Is AMD still undefeated in budget builds?

[mohit9206]

Yes i agree that when it comes to budget cpu's,Intel's Pentium and i3 are good for gaming but on other forums especially on Toms,i always see even the most experienced forum members recommending the fx-6300 over the i3 and athlons over the pentiums citing only games like bf4 without considering other games and comparison benchmarks considering even today most games still only heavily use 2 cores and use other cores lightly.

 

[Erenhardt]

I wonder how much impact would real world usage conditions do to the performance of 2core intel and 4 core amd.

I mean, those test are done with everything closed, only game running. What happens when you have some background tasks running. Torrent client, skype, music playback, etc.

I imagine 2 intel cores would suffer more than 4 amd cores. I think the %utilization of 4 core CPU is a lot lower than 2core.

 

[ShintaiDK]

I wonder how much impact would real world usage conditions do to the performance of 2core intel and 4 core amd.

I mean, those test are done with everything closed, only game running. What happens when you have some background tasks running. Torrent client, skype, music playback, etc.

I imagine 2 intel cores would suffer more than 4 amd cores. I think the %utilization of 4 core CPU is a lot lower than 2core.

I think you overestimate background utilization wastly.

 

[Erenhardt]

I think you overestimate background utilization wastly.

Nope, I've seen what Mantle does when there is something running in the background. As simple as windows task manager is enough to make a noticable difference when CPU is running at its limits.

EDIT:
To make my point clearer

2 core CPU have only 3% of single core processing power left
4 core CPU have 36% of single core processing power left

 

[ShintaiDK]

As simple as windows task manager is enough to make a noticable difference when CPU is running at its limits.

Run perfmon and see for yourself. And something like taskmgr wouldnt even use 1% at 800Mhz.

You also compare apples and oranges.

 

[2is]

Nope, I've seen what Mantle does when there is something running in the background. As simple as windows task manager is enough to make a noticable difference when CPU is running at its limits.

EDIT:
To make my point clearer

2 core CPU have only 3% of single core processing power left
4 core CPU have 36% of single core processing power left

I have task manager as well as msi afterburner running quite regularly on a 2nd monitor to monitor usage and have never noticed any amount of performance degradation in the least no matter what game I play vs having them closed... If this is something you've experienced, it must be an AMD thing. In which case, your imagination of 4 AMD cores being > two intel cores is just that... Imagination.

 

[Erenhardt]

My theory is in line with tests done by AtenRa in Thief:
http://forum.oktabit.gr/topic/thief-mantle-cpu-scaling-and-power-evaluation

Look at that G3420 performing the same as fx-8320! (minus minimums)

My theory says it is because big part of fx is simply idling while g3420 runs at 99%.

Lets see when we enable mantle - which should increase CPU utilization:

G3420 gains noting (some minimums) while FX gets +50%. Even athlon 750k is noticably faster (10%) than g3420.

I think background tasks would have comparable effect on gaming performance.

 

[Erenhardt]

I have task manager as well as msi afterburner running quite regularly on a 2nd monitor to monitor usage and have never noticed any amount of performance degradation in the least no matter what game I play vs having them closed... If this is something you've experienced, it must be an AMD thing. In which case, your imagination of 4 AMD cores being > two intel cores is just that... Imagination.

Are you using 2 thread intel celeron or 8 thread core i7?

 

[ShintaiDK]

My theory is in line with tests done by AtenRa in Thief:

And you couldnt have picked a more untrustworthy source.

Now as I told you. Run perfmon and see what your background usage is. Then we dont have to filter out all the rubbish.

 

[monstercameron]

And you couldnt have picked a more untrustworthy source.

Now as I told you. Run perfmon and see what your background usage is. Then we dont have to filter out all the rubbish.

how is the source untrustworthy or more untrustworthy than any other source?

 

[blackened23]

how is the source untrustworthy or more untrustworthy than any other source?

Graph from some random forum member. I'm sure that wasn't manipulated for a specific outcome, right? We have assurances that the person making that graph is an objective source, right? Or is it some wingnut AMD fan trying to show a specific result? I don't know. It's a forum member from somewhere. Are they an AMD reseller? AMD sponsored? I don't know. We don't know anything about this supposed source. Conversely, I could take a source like guru3d and know who's behind it, and trust it based on precedent. But this graph, don't know the source, random forum member from some website in Greece. It's not a review website. I can trust some review websites. Random forum member? Nah.

 

[2is]

Are you using 2 thread intel celeron or 8 thread core i7?

I've been running task manager pretty regularly while gaming for well over a decade now and I've never noticed any bit of a performance loss. This spans processors starting from a single core 1.33GHz Athlon Thunderbirds to i7's

 

[shady28]

And you couldnt have picked a more untrustworthy source.

Now as I told you. Run perfmon and see what your background usage is. Then we dont have to filter out all the rubbish.

That's ridiculous. A real-world test would be simple to do given similar systems, however for whatever reason these big-ticket benchmark sites no longer do that kind of test.

But they did, for a short time.

Here you see in 2010, an i3-540 vs some low end Phenom II x4 chips.

So you start off with a very near tie between an i3-540 and a Phenom II x4 940/945 while just doing the game benchmark, with the Phenom taking a slight lead (~1-2% faster).

Then they do the same test with just one extra task - a RAR archive creation - in the background. The result is a 20% higher minimum frame rate and a 25% higher max frame rate on the Phenom.

Absolutely nothing you have said in this thread would lend any credence to thinking the results would be any different on say a 6-core FX-6300 vs a 2-core i3. In fact, I suspect the difference would be greater.

Then you've got stuff like this, where gaming while streaming shows 8-core AMD with a significant lead vs quad core Intels.

https://teksyndicate.com/videos/amd...s-3820-gaming-and-xsplit-streaming-benchmarks

I mean really, an i3 is just not a good choice for any kind of power user.

 

[BSim500]

2 core CPU have only 3% of single core processing power left

I mean, those test are done with everything closed, only game running. What happens when you have some background tasks running. Torrent client, skype, music playback, etc.

2x phenom II X2 550 cores does not equal 2x Intel Haswell cores. Even a Pentium G3420 is 66% faster per core vs an X2 550, and an i3-4130 something like 85% faster per core + another 20-30% Hyper-Threading on top vs the Pentium with 4 threads... It's amazing people still just count the cores and ignore the massive 60-85% disparity in IPC. Like the past few years didn't happen and it's still 2010...

I've been running task manager pretty regularly while gaming for well over a decade now and I've never noticed any bit of a performance loss. This spans processors starting from a single core 1.33GHz Athlon Thunderbirds to i7's

Agreed. A lot of "background task" estimates are wildly overrated. Task manager (I prefer Process Explorer myself) barely uses 1%. Firefox something like 1-3% with 50 tabs open. Even having a Hauppauge TV card installed uses something like 0.02% idle and 1.5% recording. If people are genuinely seeing a lot of CPU being "eaten up by background tasks", they should seriously do a scan for viruses / trojans / keyloggers or install a non buggy driver... My low-end i3 HTPC rig shows 97-98% idle most of the time, and that's with a TV card recording + NextPVR running (and 1% of that is Process Explorer itself)...

Same with unrealistic "real world" usage of running a game and playing music and using Skype all at the same time. How the hell is the person on the other end of Skype going to hear what you're saying with music + a game soundtrack blaring out? And if the game is in the foreground, then you can't see Skype. And if Skype is in the foreground, then the game is usually paused... If you're talking on Skype / the phone you usually mute music (which itself uses all of 1% in Winamp)... There are perfectly valid things that a quad-core is faster than dual-core at (video encoding, etc), but there's also a ton of deliberately absurd exaggerations of "dual-core is dead" involving throwing out unrealistic scenarios like running a WinRAR benchmark whilst simultaneously playing Crysis at glorious 1280x1024 resolutions & 0AA (like it's still 1999), etc, to 'prove a point'...

Edit: And even the slowest Haswell i3 runs BF4 just fine.

 

[SlowSpyder]

And you couldnt have picked a more untrustworthy source.

Now as I told you. Run perfmon and see what your background usage is. Then we dont have to filter out all the rubbish.

I have no idea if Erenhardt is right or wrong regarding background tasks possibly effecting a 2C/4T Intel more than a two module AMD CPU. But at least he's showing some reasoning for his line of thought, all you've done so far is suggest AtenRa is a liar. Do you have an example of something he hasn't been truthful about? If not, kind of a toll comment on your part.

 

[Erenhardt]

2x phenom II X2 550 cores does not equal 2x Intel Haswell cores. Even a Pentium G3420 is 66% faster per core vs an X2 550, and an i3-4130 something like 85% faster per core + another 20-30% Hyper-Threading on top vs the Pentium with 4 threads... It's amazing people still just count the cores and ignore the massive 60-85% disparity in IPC. Like the past few years didn't happen and it's still 2010...

You totally missed my point.

There is a CPU utilization pattern.

2 thread CPU is using one thread pegged at 99% and the other at 98% if there is enough graphics power (CPU bound scenario). So, 3% of single core time is unused.
The same scenario with 4 thread CPU is much different. Not a single thread is utilized above 95%, two are in the low 90's and one is even lower. More than 30% of single thread is unused, which can be dedicated to background tasks.

All that is based on one of the most multithreaded games we have.

The Difference in IPC have nothing to do here.

There is only 3% of single thread free to use in 2 thread CPU and more than 30% in thread time in 4 thread CPU.

Simple math, 2-thread CPU needs to have 10 times higher IPC to have as much free processing power as 4-thread CPU in this scenario.

It can be looked at from different angle. 2-thread CPU is running at 98,5%[(99+98)/2] of its capability, while 4-thread CPU at 91% [(95+91+91+87)/4].

There is only 1,5% free CPU time in first example, and 9% free CPU time in second.

Quick math: First CPU needs to have 9 times more IPC to have the same amount of unused processing power as the second one given the usage scenario.

I agree that intel IPC is superior to AMD (funny how fx8350 is 8 core when we talk about IPC and 4 core when we talk about possible performance in synthetic tests - whatever suits the situation)

As to the source of mantle benchmarks...Well it shows intel is fastest, and it shows g3420 in very good light in dx. Or did I misunderstood you, and you were trying to say AtenRa is biased towards intel?

 

[Erenhardt]

I've been running task manager pretty regularly while gaming for well over a decade now and I've never noticed any bit of a performance loss. This spans processors starting from a single core 1.33GHz Athlon Thunderbirds to i7's

Software changed quite a bit. I have still 32MB hard drive in the closed. Titanfall needs 50 GBs.

Nasa flew to the moon with less processing power than people put in their pockets when going outside, which now is not enough to go through facebook!

Not sure if I understand the point you are trying to make.

 

[frozentundra123456]

Seems kind of academic to me. Guess I grew up when cpus were a lot less powerful, but it just makes sense to me if one is trying for maximum performance in a cpu intensive task like a game that they would minimize background tasks. I mean can anyone really skype while they are playing a game? As for downloading, dont think that would be as much of a burden on the cpu as on the internet bandwidth if you were trying to play online.

 

[Erenhardt]

Seems kind of academic to me. Guess I grew up when cpus were a lot less powerful, but it just makes sense to me if one is trying for maximum performance in a cpu intensive task like a game that they would minimize background tasks. I mean can anyone really skype while they are playing a game? As for downloading, dont think that would be as much of a burden on the cpu as on the internet bandwidth if you were trying to play online.

I was always behind a generation or two when it comes to computer hardware. I always had to kill all background tasks to have enjoyable gameplay.

And here, we are discussing the same situation, but instead using old top of the line hardware, we are using bottom end current generation. Both situations lead to CPU bottleneck we are discussing.

 

[chimaxi83]

Graph from some random forum member. I'm sure that wasn't manipulated for a specific outcome, right? We have assurances that the person making that graph is an objective source, right? Or is it some wingnut AMD fan trying to show a specific result? I don't know. It's a forum member from somewhere. Are they an AMD reseller? AMD sponsored? I don't know. We don't know anything about this supposed source. Conversely, I could take a source like guru3d and know who's behind it, and trust it based on precedent. But this graph, don't know the source, random forum member from some website in Greece. It's not a review website. I can trust some review websites. Random forum member? Nah.

You don't know anything, but you freely throw out the accusations all over the forums, don't you?

 

[ShintaiDK]

Snip

But RAR is not what you consider regular background tasks.

 

[ShintaiDK]

I have no idea if Erenhardt is right or wrong regarding background tasks possibly effecting a 2C/4T Intel more than a two module AMD CPU. But at least he's showing some reasoning for his line of thought, all you've done so far is suggest AtenRa is a liar. Do you have an example of something he hasn't been truthful about? If not, kind of a toll comment on your part.

I told him to run perfmon rather than making random assumptions from random benchmarks that is entirely unrelated.

 

[Erenhardt]

I told him to run perfmon rather than making random assumptions from random benchmarks that is entirely unrelated.

Can you somehow relate to the situation we are talking about? Or you are happily sitting on intels 8 thread CPU?

Are you suggesting someone 1.0L V2 car, because you are happy with how your 8.0L V16 performs? We know 8.0L beast will pull camping trailer and still be faster than most vehicles on the road, but that doesn't mean 1.0L V2 will be the same.

Have you tried mantle and saw what happens when the most basic tasks run in the background?

As to personal attacks on AtenRa...
Here, have your beloved PcLab:

<10% gain. 56 FPS with 290X

+60% gain. 63 FPS with 290X

Conclusion: There is a lot of untapped potential in 4+ Thread CPUs which can be used to run background tasks. At the same time those background tasks will take away CPU time that was previously dedicated to game when there are only 2 thread that are already fully used.

 

[NTMBK]

Graph from some random forum member. I'm sure that wasn't manipulated for a specific outcome, right? We have assurances that the person making that graph is an objective source, right? Or is it some wingnut AMD fan trying to show a specific result? I don't know. It's a forum member from somewhere. Are they an AMD reseller? AMD sponsored? I don't know. We don't know anything about this supposed source. Conversely, I could take a source like guru3d and know who's behind it, and trust it based on precedent. But this graph, don't know the source, random forum member from some website in Greece. It's not a review website. I can trust some review websites. Random forum member? Nah.

You make a valid point; we don't know who anyone on this forum is. He could be paid by AMD. I could be in VIA's pocket. Heck, you could be the CEO of Intel! Think about it:

blackened23
blackened
BlacKened
BK
Brian Krzanich

Coincidence? You decide :hmm:

 

[BSim500]

Here, have your beloved PcLab:

What's funny is that the nearest price equivalent to the FX-6300 is the i3, which you "accidentally" ommitted, presumably because it managed to match an FX-8350 / FX-6350 and beat an FX-4350 by 10% - even at a whopping 1.2GHz disadvantage:-

http://pclab.pl/art56897-3.html

Simple math, 2-thread CPU needs to have 10 times higher IPC to have as much free processing power as 4-thread CPU in this scenario. It can be looked at from different angle. 2-thread CPU is running at 98,5%[(99+98)/2] of its capability, while 4-thread CPU at 91% [(95+91+91+87)/4]. There is only 1,5% free CPU time in first example, and 9% free CPU time in second. Quick math: First CPU needs to have 9 times more IPC to have the same amount of unused processing power as the second one given the usage scenario.

Your maths are woefully flawed. A 2x IPC is essentially a doubling in a CPU's power whether code is multi-threaded or not. A 10x higher IPC is an increase in CPU power of a staggering 900% - again whether code is well threaded or not. You're basically arguing a 3.2GHz CPU vs the same CPU equivalent essentially OC'd to 32GHz will both have the same % load, and that the extra 10x power will lower CPU usage by only 10%. Um, no... A 10x increase in CPU power IPC / clock speed would lower 80-90% CPU usage to well under 20%. This is already seen in 1990's games that sucked up 99% usage on a 466Mhz Celeron that now barely pull 15-20% on a 3GHz Haswell Pentium today. A 900% increase in CPU performance doesn't = a 10% increase in idle time, because the load % is not static as IPC / clock speed changes, ie, 2.7Ghz used out of 3.0GHz (90%) is not going to remain 90% when that 2.7Ghz used is now out of a 10x more powerful CPU with "horsepower" equivalent to the same CPU clocked at over 25-32GHz, (to use your example)...

Or did I misunderstood you, and you were trying to say AtenRa is biased towards intel?

You misunderstood me. In fact I think you have me mixed up with someone else as I never said anything about AtenRa being "biased towards Intel". It wasn't me who you originally responded to, either, I merely pointed out the seemingly popular absurdity in "stacking the deck" with running totally unrealistic "background tasks" to 'prove a point'. Eg, in the real world, you can't even ergonomically use Skype & gaming at once simply because your eyeballs won't be able to focus / concentrate on the game whilst using Skype (unless you're cross-eyed). It's like claiming the advantage of a quad-core tablet is doing a crossword puzzle, watching TV and holding a video conference call all at the same time - It's simply not realistic multi-tasking even on 8/16/32-core CPU purely due to human ergonomics / concentration limitations, and the way all these "typical background tasks" only ever seem to spring up when discussing 2-core Intel's vs 4-core AMD's budget builds yet disappear when i5/i7's come into the equation, is pretty obvious bias. ie, some people are embarrassed that a 2-core Intel i3 beats even 6-8 core AMD chips in some 2014 games even when clocked over 1GHz slower, so try and "move the goalposts" by pretending people on a budget "run more in the background" when gaming than they actually do. In reality the opposite is true - people on a budget with common "get the most out of your bang-per-buck hardware" sense are those intelligent enough to run "heavy background stuff" when they aren't at their PC (eg, encoding a queue of videos overnight / when at work, virus scanning when eating dinner, etc) so that even heavy tasks doesn't impact their weaker CPU's when they're sitting in front of them. Common sense time management.

But RAR is not what you consider regular background tasks.

Precisely. Nor does anyone run 2x separate benchmarks simultaneously on any CPU as you wouldn't get an accurate reading on either. Downloading a file in a web browser, listening to MP3's, etc, literally use 1% CPU usage. You hear the same thing with console vs PC : "Consoles are better than PC's man because Windows and all that stuff on a PC sucks up at least, let's say 30-40% background CPU! Yeah - that's it - 40%!". LOL. What "stuff" would that be exactly? Objective metrics anyone? For heavy future mainstream gaming I'd definitely recommend a quad i5 over an i3, but some of the excuses people make for why "i3's can't run games" (whilst there are hundreds of Youtube FRAPS vids to the contrary) for budget builds by making up unrealistic "typical background task usage" scenarios or comparing $70 Pentium's to double the price $140 FX-6350's is pretty silly.