CPU Questions

[Srnewbee]

I am not trying to start a fanboy war here. I have copule of questions.

1. I am just wondering why AMD threadrippers underperform their intel counterparts in gaming performance?

 

1. AMD has more cores

2. AMD has more threads

3. uses smaller fabrication

 

I am aware that per core to core intel is faster but its hard for me to believe that a single core performance of an AMD chip and Intel can be that vastly different. 

I think most games are not optimized for multiple cores and threads so AMD keeps on showing crappy benchmarks across the board.

 

2. My second question is, when do you think most games, meaning more than 50% of the titles, will be optimized for multiple cores?

3. My third question is, can intel really continue to compete with AMD when they are stuck at manufacturing 14nm dies? They are trying to pull a fast one by stating that they use the 10nm on some mobile canon lake processors but, lets be real. The most powerful consumer grade desktop processor from Intel is 14nm and they wont be rolling 10nm out until 2019 winter.

 

Looking forward to hearing your thoughts on this!

[Fasauceome]

AMD is worse than Intel because their microarchitecture is different. On a fundamental level, coffee lake is more effective than Zen. It allows for better use of the power being pushed into the CPU, and generally just results in more performance in every way.

 

it'll probably take awhile for games in the mainstream to take advantage of lots of cores, because lots of Gamers still only have quad cores due to budget restrictions. To phase out that level of Technology would also phase out a lot of their consumer base.

[Srnewbee]

would it be possible to build game engines that utilize scalable core use?

[Enderman]

They are completely different architectures. Just like riding a bus is different from riding a train.

More cores and threads are useless if the program can't use them.

99% of games use less than 8 threads.

Therefore the higher core performance in an intel CPU results in higher game fps.

[Fasauceome]

Just now, Srnewbee said:

would it be possible to build game engines that utilize scalable core use?

Possible but not popular at the moment.

[bleedblue]

More cores/threads != more fps. In fact the 7980XE which has 18 cores performs worse than the 4core 8 thread 7700k in games. Intel still has the superior single threaded performance which is more important in most games.

 

[Enderman]

Just now, Srnewbee said:

would it be possible to build game engines that utilize scalable core use?

This is difficult because a lot of stuff in a game happens sequentially and cannot be parallelized.

You can't calculate a bullet path before you press your mouse button to fire a gun.

 

Stuff like video rendering can be parallelized because it's easy to have many cores render each a small part of an image and then put the whole thing together at the end.

[Fasauceome]

Just now, bleedblue said:

More cores/threads != more fps. In fact the 7980XE which has 18 cores performs worse than the 4core 8 thread 7700k in games. Intel still has the superior single threaded performance which is more important in most games.

 

Side note, that 7980xe is no slouch. I tossed one into an Asus x299 deluxe board with a dark rock pro 4 and the thing made a 36% overclock no sweat. Pretty damn impressive for a thermal package that severe.

[xg32]

the main reason is just raw single core speed, amd with ddr3200 b-die even has around 10% better ipc than intel atm, but it doesnt make up the difference between 4.2ghz and 5.2ghz, it's not rocket science how amd is about 10% slower than intel in a decent amount of games, but that's where intel's advantage ends. 

 

If amd can somehow get to 4.7/4.8 intel would need 5.5~ to keep the lead, which is what the rumored speed of the 9700/9900k is looking to be.

[GoldenLag]

8 minutes ago, xg32 said:

If amd can somehow get to 4.7/4.8 intel would need 5.5~ to keep the lead, which is what the rumored speed of the 9700/9900k is looking to be

Update on the 14nm+++ speed seems that 5.4/5.5 is doable if the voltage is 1.5+. Meaning yes its possible, but the voltage will kill the CPU very quickly

 

 

Currently the only true advantage has is their node is amazing, and their ringbus core layout. 

 

Sadly ringbus scales badly with extra cores, and they may be loosing their node advantage. 

[Enderman]

35 minutes ago, GoldenLag said:

Update on the 14nm+++ speed seems that 5.4/5.5 is doable if the voltage is 1.5+. Meaning yes its possible, but the voltage will kill the CPU very quickly

 

 

Currently the only true advantage has is their node is amazing, and their ringbus core layout. 

 

Sadly ringbus scales badly with extra cores, and they may be loosing their node advantage. 

5GHz with maybe 1.2-1.3v will be great though.

[dgsddfgdfhgs]

the thing is that game developed based on intel cpu before ryzen was release. Sure this will benefit intel.

I dont play multiplayer online or stream, and I never see my ryzen 1700 usuage above 40% used. but I already see people reported their 8700k usage was too high

even people say intel is better on very high end gpu, you will never see the difference , like 110fps vs 120fps

[YakityYak]

I like intel because the name sounds funny hehehe

[GoldenLag]

7 minutes ago, Enderman said:

5GHz with maybe 1.2-1.3v will be great though.

considering 5-5,2ghz was about 1.38-1.45 im expecting about 1.35 volts for 5ghz, but thats purely speculation. the only difference between +++ and ++ is the boost speeds as far as i know.

[Quazie]

1 hour ago, bleedblue said:

More cores/threads != more fps. In fact the 7980XE which has 18 cores performs worse than the 4core 8 thread 7700k in games. Intel still has the superior single threaded performance which is more important in most games.

 

This is where I ask a simple question.

Why does this matter when multitudes of players do not just run the game, Third party com's, streaming, or even running music video on youtube while gaming.

These all put the computer into multitasking and load the CPU more, but why do reviewers never simulate these more real world scenarios?

 

If you are telling people one chips is better than another, then the tests run should reflect real world situations for a fairer analysis.

[Srnewbee]

16 minutes ago, Quazie said:

This is where I ask a simple question.

Why does this matter when multitudes of players do not just run the game, Third party com's, streaming, or even running music video on youtube while gaming.

These all put the computer into multitasking and load the CPU more, but why do reviewers never simulate these more real world scenarios?

 

If you are telling people one chips is better than another, then the tests run should reflect real world situations for a fairer analysis.

I dunno about you but I close all my background apps before I game. I think people have a different take on "real world analysis"

[Quazie]

9 minutes ago, Srnewbee said:

I dunno about you but I close all my background apps before I game. I think people have a different take on "real world analysis"

I use other apps at times, most of the time just com's.

But I would be interested in the effects of streaming while gaming on the CPU.

If it is just Intel vs Intel or AMD vs AMD that's not so bad, But Intel vs AMD which as was pointed out use different architectures it will favour single core every time.

A review should be unbiased and fair.

[dgsddfgdfhgs]

this is how biased and unrealistic for real world performance.

they are all tested with a single program/ game running, best GPU with AA and vsync OFF and claims intel better at gaming.....

 

[Quazie]

4 minutes ago, dgsddfgdfhgs said:

this is how biased and unrealistic for real world performance.

they are all tested with a single program/ game running, best GPU with AA and vsync OFF and claims intel better at gaming.....

 

Thank you my point exactly ..

[Stefan Payne]

3 hours ago, Srnewbee said:

1. I am just wondering why AMD threadrippers underperform their intel counterparts in gaming performance?

Lower Clockrate, boost clock and a bit higher latency.

 

But that is already beeing partly adressed with Zen+ and will be further improved with the 7nm Parts next year.

 

would it be possible to build game engines that utilize scalable core use?

 

 

TL;DR, ACHTUNG DEUTSCH:
https://www.planet3dnow.de/vbulletin/threads/431710-32-Kerne-fuer-Computerspiele-sind-kein-Problem-Dan-Baker-Oxide