Do I need more cores for games

[ThurraxKidd]

About to upgrade my gaming computer from my first gen core i5 dual core 650 to a ryzen based platform. But what I want to know is should I really shell out the extra cash for the six core r5 2600 or will I be fine grabbing the r3 2200g. Will there be any real use for the two extra cores of the ryzen 5 in gaming or is it like just a to have it kind of thing?

[KarathKasun]

You mean two cores and 6 extra threads.

 

SMT has a positive impact when core count is an issue, and some games can use 8+ threads at this point.

 

4C/8T is the MINIMUM I would get for a PC focused on gaming at this point in time.  Some games can still run into hitching and stuttering with that though.

[Eigenvektor]

That depends on the type of game and type of GPU you want to pair it with. Many games are starting to utilize more cores than just two or four and for a high-end GPU at least six or eight are recommended now.

 

Also, there's no i5 2600, but an i7 2600, which is very old. Did you mean an R5 2600?

[ThurraxKidd]

2 minutes ago, Eigenvektor said:

That depends on the type of game and type of GPU you want to pair it with. Many games are starting to utilize more cores than just two or four and for a high-end GPU at least six or eight are recommended now.

 

Also, there's no i5 2600, but an i7 2600, which is very old. Did you mean an R5 2600?

And as for the types of games. I intend to play some what newer games. My current steam library is mostly f2p games and games I bought that released in 2012 and 2013. But I intend to pair it with a 1080p gpu like a 570 or 580, possibly a 5500 when they release. I do have a preference for open world action adventure games as well as action rpgs. Would that still incite an r5 recommendation?

[KarathKasun]

10 minutes ago, ThurraxKidd said:

And as for the types of games. I intend to play some what newer games. My current steam library is mostly f2p games and games I bought that released in 2012 and 2013. But I intend to pair it with a 1080p gpu like a 570 or 580, possibly a 5500 when they release. I do have a preference for open world action adventure games as well as action rpgs. Would that still incite an r5 recommendation?

Dont get a G series CPU if you are getting a dedicated GPU.

 

Get an R5 2600 or 3600 at minimum.

[Fasauceome]

Might as well get the 2200G now while it's still capable and of course you have a solid upgrade path down the line for as many cores as you want, as long as you don't cheap out on the motherboard.

[dgsddfgdfhgs]

15 minutes ago, ThurraxKidd said:

I do have a preference for open world action adventure games as well as action rpgs. Would that still incite an r5 recommendation?

get 2600 together with a gpu

a 2200g isnt that much better than your old i5

[vanished]

4C/8T is budget territory these days, the absolute minimum I'd ever suggest.  The thing is many games, despite what LTT keeps saying, can actually use 6 or more cores, and the slower your cores are, the more apparent this becomes.  4C being budget means that they are also likely the slowest 4C you can get, and so for that reason going to 6 really does help

[SpiralTTGL]

since the push for more cores has become the new trend game engines and developers will be making use of them its def the way to go for future proofing your rig for next 3-5 years

[Drak3]

9 minutes ago, Ryan_Vickers said:

The thing is many games, despite what LTT keeps saying, can actually use 6 or more cores

There have been games that use more than 4 cores for awhile.

But for a long time, the benefit was minimal. For the games themselves, it still is outside of edge cases like Far Cry 5 (which does whatever the fuck it wants). The main perk is background tasks not interfering with the game.

[vanished]

18 minutes ago, Drak3 said:

There have been games that use more than 4 cores for awhile.

But for a long time, the benefit was minimal. For the games themselves, it still is outside of edge cases like Far Cry 5 (which does whatever the fuck it wants). The main perk is background tasks not interfering with the game.

Far Cry is an example of one of a small number of games that have not advanced enough to use more cores yet, but most have, and it's not just background tasks.  A proper system at idle uses maybe 2% CPU on average, not enough to make any meaningful impact on the game.  If you're infested with loads of malware and are idling with 1 or 2 whole cores pinned to 100% at all times sure, but that's not most people.  We saw this even with first gen Ryzen how the slower hex cores could keep up to or even meaningfully pass the faster quad cores, and it's been years since then, with more and more games being more and more multi-threaded.

 

You can also see here a huge lead for the 8700k vs the 7700k

 

Way more than can be explained by IPC or clockspeed differences alone.  Additional evidence for the benefit being due to core count and not other improvements is the few games shown here that don't show any benefit to the 8700k - obviously titles for which the 4 core limit still holds true.  This should eliminate any question of whether the previous example overstates the benefit of 6 cores due to comparing against only a 4/4 and not a 4/8.

 

It's true that having more cores will allow you to not be negatively impacted by significant background spikes like OneDrive, system updates, etc. but most of the time background tasks won't add up to anything meaningful, and these spikes can be controlled in other ways.  Most of all though this is not the only benefit.  These tests should show that performance is meaningfully better with core counts over 4, even 4C/8T, throughout gameplay, not just as a means to avoid rare dips.

[Flying Sausages]

For today gaming, 4 core is min and 6 core is normal for gaming.

[Drak3]

1 minute ago, Ryan_Vickers said:

Far Cry is an example of one of a small number of games that have not advanced enough to use more cores yet

That's not true. Most games don't really make much use of hexacores.

 

https://www.gamersnexus.net/guides/3423-intel-i7-7700k-revisit-benchmark-vs-9700k-2700-9900k

 

We see the humbled 7700K in roughly the middle of the pack of higher core count chips. The trend is that the chips that perform significantly better are those that clock higher, have seen refinements to memory subsystems, and have more cache.

 

21 minutes ago, Ryan_Vickers said:

A proper system at idle uses maybe 2% CPU on average

Both my 5930K and 9750H systems idle anywhere from 2 to 24%. The upper end being caused by Windows processes.

[dgsddfgdfhgs]

28 minutes ago, Ryan_Vickers said:

Far Cry is an example

ubisoft games shall never be used as any reference.

too well known for bad optimisation

 

 

 

[vanished]

6 minutes ago, Drak3 said:

That's not true. Most games don't really make much use of hexacores.

 

https://www.gamersnexus.net/guides/3423-intel-i7-7700k-revisit-benchmark-vs-9700k-2700-9900k

 

We see the humbled 7700K in roughly the middle of the pack of higher core count chips. The trend is that the chips that perform significantly better are those that clock higher, have seen refinements to memory subsystems, and have more cache.

 

Both my 5930K and 9750H systems idle anywhere from 2 to 24%. The upper end being caused by Windows processes.

They only tested 3 games - F1, Far Cry, and ACO.  We already knew about Far Cry so I won't bring that up again.  Looks like F1 is a similar story so you're right about it, but ACO seems to indeed like more cores

 

Looking at the 8700k vs 7700k both at 5 GHz, (well actually the 7700k has a 100 MHz advantage) the average is about 15% better and the lows (where it can really show) are 22% better.  Stepping up to the 9900k it looks like much the same story again, and we all know Intel didn't make that much of an improvement between generations to architecture, etc.

[KarathKasun]

6 minutes ago, Drak3 said:

That's not true. Most games don't really make much use of hexacores.

 

https://www.gamersnexus.net/guides/3423-intel-i7-7700k-revisit-benchmark-vs-9700k-2700-9900k

 

We see the humbled 7700K in roughly the middle of the pack of higher core count chips. The trend is that the chips that perform significantly better are those that clock higher, have seen refinements to memory subsystems, and have more cache.

 

Both my 5930K and 9750H systems idle anywhere from 2 to 24%. The upper end being caused by Windows processes.

Just going to point out that 7700k minimums are generally lower on those older games, which is where performance matters.

[Drak3]

2 minutes ago, Ryan_Vickers said:

but ACO seems to indeed like more cores

9600K

R7 2700

8600K

2700X

R7 1700 OC'd

R5 2600

R7 1700

R5 1600

 

Those all lost to the 7700K at stock. Those all have more cores.

Where they're lacking is cache performance. Not cores. The vast majority of games don't give a shit about hexacores, because games are highly serial in nature.

 

And are you honestly going to tell me that 100+ FPS is now considered the bare minimum?

 

 

 

BTW, according to this, the 5775C, a quadcore with moderate clockspeeds (3.7GHz TURBO) is too close to the 9900K given the clock difference and that ACO supposedly likes more than 4 cores.

https://www.gpucheck.com/game_gpu/assassins-creed-origins/nvidia-geforce-rtx-2080-ti/intel-core-i7-5775c-3-30ghz/low

[KarathKasun]

3 minutes ago, Drak3 said:

9600K

R7 2700

8600K

2700X

R7 1700 OC'd

R5 2600

R7 1700

R5 1600

 

Those all lost to the 7700K at stock. Those all have more cores.

Where they're lacking is cache performance. Not cores. The vast majority of games don't give a shit about hexacores, because games are highly serial in nature.

 

And are you honestly going to tell me that 100+ FPS is now considered the bare minimum?

 

 

 

BTW, according to this, the 5775C, a quadcore with moderate clockspeeds (3.7GHz TURBO) is too close to the 9900K given the clock difference and that ACO supposedly likes more than 4 cores.

https://www.gpucheck.com/game_gpu/assassins-creed-origins/nvidia-geforce-rtx-2080-ti/intel-core-i7-5775c-3-30ghz/low

Many have fewer threads or lower clocks, simple as that.  You cant compare apples and oranges.

[vanished]

5 minutes ago, Drak3 said:

9600K

R7 2700

8600K

2700X

R7 1700 OC'd

R5 2600

R7 1700

R5 1600

 

Those all lost to the 7700K at stock. Those all have more cores.

The 6/6 Intel chips are only "losing" by a small margin.  It's call it roughly even, which actually supports what I'm saying - you can trade clockspeed for core count, so long as the game can use that many cores, and the clockspeed loss is not too great, which brings me to the AMD chips, which btw have more cache like you say is the reason for the improvements - they're losing because they're too slow.

Quote

And are you honestly going to tell me that 100+ FPS is now considered the bare minimum?

No not at all.  No idea where you got that idea from.

Quote

BTW, according to this, the 5775C, a quadcore with moderate clockspeeds (3.7GHz TURBO) is too close to the 9900K given the clock difference and that ACO supposedly likes more than 4 cores.

https://www.gpucheck.com/game_gpu/assassins-creed-origins/nvidia-geforce-rtx-2080-ti/intel-core-i7-5775c-3-30ghz/low

I don't know what you're trying to show here, I can't see how or where that page is making a comparison between those chips

[Drak3]

3 minutes ago, KarathKasun said:

Many have fewer threads or lower clocks, simple as that.  You cant compare apples and oranges.

1. Most games don't see much performance uplift from multithreading.
2. Even at somewhat lower clockspeeds, scaling upwards with more cores will generate a performance uplift (hence why, despite lower clockspeeds, the octocore Ryzens were faster at various workstation workloads that scale with more cores than higher clockspeed chips like the 7700K).

1 minute ago, Ryan_Vickers said:

No not at all.  No idea where you got that idea from.

1 hour ago, Ryan_Vickers said:

4C/8T is budget territory these days, the absolute minimum I'd ever suggest.    4C being budget means that they are also likely the slowest 4C you can get, and so for that reason going to 6 really does help

 

2 minutes ago, Ryan_Vickers said:

I don't know what you're trying to show here, I can't see how or where that page is making a comparison between those chips

Scroll down. It lists 23 processors released last year, including the 9900K. It's at 225FPS with low settings, the 5775C is 207.

[KarathKasun]

2 minutes ago, Drak3 said:

1. Most games don't see much performance uplift from multithreading.
2. Even at somewhat lower clockspeeds, scaling upwards with more cores will generate a performance uplift (hence why, despite lower clockspeeds, the octocore Ryzens were faster at various workstation workloads that scale with more cores than higher clockspeed chips like the 7700K).

You do not get a linear increase in performance with multithreading unless its an embarrassingly parallel task.  Games still fall back on a single state management thread, meaning that you do gain performance from adding cores as long as you have some task that is still sharing a core with the primary thread.

 

Low settings is also not a proper CPU benchmark.  Low settings reduce the number of item states in flight in nearly all game engines (less items drawn and tracked), which makes the benchmark almost completely about performance in the primary state management thread instead of game performance for a given core count.

[GoldenLag]

1 hour ago, Ryan_Vickers said:

We saw this even with first gen Ryzen how the slower hex cores could keep up to or even meaningfully pass the faster quad cores, and it's been years since then, with more and more games being more and more multi-threaded.

you can also sort of see this in regarding 6c/6t CPUs vs 6c/12t where even tho the 6c/6t has traditionally been better, the lack of threads to play with just makes the ht parts stretch ahead

[vanished]

1 minute ago, Drak3 said:

Scroll down. It lists 23 processors released last year, including the 9900K. It's at 225FPS with low settings, the 5775C is 207.

I see now.  Well I don't know where these results are coming from but I'm reasonably sure they're total BS.  We can easily sanity check them against the GN results which I think everyone trusts.  GN got an average of 135.4 fps with the 9900k at stock running 1080p Medium on a 2080 Ti.  This site claims that same config should clock in at 179 fps.  Without a very good explanation for that I'd rather not use anything else from there, especially since we already have an extensive list of CPUs tested on that game, including 1% and 0.1% lows (which critically that other site has left out).

 

As for the quote, It's not like 4/8 is unusable for gaming, but I stand by what I said.  It's bottom of the stack now, with the vast majority of chips out there being at least 6/6 if not 12/24.  A CPU is somewhat of a long term investment given the cost and hassle of upgrading since it usually requires upgrading the motherboard and possibly RAM too, so coming in at the bottom of the stack, especially at a time when things are quickly moving upward seems to me like a bad idea.

 

I found some other videos and benchmarks and in terms of does 6 help, I'd say this:

• AotS Escalation: Yes
• Civ 6: No
• F1 2017: No
• BF1: Yes, though mind that you're not GPU bottlenecked.  That's easy to do so probably not much of a concern.

As for does 8 help over 6, I've not seen a strong sign for that.  As mentioned earlier ACO is perhaps showing a benefit but that's only one title and everything else I've seen recently I'd call a No for that comparison.

 

I'm more than happy to look into this cache vs cores situation and would like to see it proven one way or another.  I think for that we need to see two CPUs with largely similar specs, but where one has much more cache perform better in many games (or not, if that is the case).  Or if you can think of another way to test then let me know.

[Sakkura]

12 minutes ago, Ryan_Vickers said:

As for the quote, It's not like 4/8 is unusable for gaming, but I stand by what I said.  It's bottom of the stack now, with the vast majority of chips out there being at least 6/6 if not 12/24.  A CPU is somewhat of a long term investment given the cost and hassle of upgrading since it usually requires upgrading the motherboard and possibly RAM too, so coming in at the bottom of the stack, especially at a time when things are quickly moving upward seems to me like a bad idea.

Not to disagree, just want to point out a couple things: First, the difference between 4c8t and 6c6t generally isn't that big even in highly parallelized applications, as the extra threads compensate for the fewer cores. And second, the Ryzen 3 2200G that OP is actually looking at is a 4c4t chip rather than 4c8t. I'd really not want to be getting a 4c4t chip in this day and age.

[Drak3]

10 minutes ago, KarathKasun said:

ow settings is also not a proper CPU benchmark

Change the selection. You'll see that the roughly 20FPS difference is maintained across medium, high, and ultra.

 

Given that it's a 5775C, a moderate clocking quadcore running on stock DDR3, that shouldn't be possible if the assumption that ACO likes cores.

 

But that's not the case. Games don't care about your extra cores. Those are nigh worthless to the games. What they care about is cache, the ability to store more information in a faster storage medium to be able to bang out more computations in a serialized workload is king.

 

Your 9900K performs better because it has 16MB of L3 on tap and a significantly higher clockspeed. 8MB more than the 7700K.

On the other hand, the 9600K has a meager 9MB and a marginally higher clockspeed. They perform within margin of error.

And Zen's architecture renders it as effectively being 8+8MB of cache with a performance penalty that using more than 8MB of cache results in. That impacts gaming performance.

5 minutes ago, Ryan_Vickers said:

We can easily sanity check them against the GN results which I think everyone trusts.

I trust it for comparison against his other benchmarks. Same with GPU check.

 

For what I could actually expect in the real world, I don't trust either.