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

[ivan134]

5 minutes ago, Majestic said:

Not sure yet, the freesync implementation of this panel isn't stellar. So I might just upgrade to a better Nvidia card and just deal with the tearing.

Which is not as bad on a 144hz panel asl it is on a 60hz panel.

 

Image quality of this panel is something else though. The contrast is sick.

Which driver? From time to time we get drivers that bork FreeSync.

[ivan134]

Double post

[Majestic]

5 minutes ago, ivan134 said:

Which driver? From time to time we get drivers that bork FreeSync.

The latest one.

[ivan134]

10 minutes ago, Majestic said:

The latest one.

So what's wrong? You still get tearing? What's the range btw?

[Majestic]

3 minutes ago, ivan134 said:

So what's wrong? You still get tearing? What's the range btw?

Well, there's where this topic got started  

 

It's range is fairly limited, 70-144hz. And now you see the crux of the issue, hitting 70-144hz (even when lowering settings) is not possible with the i5 in modern games. And LFC is dreadful tbh.

[ivan134]

2 minutes ago, Majestic said:

Well, there's where this topic got started  

 

It's range is fairly limited, 70-144hz. And now you see the crux of the issue, hitting 70-144hz (even when lowering settings) is not possible with the i5 in modern games. And LFC is dreadful tbh.

Jesus. Wow that is really bad. I didn't even think such a range existed.

[SkilledRebuilds]

I5 4690 user here. Decent but likely running out of superhigh fps legs already...

 

Im using a GTX970 OC fine, but rma'd a two day old used 980ti strix card with warranty.

When it ir a replacement arrives it'll be interesting.

But Im a fan of DSR so either way Im covered.

[Zodiark1593]

I wonder, since games are designed with many smaller cores in mind, would heavy context switching penalize the i5's to the point where they lose out greatly? If the threads are demanding as well, context switching penalties will certainly stall things. Nonetheless, I find it interesting that in such a short time, we went from i3s being capable gaming cpus, to Essentially leapfrogging i5's. I'm quite grateful that my gaming preferences don't consist of these "modern" AAA titles. 

 

One comparison I would have liked to see would be something like 8 Atom cores vs an i5, though now that I consider, the demands for 30 fps (consoles) and 60 are entirely different. 

[Majestic]

57 minutes ago, Zodiark1593 said:

the demands for 30 fps (consoles) and 60 are entirely different. 

This is likely the case. Consoles with high corecount on low-overhead API's aiming (in terms of drawcalls) for 30fps. And we pair less threads on a higher overhead API (DX11/12) at double the framerate with likely higher geometry resolution. Something's gotta give.

[Zodiark1593]

9 minutes ago, Majestic said:

This is likely the case. Consoles with high corecount on low-overhead API's aiming (in terms of drawcalls) for 30fps. And we pair less threads on a higher overhead API (DX11/12) at double the framerate with likely higher geometry resolution. Something's gotta give.

Exactly what I was thinking just now. It isn't as though the i5's are too weak. It's just that with the workload of juggling numerous threads, coupled with the demands for higher framerate and driver overhead, i5's can't function anywhere close to optimum efficiency. 

 

i7s, with Hyperthreading can put through another thread in unused functional units, but probably of greater relevance is that they can juggle twice as many threads without hitting any penalties for switching between threads. With the increased workload of pushing 60 frames and dealing with DX11 drivers, this probably makes a substantial difference. 

[Majestic]

@Zodiark1593

 

[xAcid9]

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

 

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

Spoiler

4C/4T - 73 fps average

Spoiler

 

[Majestic]

Sold that god awful 480 and got me a 1070. Expect a (rant) update soon.

[JoostinOnline]

10 minutes ago, Majestic said:

Sold that god awful 480 and got me a 1070. Expect a (rant) update soon.

Hopefully that's not how you advertised it.

[Majestic]

31 minutes ago, JoostinOnline said:

Hopefully that's not how you advertised it.

It wasn't the card, it was mostly the drivers. The card was...well...as good as any 480 will get. Having spent time on them personally, I can honestly say they suck balls (the drivers). And not in the "i have no issues" category, but mostly in frame delivery which doesn't show up in results, and a much bigger CPU overhead. OpenGL was also a disaster, meaning Dolphin ran like crap in games that required openGL (DX causes gltiches in some games).

[Citadelen]

I remember disagreeing with you quite a lot on this subject, since then I've see the error of my ways. I concede @Majestic, AMD has a terrible DX11 driver overhead.

 

I can't wait to ditch this god aweful FX-9590 dear god it's shit. My GPU is on its death bed as well so everything is a stuttery mess for me atm.

[Lare111]

I tried to play Watch Dogs 2 without hyper threading and there was way too much stuttering. Constant 60FPS was impossible in populated areas and I had 100% CPU usage. Even with hyper threading enabled I still get 90-100% CPU usage on my i7 6700 and I'm getting some bottlenecking. Ghost Recon: Wildlands with the same engine is also pretty CPU intensive.

 

Then I tried to play Witcher 3 without hyper threading and I got stuttering in Novigrad. In GTA V and BF1 MP I also had constant 100% CPU usage.

 

So if you wan't constant 60FPS I'd say i5 isn't enough.

[Clyne]

I'd agree with this, even playing games(WD2/ROTTR) with my i5-6500 and R9 Fury wasn't a pleasant experience. 

Currently using CL17 2448 12GB RAM, waiting for my CL15 3200 16GB RAM set to come it.

I'll probably do a performance difference between the two, as having faster ram would 

WD2 has noticable FPS issues with 100% CPU usage above the board, and it drops to low 40s when driving through the city. 

[sgloux3470]

First the multithreading came for the Pentium G3528 and it's soaring overclocks and IPC.

 

Second for the i3, once a champion of budget builds despite it's locked clockspeeds.

 

Third the i5, for while it had power and overclocking it was crippled by a lack of hyperthreading for the sake of market segmentation.

 

Even if the i5 edges out the 1400X/1500X/1600/1600X I think it's still more compelling to buy the AMD alternative simply because it has more cores/threads.  IPC is within spitting distance so it's simply an issue of clockspeed.  I think the smart move right now is a 1400X or the 1600. 

 

 

 

 

 

 

 

[Majestic]

2 hours ago, Bubblewhale said:

I'd agree with this, even playing games(WD2/ROTTR) with my i5-6500 and R9 Fury wasn't a pleasant experience. 

The problem is that CPU bottlenecking means bad framepacing. You really want to be GPU-bound. Right now with the Nvidia drivers i'm having a bit more headroom with the i5 and I can generally get the game to run GPU bound.

[xAcid9]

2 hours ago, Majestic said:

The problem is that CPU bottlenecking means bad framepacing. You really want to be GPU-bound. Right now with the Nvidia drivers i'm having a bit more headroom with the i5 and I can generally get the game to run GPU bound.

Can you include benchmarks with your new 1070 for comparison?

[Majestic]

27 minutes ago, xAcid9 said:

Can you include benchmarks with your new 1070 for comparison?

Will be filling in this added post

https://linustechtips.com/main/topic/749867-is-the-i5-still-enough-bottlenecking-explained-with-an-rx-480/?do=findComment&comment=9492663

 

Watch Dogs 2 is now filled in, make sure to copy link and open in new tab for max detail. I've added 60fps lines.

 

But take note, I have noticed something when testing, which I've explained here:

https://linustechtips.com/main/topic/757831-msi-afterburner-missing-interrupts/?do=findComment&comment=9579956

[Princess Luna]

Upsetting that youtube builders continues to mislead inexperienced users into expending a shit load of money for super higher end z270 board and water cooler for just an i5 7600k and even worse is that only then these users come to places like here on the forum to complain how their very expensive processor combo is hitting 100% on BF1 and making the game experience bad...

 

There is only so much we can do and honestly I get tired time to time repeating over and over how people should avoid over spending on an unlocked i5, because bad youtubers makes people biased.

[Majestic]

1 hour ago, Princess Cadence said:

Upsetting that youtube builders continues to mislead inexperienced users into expending a shit load of money for super higher end z270 board and water cooler for just an i5 7600k and even worse is that only then these users come to places like here on the forum to complain how their very expensive processor combo is hitting 100% on BF1 and making the game experience bad...

 

There is only so much we can do and honestly I get tired time to time repeating over and over how people should avoid over spending on an unlocked i5, because bad youtubers makes people biased.

Honestly? I think there are serveral issues with these youtubers. And not much of it is with nefarious intent.

 

A. They common saying "entrepeneurs work 80 hours a week to prevent working 40 hours for someone else" is the most predominant one. People like Jay from JayzTwoCentz, Brian from Tech (yes) City or Steve from Gamers Nexus are usually working a lot of time of the day on their content. Endlessly benchmarking or whatever it is they do. They rely mostly on the metrics to convey whether something is feasible or not. They usually don't play the games themselves for hours on end to notice the subjective experience is actually pretty bad. They also don't test all games, or when they do it's usually something easily replicated like a benchmark tool or single player. 

 

B. As demonstrated by my other post, youtubers who use software analysis instead of proper FCAT systems will have extremely unrealiable data. Afaik youtubers like Jay, Austin, Joker, Brian, Steve (hardware unboxed) all don't have FCAT systems. So whatever metrics they post are done with hardware polling, which can be influenced by several factors. From biased reporting to the more obvious missing or interrupts and averaging (100ms minimum for msi afterburner).

 

C. Most of them don't have science backgrounds or engineering degree's. Whilst some of them do a good job at interpreting data, whenever they run into a problem it's painfully obvious they can't narrow down a likely suspect. Their overall knowledge and technical knowhow is lacking. It's the reason Totalbiscuit probably stopped with his port report video's, even when I persuaded him via e-mail to for the love of god enable MSI Afterburner on-screen, he still didn't know how to recognize a 100% flat-out core in his Far Cry 4 analysis.

 

And then there is the elephant in the room, and something you can never prove; non-disclosed bias. Given the rough climate on youtube in terms of monetization it's not unfair to assume some people are willing to accept partnerships, deals or advertisement without disclosing it. This is however, in what now appears an apophasis, something I don't like dabbling in unless clear evidence is presented.

 

That said, Gamers Nexus has made the case in recent video's that the i5 is starting to buckle under the pressure.

 

[xAcid9]

I think most of them use PresentMon/OCAT or "good ol" FRAPS for frametime measurements.