Is the i5 still enough? Bottlenecking explained with an RX-480.

[i_build_nanosuits]

i7-4770K @ 4.3ghz with 42 multiplier on the ring bus, 2000mhz CL9 ram (OC) with GTX 980ti 1440mhz core 8100mhz memory...all games run SUPER smooth at 2560x1440 on maximum settings (except anti-aliasing of course, i usually use FXAA or something)...CPU load anywhere from 65 to 95% in BF1 and watchdogs 2, no stuttering what so ever, flawless gaming experience.

 

pop into BIOS, disable hyper-threading, boot windows, launch BF1...playable, but sub-par experience, intermitent stuttering when intense battle is raging, 100% CPU load across the board, GPU load bouncing between 85 and 99%, more often than not 92-95% GPU load (at 1440p) CPU bottleneck land, i wouldn't play BF1 on this 24/7...

 

Launch watchdogs 2...stutterfest...unplayable...feel like i have a shitty PC...

 

Pop back into BIOS, enable hyper-threading. done!

[xAcid9]

On 3/13/2017 at 9:04 PM, xAcid9 said:

Sorry forgot to post this yesterday, GTA:V with your settings. 

 

4C/8T - Slightly longer because i hit couple of cars.  75fps average

  Hide contents

4C/4T - 73 fps average

  Reveal hidden contents

 

Today someone was kind enough to give his old Z77 board to me, first i thought kinda pointless because of my locked processor.. apparently not.

 

Overclocked my RAM to 1800mhz and BCLK to 103, yield me a pretty good gain in some CPU intensive games especially GTA:V

 

4C/8T - 88.5 fps average.

Spoiler

I'm actually hitting GPU bottleneck in many places around Los Santos now and also MSI AB 100ms polling hit the FPS quite abit. Pretty satisfied for $0 cost and 30mins work.

[Majestic]

1 hour ago, xAcid9 said:

Today someone was kind enough to give his old Z77 board to me, first i thought kinda pointless because of my locked processor.. apparently not.

 

Overclocked my RAM to 1800mhz and BCLK to 103, yield me a pretty good gain in some CPU intensive games especially GTA:V

 

4C/8T - 88.5 fps average.

  Reveal hidden contents

I'm actually hitting GPU bottleneck in many places around Los Santos now and also MSI AB 100ms polling hit the FPS quite abit. Pretty satisfied for $0 cost and 30mins work.

Dude, memory scaling is pretty insane these days. I upgraded to 2400 CL11 when I doubled my ram from 8 to 16. That alone was worth the investment. Tomb Raider's difference between stock and OC is mainly the memory.

[Citadelen]

Very interesting video on how AMD and Nvidia handle scheduling and thread management differently, but also about why Ryzen performs so bad in some games.

[sgloux3470]

3 hours ago, Citadelen said:

Very interesting video on how AMD and Nvidia handle scheduling and thread management differently, but also about why Ryzen performs so bad in some games.

 

I loved this video.  It really points out that while AMD and Nvidia are very closely matched, there are some very significant fundamental hardware differences that can definitely change how games should be programmed between the two. (Which is also something we're seeing a little with Ryzen's CCX system.)

[Majestic]

Added two more games, can anyone test how much better Catalyst runs on something with more than 4 cores on 1080p with a 1070 or higher?

[maartendc]

Thanks for these benchmarks, pretty interesting stuff! You put a lot of time and effort into these!

 

I am running a i5 6600 and a 980Ti, but I game at 1440P. My monitor is 60Hz, so I am not trying to run anything beyond 60 FPS anyway. I get a solid 60 FPS in most games I play at Ultra settings (Battlefield 1, Witcher 3, etc.)

 

I think my CPU usage is in the 60-80% range on most of these games, so not really seeing any major bottlenecking at 1440P, 60 Hz. So this makes me wonder how good of a comparison your OC haswell is to the Locked Skylake. I agree with you that theoretically, the OC haswell and locked Skylake should perform similarly. But I am not seeing the overall bottlenecking that you are seeing in Battlefield 1 or Witcher 3 (havent played all the games you tested). 

 

I haven't really looked into frame times for my system, but I am getting a good experience in Battlefield 1 etc. The only caveat is Playerunknown's Battlegrounds, where I do get framedrops ocassionaly, but this is a Early Access game, so it is not fully optimized.

 

I am going to take a look at my frametimes to see if I am getting any stutters. Did you measure these with MSI Afterburner? I did not know it could measure frametimes actually, interesting!

 

I totally agree with your point that benchmarks usually only show average FPS and not frame times. This makes a big difference. Although most websites are getting better at this, slowly.

 

 

 

 

 

[Majestic]

@maartendc Well, these were done at 1080p both with a 480 and 1070 with unlocked framerate. You need to look at the specific framerates the games are getting to see if it'd effect you. It's why I specifically mention for each game whether they stay above 60fps. Also, none of the games you mentioned are in the test. As for the difference between skylake and haswell.

 

 

 

 

Here you can check a CPU intensive game like Crysis 3. The top chart is stock, bottom chart is OC. You can compare the stock skylake scores with the bottom OC haswell scores.

[maartendc]

41 minutes ago, Majestic said:

@maartendc Well, these were done at 1080p both with a 480 and 1070 with unlocked framerate. You need to look at the specific framerates the games are getting to see if it'd effect you. It's why I specifically mention for each game whether they stay above 60fps. Also, none of the games you mentioned are in the test. As for the difference between skylake and haswell.

(...)

 

Here you can check a CPU intensive game like Crysis 3. The top chart is stock, bottom chart is OC. You can compare the stock skylake scores with the bottom OC haswell scores.

Huh, interesting, so an OC haswell even outperforms a stock Skylake. GO figure.

 

Well thinking about it some more, I find it hard to believe that the i5 would have so much trouble reaching 60 FPS in all those games, regardless of resolution? Forgive me, but I have trouble reading the graphs, cannot really see minimum, / average FPS.

 

In my experience, the i5's get around 90-110 fps in most current games when CPU bound (so either low-res or very powerful GPU). The i7's are usually getting about 10 or 20 fps extra when CPU bound (so from 110 to 130 fps etc.) This should be regardless of resolution used, this is just how many frames the CPU can maximally push out when CPU bound.

 

My i5 6600 is getting 90 fps average in Battlefield 1, which is supposed to be a really CPU intensive game in 64 player multiplayer..  Again, not testing the same games here, but these benchmarks are surprising to me.

[TheDankKoosh]

At this point do you think locked i7 on a midrange board (b250 for example) with a cryorig c7  (or any other cheap and good cooler) is now the better buy? This is how I see it at this point, but I wanted to hear others opinions.

[Majestic]

44 minutes ago, maartendc said:

Huh, interesting, so an OC haswell even outperforms a stock Skylake. GO figure.

 

Well thinking about it some more, I find it hard to believe that the i5 would have so much trouble reaching 60 FPS in all those games, regardless of resolution? Forgive me, but I have trouble reading the graphs, cannot really see minimum, / average FPS.

 

In my experience, the i5's get around 90-110 fps in most current games when CPU bound (so either low-res or very powerful GPU). The i7's are usually getting about 10 or 20 fps extra when CPU bound (so from 110 to 130 fps etc.) This should be regardless of resolution used, this is just how many frames the CPU can maximally push out when CPU bound.

 

My i5 6600 is getting 90 fps average in Battlefield 1, which is supposed to be a really CPU intensive game in 64 player multiplayer..  Again, not testing the same games here, but these benchmarks are surprising to me.

Yeah, go figure the i5-4670K is from 2013. Not hard to imagine Ryzen catching Intel with their pants down, they were lazy af.

 

It's 20fps/division. So if the line is above 3 divisions, it's above 60fps. Min. fps is the lowest point on the graph. I put it this way because it gives the most complete picture, eventhough it requires more knowledge from the end-user.

 

Not sure why you're pulling those numbers out of thin air. As you can see the unconstrained numbers in the graphs I showed you. 

 

But I didn't test BF1, I tested the games I listed. Not all games are equally taxing. BF1 was also made for 64 players, those games I listed are made for 30fps. 

 

4 minutes ago, TheDankKoosh said:

At this point do you think locked i7 on a midrange board (b250 for example) with a cryorig c7  (or any other cheap and good cooler) is now the better buy? This is how I see it at this point, but I wanted to hear others opinions.

Nah, the Ryzen 5 are the better buy when going with new parts. Secondhand, you might do right by some i7 or Xeon.

[TheDankKoosh]

Just now, Majestic said:

 

 

Nah, the Ryzen 5 are the better buy when going with new parts. Secondhand, you might do right by some i7 or Xeon.

I was more so speaking about the intel platform, I know that r5 is a much better value compared to most intel offerings. 

[maartendc]

1 hour ago, TheDankKoosh said:

At this point do you think locked i7 on a midrange board (b250 for example) with a cryorig c7  (or any other cheap and good cooler) is now the better buy? This is how I see it at this point, but I wanted to hear others opinions.

I mean, do we really know that the current 4-core i7's do much better anyway? It is still only 4 cores, only adding multithreading. Does that really help out the frametimes in these games by that much? I don't really know where to look for good benchmarks on frametimes for CPU's, but I doubt it. I don't know if games take that much advantage of multithreading.

 

As shown in the graphs posted by @Majestic in the post above, a i7 multithreaded Haswell (i7 4770K) is beat by the non-multithreaded i5 6600K in OC mode. So the multithreading doesnt really offer that much benefit. These are average FPS though, don't know how multithreading impacts frametimes.

 

You might indeed be better off looking to Ryzen 1600, an affordable 6 core, or Ryzen 7, At least until Intel brings 6 and 8 cores to the mainstream i5 and i7 products.

[TheDankKoosh]

3 minutes ago, maartendc said:

I mean, do we really know that the current 4-core i7's do much better anyway? It is still only 4 cores, only adding multithreading. Does that really help out the frametimes in these games by that much? I don't really know where to look for good benchmarks on frametimes for CPU's, but I doubt it. I don't know if games take that much advantage of multithreading.

 

As shown in the graphs posted by @Majestic in the post above, a i7 multithreaded Haswell (i7 4770K) is beat by the non-multithreaded i5 6600K in OC mode. So the multithreading doesnt really offer that much benefit. These are average FPS though, don't know how multithreading impacts frametimes.

 

You might indeed be better off looking to Ryzen 1600, an affordable 6 core, or Ryzen 7, At least until Intel brings 6 and 8 cores to the mainstream i5 and i7 products.

In many cpu bound games the i7 should win, with the ones that depend on single core would have the 6600k winning, but either way the minimum I would consider for gaming would be an ivy bridge i7 (intel) or r5 1400 (amd) just because I'm very strict about how smooth my games feel. Others may feel different, but this is how I see things. 

[ivan134]

35 minutes ago, maartendc said:

I mean, do we really know that the current 4-core i7's do much better anyway? It is still only 4 cores, only adding multithreading. Does that really help out the frametimes in these games by that much? I don't really know where to look for good benchmarks on frametimes for CPU's, but I doubt it. I don't know if games take that much advantage of multithreading.

 

As shown in the graphs posted by @Majestic in the post above, a i7 multithreaded Haswell (i7 4770K) is beat by the non-multithreaded i5 6600K in OC mode. So the multithreading doesnt really offer that much benefit. These are average FPS though, don't know how multithreading impacts frametimes.

 

You might indeed be better off looking to Ryzen 1600, an affordable 6 core, or Ryzen 7, At least until Intel brings 6 and 8 cores to the mainstream i5 and i7 products.

I'm on my phone so it's hard to link. Look up techspot's Ryzen reviews and you'll see that i5s have bad minimums and in very CPU intensive games, they cause stuttering.

[Majestic]

11 hours ago, maartendc said:

I mean, do we really know that the current 4-core i7's do much better anyway? It is still only 4 cores, only adding multithreading. Does that really help out the frametimes in these games by that much? I don't really know where to look for good benchmarks on frametimes for CPU's, but I doubt it. I don't know if games take that much advantage of multithreading.

 

As shown in the graphs posted by @Majestic in the post above, a i7 multithreaded Haswell (i7 4770K) is beat by the non-multithreaded i5 6600K in OC mode. So the multithreading doesnt really offer that much benefit. These are average FPS though, don't know how multithreading impacts frametimes.

 

You might indeed be better off looking to Ryzen 1600, an affordable 6 core, or Ryzen 7, At least until Intel brings 6 and 8 cores to the mainstream i5 and i7 products.

 

 

The i7's are still better at the moment. But the games still need to be optimized for Ryzen.

[tom_w141]

This is still a thread since the launch of R5s?? R5s make i5s completely redundant unless they are snapped up as sub $150 used parts. 4 thread processors are at their limit now with 95-100% utilisation in games and no headroom for the future, no ability to multitask and streaming is out of the question. All which you can do on the unlocked, cheaper R5s, the gem of which being the 6c/12t 1600 for less than a 7600k.

[Majestic]

@tom_w141  It was actually drafted before Ryzen hit. And aren't you repeating what is being said? Don't understand the confusion.

[tom_w141]

2 minutes ago, Majestic said:

@tom_w141  It was actually drafted before Ryzen hit. And aren't you repeating what is being said? Don't understand the confusion.

I'm aware of that  Just wondering why OP wasn't updated to reflect it? In the current climate I'd never recommend an i5 over an R5 unless Intel drop the pricing by $80-100 or change what an i5 is in the future. Perhaps the coffeelake i5 will be 6c/6t, that's the only way I see it staying relevant.

[maartendc]

11 hours ago, Majestic said:

The i7's are still better at the moment. But the games still need to be optimized for Ryzen.

Right, but those are Average FPS.

 

I was referring more to the question of frametimes. We all know that average FPS is better on i7's and even i5's versus Ryzen, but what is more important in many cases is the better frame times on Ryzen. And like you said, games still need to be optimized for Ryzen and 6+ cores, so it might get even better in the future in terms of average fps as well.

 

As more and more people are becoming aware of, it is better to have a slower average fps, as long as the frametimes are better. The better frame time performance will look smoother rather than a higher average FPS with big framedrops.

[MyName13]

So according to this gaming with Pentium g4560 is impossible?Nobody mentioned insane stuttering in Pentium benchmarks.Does this mean that gtx 1050 ti and ryzen 1400 are a better combination than ryzen 1200 and rx 570?

[NicholasNicholasNicholas]

5 hours ago, MyName13 said:

So according to this gaming with Pentium g4560 is impossible?Nobody mentioned insane stuttering in Pentium benchmarks.Does this mean that gtx 1050 ti and ryzen 1400 are a better combination than ryzen 1200 and rx 570?

facepalm

no one said it was impossible, but its definitely not a good experience. 

and you cant generalize like that, it depends on the game. in some the 1200 + 570 would be better and in some the 1400 +ti would be better

[TheDankKoosh]

1 minute ago, Nickathom said:

 in some the 1200 + 570 would be better and in some the 1400 +ti would be better

Eh, I'd say the experience of overclocking something like the first system would be a better experience in nearly any game compared to that second system. An overclocked 1200 would perform like a 7600k at stock and then the 30% better gpu power would still be utilized well.

[NicholasNicholasNicholas]

Just now, TheDankKoosh said:

Eh, I'd say the experience of overclocking something like the first system would be a better experience in nearly any game compared to that second system. An overclocked 1200 would perform like a 7600k at stock and then the 30% better gpu power would still be utilized well.

an overclocked 1200 would not be that good. it would be like a 7500 at most

generally the 1400 is tied with the 7600k

[TheDankKoosh]

1 minute ago, Nickathom said:

an overclocked 1200 would not be that good. it would be like a 7500 at most

generally the 1400 is tied with the 7600k

Ryzen, at least from all the benchmarking that I've watched has about a 5% decrease in ipc from skylake/kabylake, so at this point it is just up to game devs to support ryzen, even if the game is a few years old. GTA 5 is notorious for example, to work like shit on ryzen.