nonsensical. I work with 100+ tabs open. Games run fast. I never close anything on my PC.

 

[Abwx]

Nonsense. You're making stuff up and your hypothetical is nonsensical. I work with 100+ tabs open. Games run fast. I never close anything on my PC.

If it was non sense then those numbers wouldnt exist :

http://www.computerbase.de/2015-10/...m-multitasking-test-the-witcher-3-plus-winrar

 

[Yuriman]

If it was non sense then those numbers wouldnt exist :

http://www.computerbase.de/2015-10/...m-multitasking-test-the-witcher-3-plus-winrar

There's an important piece of data missing in those tests. In Cinebench + WinRAR, they list the Cinebench score, but how much longer did it take to decompress their test WinRAR file? That's critical.

Analogy: If I wanted to compare your multitasking ability to mine, by having you stack cans with one hand while shuffling cards with the other, and it took you twice as long to stack cans vs with both of your hands free, while it only took me 10% longer, that completely ignores the fact that I might have been focusing far less on the cards than you were. Useless comparison without both pieces of information.

 

[Erenhardt]

There's an important piece of data missing in those tests. In Cinebench + WinRAR, they list the Cinebench score, but how much longer did it take to decompress their test WinRAR file? That's critical.

Analogy: If I wanted to compare your multitasking ability to mine, by having you stack cans with one hand while shuffling cards with the other, and it took you twice as long to stack cans vs with both of your hands free, while it only took me 10% longer, that completely ignores the fact that I might have been focusing far less on the cards than you were. Useless comparison without both pieces of information.

Go to second graph. It shows Winrar times.

 

[Sweepr]

If it was non sense then those numbers wouldnt exist :

http://www.computerbase.de/2015-10/...m-multitasking-test-the-witcher-3-plus-winrar

Lol, just noticed Skylake Core i5 is 15.6% slower at the isolated WinRAR test but only 10.5% slower when running WinRAR + The Wicther 3 compared to the FX8350. There goes your fallacy about multitasking.

And we already know what ComputerBase recommends, they chose Core i5 instead of the higher-clocked FXs in their latest guide (like any sane person would).

http://www.hardware.fr/getgraphimg.php?id=223&n=17

Def is the same at 1080p ultra, and other scenes than the ones used by Hfr are not reproducible according to their reviewer.

And remember, anything Intel below a i7 wont provide those frame without stuttering if there s anything ese that the game activated, with a i3 you couldnt even download a youtube video using wifi while there s a tab open with some flash in the background, all things that are trivial for a FX6.

So both Hardware.fr and PCLab put the latest Core i5 above all AMD CPUs in this game. Thanks for the info.

Crysis 3 is a very different game depending on which level you play,root of all evil compared to welcome to the jungle compared to any other level will give you completely different results so you cannot compare benches from different sites.

This. Not that FX fares much better in other levels. Barely matches older Core i5s @ Welcome to the Jungle.

The conversation was about 4k gaming -- don't try to move the goalposts.

Every game is GPU bound at 4K! [/B]The AMD FX chips seem to minimize frame time variance better when playing at those 4k resolutions (versus Intel CPU's). This isn't a difficult concept to grasp.

You're basing this on what, exactly? There's more recent TweakTown benchmarks than the one you posted, and they show that 3-year-old Core i7-4770K is faster than the FX with GTX980 Tis in SLI @ 4K.

And considering that your 4K (GTA V) results are completely different from this:

Where performance was very similar for all chips (GPU bound), I find them dubious at best.

 

[Abwx]

Lol, just noticed Skylake Core i5 is 15.6% slower at the isolated WinRAR test but only 10.5% slower when running WinRAR + The Wicther 3 compared to the FX8350. There goes your fallacy about multitasking.

It does relatively well here but if the second task is FP instead of integer the result is, well, awsome..

You could game while rendering but this latter task would collapse by 68%, that is 3x the time while the FX would see the rendering task losing only 39%, not counting that it start from a higher FP score than the i5 6600K despite the FP soft being Cinebench R15...

So much about the multitasking fallacy, i5s are not always good at it, basically it must be Integer + Integer but not FP + Integer.

So both Hardware.fr and PCLab put the latest Core i5 above all AMD CPUs in this game. Thanks for the info.

The latest, and with the limitations mentioned above while it is years that you are hyping the IB and HW i5s...

 

[Sweepr]

The latest, and with the limitations mentioned above while it is years that you are hyping the IB and HW i5s...

Only 220W FX9590 can put up a challenge to 2012/2013-era Core i5s, and this one is definitely not the most popular Vishera CPU. Anyway your discussion is pointless, a slight OC would put 4 year old Core i5 3570K out of reach for all AMD chips @ Hardware.fr. Also really strange that Haswell is slower than Ivy Bridge in their test, different from PCLab:

 

[VirtualLarry]

So much about the multitasking fallacy, i5s are not always good at it, basically it must be Integer + Integer but not FP + Integer.

This, my friends, is complete ... bull sheet.

i5 CPUs do NOT support HyperThreading. So there is NO SHARING between Integer or FP tasks, on a core. None. The L3 cache which is shared, IS affected, so it can depend on the cache locality of what is running on all four cores. But again, that has NOTHING to do with whether the tasks running on the CPU stress the INT or FP capabilities of the core.

Each thread is scheduled on a FULL CPU CORE. THERE IS NO SHARING WITHOUT HYPERTHREADING.

Methinks Abwx doesn't understand OS CPU scheduling at all.

 

[Abwx]

This, my friends, is complete ... bull sheet.

i5 CPUs do NOT support HyperThreading. So there is NO SHARING between Integer or FP tasks, on a core. None. The L3 cache which is shared, IS affected, so it can depend on the cache locality of what is running on all four cores. But again, that has NOTHING to do with whether the tasks running on the CPU stress the INT or FP capabilities of the core.

Each thread is scheduled on a FULL CPU CORE. THERE IS NO SHARING WITHOUT HYPERTHREADING.

Methinks Abwx doesn't understand OS CPU scheduling at all.

In this test there s no scheduling as it would amount to serialize the tests wich is not the point of this bench, the exe units are stressed simultaneously by the two softs and what is measured is the ability to sustain the throughputs.

If the schedulers where acting this would just yield the times sum of the two tests done separately, as for i5 not supporting HT this is irrelevant, as discussed ad nauseam it s likely the cache size that is the (willfull..) segmenting factor, at some point there s not enough bytes/thread in the L3.

 

[Phynaz]

No scheduling...it must be the transconductance that eliminates the need .

 

[VirtualLarry]

In this test there s no scheduling

The OS scheduler is always active.

at some point there s not enough bytes/thread in the L3.

So now you admit that your earlier statement had no basis in fact.

 

[SlowSpyder]

Just bought Crysis 3. So far it plays very well on my 4.4GHz FX8320E. 45-60FPS, all settings maxed except using 2xMSAA, R9 390, 19x12 resolution. We'll see how it goes, but so far so good, graphics are very sharp so far... good looking game!

 

[VirtualLarry]

the exe units are stressed simultaneously by the two softs
as for i5 not supporting HT this is irrelevant

The i5 is a 4C/4T CPU. Each core has DEDICATED int and FP resources. Since there is no HT, they are not shared. They are NOT stressed simultaneously, the threads are time-sliced, and scheduled within their own quantum by the OS scheduler, on their own dedicated cores. Again, no HT, no exe resource sharing.

The i5 does NOT have just one huge shared core, that runs four HyperThreads. There are four separate cores, each with dedicated execution resources. (Unless Intel has been lying all this time, and their CPUs only have one core...)

Edit: The only things on the CPU that are shared, are the memory controller and the L3 cache (the uncore - also things like the PCU, the power control unit).

 

[TheELF]

In this test there s no scheduling as it would amount to serialize the tests wich is not the point of this bench, the exe units are stressed simultaneously by the two softs and what is measured is the ability to sustain the throughputs.

No,except for cinebench every other test they do does not use 100% of all the CPUs, they are finite tests,the game gets limited by the gpu or by the amount of threads it runs and they don't run the winrar benchmark but really do some compression which means that it will only use a percentage of some CPUs before mem/storage becomes the bottleneck.

Bottom line,of course both the game and winrar are going to use a much larger percentage of the dual core i3 then from the 8 cored fx-8xxx.

If you want to measure the ability to sustain the throughput you will have to find two things to run that can run with 100% usage on any CPU when running alone.
Not two things that only run with some percentage of a high core CPU,this way you only measure how much of your multicored CPU does not get used by everyday tasks,in other words how much of your cash gets wasted if you do only one thing on your CPU.

 

[Abwx]

The i5 is a 4C/4T CPU. Each core has DEDICATED int and FP resources. Since there is no HT, they are not shared. They are NOT stressed simultaneously, the threads are time-sliced, and scheduled within their own quantum by the OS scheduler, on their own dedicated cores. Again, no HT, no exe resource sharing.


The i5 does NOT have just one huge shared core, that runs four HyperThreads. There are four separate cores, each with dedicated execution resources. (Unless Intel has been lying all this time, and their CPUs only have one core...)

Edit: The only things on the CPU that are shared, are the memory controller and the L3 cache (the uncore - also things like the PCU, the power control unit).

HT has nothing to do with it, Intel s uarch use exe clusters and the Integer exe units (ALUs) are in the same clusters as the FP exe units, HT is at the front end level and not within the exe ressources, it act as a second reservation unit within the actual reservation station..

The difference between the i5 and the i7 is that the latter has added cache to serve 4 other threads, the same way i3s have the necessary added cache to deal with 2 more threads than the Pentium/Celeron.

No,except for cinebench every other test they do does not use 100% of all the CPUs, they are finite tests,the game gets limited by the gpu or by the amount of threads it runs and they don't run the winrar benchmark but really do some compression which means that it will only use a percentage of some CPUs before mem/storage becomes the bottleneck.

Bottom line,of course both the game and winrar are going to use a much larger percentage of the dual core i3 then from the 8 cored fx-8xxx.

If you want to measure the ability to sustain the throughput you will have to find two things to run that can run with 100% usage on any CPU when running alone.
Not two things that only run with some percentage of a high core CPU,this way you only measure how much of your multicored CPU does not get used by everyday tasks,in other words how much of your cash gets wasted if you do only one thing on your CPU.

CB is a FP task and what matters is to use an FP task; it could be PovRay or 3DSMax the outcome would be the same.

Same for the game + Winrar, this latter is used as Integer code soft but it could be whatever other soft using Integer code, let say some X264 encoding, because one would eventualy play a game while his PC is encoding or rendering, isnt it...

At the end the usual serialised benches do not give a good idea of a CPU versatility, and that s all what this thread is about after all..

 

[VirtualLarry]

HT has nothing to do with it, Intel s uarch use exe clusters and the Integer exe units (ALUs) are in the same clusters as the FP exe units, HT is at the front end level and not within the exe ressources, it act as a second reservation unit within the actual reservation station..

The difference between the i5 and the i7 is that the latter has added cache to serve 4 other threads, the same way i3s have the necessary added cache to deal with 2 more threads than the Pentium/Celeron.

I'm not really sure what you're trying to say here.
Here's a Skylake die shot:

Notice the four identical cores (mirrored)?

Edit: Maybe I do need to spoon-feed you.

Each one of those four cores, has independent, dedicated int and FP pipes. Threads get scheduled on those four independent cores, by the OS scheduler, one timeslice (quantum) at a time. So it doesn't matter if the four scheduled threads are int- or FP-heavy, or a mixture of the two. Each thread gets dedicated resources.

Or were you referring to a SINGLE THREAD's execution on a core, containing either all-int, all-FP, or a mixture of the two? That would have to do with the IPC of a single thread, and wouldn't have anything to do with a benchmark where two separate application processes, each consisting of several threads, were run at the same time on the OS.

 

[MiddleOfTheRoad]

Must have been easy to find, huh?

Any reviewer can deliberately sandbag benchmarks. For a Skylake review, someone could track down some OEM DDR4 1600 Mhz memory for the benches..... Probably would make it slower than an Ivy Bridge. It wouldn't be the first time a review site has done it.

 

[Abwx]

Each one of those four cores, has independent, dedicated int and FP pipes. Threads get scheduled on those four independent cores, by the OS scheduler, one timeslice (quantum) at a time. So it doesn't matter if the four scheduled threads are int- or FP-heavy, or a mixture of the two. Each thread gets dedicated resources.

Or were you referring to a SINGLE THREAD's execution on a core, containing either all-int, all-FP, or a mixture of the two? That would have to do with the IPC of a single thread, and wouldn't have anything to do with a benchmark where two separate application processes, each consisting of several threads, were run at the same time on the OS.

For instance a core can load two operands related to a given thread, let say Winrar one, while in the same cycle loading an operand from say a CB thread, during the relevant cycle the three (or more) ops related to theses two threads can be likewise scheduled to the exe units..

So during a cycle ops from different threads can be executed, this allow maximal efficency of the CPU since exe ressources are kept more busy, a pure scheduling would amount to reduce efficency as this would amount to using exclusive cycles for each apps and the result would be the same (at best) as executing the tasks one after the other.

Now there s CPUs that cope better with more threads, particularly when theses are from separate applications.

 

[VirtualLarry]

For instance a core can load two operands related to a given thread, let say Winrar one, while in the same cycle loading an operand from say a CB thread, during the relevant cycle the three (or more) ops related to theses two threads can be likewise scheduled to the exe units..

So during a cycle ops from different threads can be executed, this allow maximal efficency of the CPU since exe ressources are kept more busy, a pure scheduling would amount to reduce efficency as this would amount to using exclusive cycles for each apps and the result would be the same (at best) as executing the tasks one after the other.

That is basically what HyperThreading does... it shares core resources between two threads. The i5 does not have HyperThreading. I think your understanding of core resources and how threads execute, is fundamentally flawed. Either that, or Intel is lying about the design and architecture of their CPUs.

To elaborate, without HyperThreading enabled on an Intel CPU, a core will NOT share threads. Each thread gets a core of its own to run on. Each core has dedicated load/store, int, FP, AGU, etc., resources.

Yes, HyperThreading has advantages for throughput, it allows cores to be more fully utilized. But the i5 doesn't have it.

 

[Abwx]

That is basically what HyperThreading does... it shares core resources between two threads. The i5 does not have HyperThreading. I think your understanding of core resources and how threads execute, is fundamentally flawed. Either that, or Intel is lying about the design and architecture of their CPUs.

To elaborate, without HyperThreading enabled on an Intel CPU, a core will NOT share threads. Each thread gets a core of its own to run on. Each core has dedicated load/store, int, FP, AGU, etc., resources.

Yes, HyperThreading has advantages for throughput, it allows cores to be more fully utilized. But the i5 doesn't have it.

This is what an out of order engine does, that s why i m saying that HT has nothing to do with these perfs penalties, it has so little to do that there s softs where HT bring little as the soft manage to get close to the full throughput with 4 threads, in Povray for instance :

http://www.computerbase.de/2014-09/amd-fx-8370e-im-test/2/#diagramm-pov-ray

 

[VirtualLarry]

This is what an out of order engine does, that s why i m saying that HT has nothing to do with these perfs penalties

EACH CPU core is an OoO engine, for EACH thread. The way you are describing the CPU, it's just one giant OoO core, that runs four threads. That's NOT what the die shot that I embedded shows. It shows four COMPLETE, INDEPENDENT cores.

Edit: If I have this right, OoO is for ILP, and HT is for TLP.

And HT has everything to do with core exe resources and contention between threads, because on an Intel CPU (currently), that's the only way multiple threads would end up sharing core resources. That doesn't happen on an i5, because it doesn't have HT enabled.

I don't know why I'm even trying to educate you about this. You don't even think threads are scheduled, on a multi-tasking OS, when you run two application processes at the same time.

The fact that multiple people have attempted to educate you about these issues, and you continue to bring up the "and WinRAR" benchmarks, to claim that an i5 supposedly sucks at multitasking, when that's clearly not true, proves that you are simply a troll or a shill.

 

[AtenRa]

@ VirtualLarry

Abwx talks about Intel Core Execution ports (Int, FPU, Mul, Add etc, etc)

 

[coercitiv]

If the scores and the archival times are from the same instance, the Cinebench + WinRaR test is very interesting. It does indeed show the FX chip to suffer less of a relative penalty when running this kind of loads, as compared to both i5 6600K and i7 6700k. The HT enabled i7 sees a smaller relative penalty in Cinebench, but the WinRaR times are considerably higher (again, relative to self).

However, other results should also be observed: the HEDT platform shows no weakness to the test, with very little degradation in both Cinebench scores and WinRaR times, a very stark contrast to Haswell scores from the mainstream platform. This indicates that either WinRaR is unable to really stress 12+ threads or that there is something else at play here other than common parts of the CPU architecture.