CPU w/ same benchmark running at higher utilization?

[caleb1]

I'm sort of a newbie and not really familiar with how CPUs work exactly, so this may be a dumb question.  But here it goes:

Old PC was a desktop with i7 6700 (4 cores, 8 threads, 8mb cache, 8 GT/s bus speed, 65W TDP, base 3.4 GHz max 4.0 GHz), 24 GB ram, GTX 745.

New PC is a laptop with i5 10300H (4 cores, 8 threads, 8mb cache, 8 GT/s bus speed, 45W TDP, base 2.5 GHz max 4.5 GHz), 16 GB ram, GTX 1650.

Ran a user benchmark test on both, they are very similar with the new i5 showing a few percent faster.

I only ever ran one game, Prepar3D, (its a flight simulator) and rarely did the CPU reach 100% with the older i7.

With the laptop, same game, same settings, same conditions, its pretty much always pegged at 100%.

Would anyone possibly be able to enlighten me as to why in the laptop, the CPU seems to be working way harder than it ever did in the desktop under the same conditions, even though both CPUs seem pretty much equivalent from the specs and benchmarks?

Is there more to determining how good a CPU is besides the specs, GHz, and benchmark scores?

Also should note that the CPU temps on the laptop are much higher than the desktop, I can see how this could slow it down but I don't know how higher temps could increase utilization.  Unless I'm not understanding it correctly.

 

Thanks.

[RONOTHAN##]

It's because your GPU is a lot faster. Before, since the GT745 is a very low end GPU, especially by modern standards, your GPU was the limiting factor and your CPU was waiting for it before doing anything. Now, since your GPU is decent, your CPU has to actually work to keep up with it, and since flight sim games are usually CPU bound, it's pegging the CPU. 

 

Temps are higher because it's a laptop. They are designed to run hot, so it probably won't slow it down. 

[caleb1]

36 minutes ago, RONOTHAN## said:

It's because your GPU is a lot faster. Before, since the GT745 is a very low end GPU, especially by modern standards, your GPU was the limiting factor and your CPU was waiting for it before doing anything. Now, since your GPU is decent, your CPU has to actually work to keep up with it, and since flight sim games are usually CPU bound, it's pegging the CPU. 

 

Temps are higher because it's a laptop. They are designed to run hot, so it probably won't slow it down. 

 

Thanks, that would make sense.  CPU cant keep up with GPU.

[caleb1]

On 9/8/2021 at 4:45 PM, RONOTHAN## said:

It's because your GPU is a lot faster. Before, since the GT745 is a very low end GPU, especially by modern standards, your GPU was the limiting factor and your CPU was waiting for it before doing anything. Now, since your GPU is decent, your CPU has to actually work to keep up with it, and since flight sim games are usually CPU bound, it's pegging the CPU. 

 

Temps are higher because it's a laptop. They are designed to run hot, so it probably won't slow it down. 

Another question I thought I'd ask....

Would anyone know what CPU would be good to use with the GTX 1650, to keep the same CPU/GPU performance ratio as the GTX 745 to the i7 6700?

Would it need to be a 6 core or 8 core rather than quad core?

[RONOTHAN##]

The CPU I'd recommend pairing with a 1650 is an i5 11400 or Ryzen 3600. However, that isn't the same performance ratio as before. With how powerful the 6700 is compared to the GT745, you'd probably need something along the lines of a 5800X. I'm not saying I recommend overspending that much on the CPU, but that's probably how high you'd need to go. 

 

As for getting a 6 or 8 core CPU instead of a quad core, while quad core chips are still fine for a lot of stuff, higher core-count chips offer a lot more flexibility and aren't that much more expensive than their relative quad cores, to the point I wouldn't recommend picking up a quad core for anything but the most budget of builds.

[caleb1]

2 hours ago, RONOTHAN## said:

The CPU I'd recommend pairing with a 1650 is an i5 11400 or Ryzen 3600. However, that isn't the same performance ratio as before. With how powerful the 6700 is compared to the GT745, you'd probably need something along the lines of a 5800X. I'm not saying I recommend overspending that much on the CPU, but that's probably how high you'd need to go. 

 

As for getting a 6 or 8 core CPU instead of a quad core, while quad core chips are still fine for a lot of stuff, higher core-count chips offer a lot more flexibility and aren't that much more expensive than their relative quad cores, to the point I wouldn't recommend picking up a quad core for anything but the most budget of builds.

That's interesting, I would need a 5800X with the GTX 1650 to achieve the same CPU usage (85 to 95 percent) as my i7 6700 did with the GTX 745? Am I interpreting that correctly?  Or would overall usage drop proportionately to number of cores I add regardless of effective speed?  Like 95-100% usage with 4 cores would be 75% with 6 or 50% with 8?  Even if they have similar benchmarks?

With CPUs, is it not so much the benchmark that measures how much of a bottleneck there is as much as the core quantity?  As most CPUs I have compared on userbenchmark are rarely more than 40 to 50% difference in effective speed, even if the number of cores is double.

Using the two you mentioned for example, a 5800X is only about 3% faster than a i5 11400 in effective speed on userbenchmark....  But it does have 8 cores vs 6...

Versus with GPUs often times some are better than others by hundreds of %s.

Thanks for helping a new pc enthusiast become better educated! Really appreciate it.

[RONOTHAN##]

24 minutes ago, caleb1 said:

5800X is only about 3% faster than a i5 11400 in effective speed on userbenchmark

Userbenchmark is a joke. Never use it for anything besides entertainment (their review of the 11900k is pretty hilarious). In reality the 5800x is a lot faster than the 11400. They have a huge bias towards Intel CPUs, i.e. they say a 11400 is faster than a 5600X, when if you head to basically every other review site the 5600X is a decent amount faster.

 

If you want to know more about how bad Userbenchmark is, watch this video:

29 minutes ago, caleb1 said:

Like 95-100% usage with 4 cores would be 75% with 6 or 50% with 8?

It really depends on the game. Some games do scale like that with cores, but not many. I don't have much experience with Prepar3D, so I can't comment on it, but usually Flight Sims are a bit more multi threaded, and so I'd expect it to actually use the other cores to some extent. What you said would be the worst case scenario (assuming it has the same clock speed, IPC, cache setup, etc.), with 4 of the cores being maxed out and nothing else being used. If the game is more multithreaded, the other cores will be used, and while it's unlikely you'd reach 100% usage on an 8 core chip, you might hit closer to 70-80%. 

 

Basically, it depends.

[caleb1]

1 minute ago, RONOTHAN## said:

Userbenchmark is a joke. Never use it for anything besides entertainment (their review of the 11900k is pretty hilarious). In reality the 5800x is a lot faster than the 11400. They have a huge bias towards Intel CPUs, i.e. they say a 11400 is faster than a 5600X, when if you head to basically every other review site the 5600X is a decent amount faster.

 

If you want to know more about how bad Userbenchmark is, watch this video:

It really depends on the game. Some games do scale like that with cores, but not many. I don't have much experience with Prepar3D, so I can't comment on it, but usually Flight Sims are a bit more multi threaded, and so I'd expect it to actually use the other cores to some extent. What you said would be the worst case scenario (assuming it has the same clock speed, IPC, cache setup, etc.), with 4 of the cores being maxed out and nothing else being used. If the game is more multithreaded, the other cores will be used, and while it's unlikely you'd reach 100% usage on an 8 core chip, you might hit closer to 70-80%. 

 

Basically, it depends.

Thanks for the reply.

I was curious because since you were saying a nearly top of the line CPU would get me the same ratio with a GTX 1650 as my i7 6700 did with a GTX 745, I was contemplating weather or not it a better GPU than this would ever be beneficial.  For example if I need a top of the line CPU to keep up with a GTX 1650 using this particular game, would getting another PC someday with a top of the line GPU be better even if I had a top of the line CPU as well?  Or is there a certain point where I am going to be CPU limited regardless of how good of a GPU I have.  But I think you answered my question by saying that having more cores will tremendously decrease utilization thereby increasing performance, as long as the game has multi-core utilization distribution support, which Prepar3D does.

So I think my answer is for the next pc to go 8 cores, then by performance won't really be CPU limited.

Thanks for your time.

[RONOTHAN##]

3 minutes ago, caleb1 said:

I was curious because since you were saying a nearly top of the line CPU would get me the same ratio with a GTX 1650 as my i7 6700 did with a GTX 745, I was contemplating weather or not it a better GPU than this would ever be beneficial

Part of the reason I said such a high end CPU is that was basically the ratio you had with your previous system, a next to top end gaming CPU with a low end GPU. For playing basically any other game than Prepar3D, it doesn't make sense to pay so much for a CPU that won't increase your performance. Plus, most other games are GPU bound rather than CPU bound, so in most other things a better GPU would come in handy.

11 minutes ago, caleb1 said:

But I think you answered my question by saying that having more cores will tremendously decrease utilization thereby increasing performance

That's not exactly what I said. Total CPU usage would go down yes, but per core usage would stay exactly where it was, and game performance would also stay the same. If the game is more multi threaded, yes performance would go up. 

 

Plus, you don't want utilization to be lower, you want it to be balanced between the CPU and the GPU. in a perfect world, they would both be at 100% usage. In reality, you're going to shoot for the GPU being at 97-98% and the CPU being somewhere above 70%. You want to monitor the GPU usage and see how that is running, and pick a CPU according to that. 

[caleb1]

3 minutes ago, RONOTHAN## said:

Part of the reason I said such a high end CPU is that was basically the ratio you had with your previous system, a next to top end gaming CPU with a low end GPU. For playing basically any other game than Prepar3D, it doesn't make sense to pay so much for a CPU that won't increase your performance. Plus, most other games are GPU bound rather than CPU bound, so in most other things a better GPU would come in handy.

That's not exactly what I said. Total CPU usage would go down yes, but per core usage would stay exactly where it was, and game performance would also stay the same. If the game is more multi threaded, yes performance would go up. 

 

Plus, you don't want utilization to be lower, you want it to be balanced between the CPU and the GPU. in a perfect world, they would both be at 100% usage. In reality, you're going to shoot for the GPU being at 97-98% and the CPU being somewhere above 70%. You want to monitor the GPU usage and see how that is running, and pick a CPU according to that. 

I understand a little better now.  With the old PC, I had that perfect balance it seemed.  CPU was between 70% and 100%, and GPU was often at 100%.

Now with the new laptop, CPU is pretty much always at 100% and GPU was around 30%, until I turned on some GPU taxing settings  (shadows, reflections, both of which I always had off on previous PC due to being GPU bound) and then the GPU went to 97-98 and I got slightly better FPS.

Could the fact GPU usage was low before I turned on GPU taxing settings indicate CPU being the bottleneck?

Either way I'll probably just need to tweak the settings to optimize CPU for now, but I am trying to gather knowledge for things to think about when I upgrade.

Specifically how good of a GPU I would need before I am CPU limited.  (like as an example, is anything better than a 1080ti useless given CPU limitations).

[RONOTHAN##]

2 minutes ago, caleb1 said:

Could the fact GPU usage was low before I turned on GPU taxing settings indicate CPU being the bottleneck?

 

In a word, yes. 

2 minutes ago, caleb1 said:

Specifically how good of a GPU I would need before I am CPU limited.

It really depends on when you go to upgrade. New CPUs are going to be coming out in the next couple months, and we don't know how much of a performance increase they're going to be, with some rumors saying a 5% increase and some saying closer to 15-20%. I probably get something like a 3080 either way, but by the time you upgrade, a 1080 could be your best bet, or a 2080 Ti could be your best bet. Plus, you might get into other games then, and while this one particular game is very CPU bound, the other games probably wouldn't be, and so getting an even higher end GPU might be a better call.

[caleb1]

2 minutes ago, RONOTHAN## said:

In a word, yes. 

It really depends on when you go to upgrade. New CPUs are going to be coming out in the next couple months, and we don't know how much of a performance increase they're going to be, with some rumors saying a 5% increase and some saying closer to 15-20%. I probably get something like a 3080 either way, but by the time you upgrade, a 1080 could be your best bet, or a 2080 Ti could be your best bet. Plus, you might get into other games then, and while this one particular game is very CPU bound, the other games probably wouldn't be, and so getting an even higher end GPU might be a better call.

Good things to think about.

And it doesn't hurt to have forward compatibility for when software technology advances as well, like you said.

Thanks for all the help!