Jump to content

CPU Performance Database


Recommended Posts

Theses are exactly my concerns ^^ Upgrade now, with Skylake MotherBoard CPU et DDR4 Ram ? Or wait until further generation, since I just want to upgrade my KSP performance, which should easily be possible with something like a 3770, still very powerful ? Waiting 2 years could be a good idea, in order to get DDR4 when it's fully operationnal et affordable.

So... For 250€, what 1155 CPU to buy, specifically for KSP ? Maybe there is some particular reference, dealing with this kinda-monocore thing... Even with Unity 5, the multithread will only allow one core per vessel, soooo... Dunno, if you have advices, i'd like to learn about it !

Link to comment
Share on other sites

Theses are exactly my concerns ^^ Upgrade now, with Skylake MotherBoard CPU et DDR4 Ram ? Or wait until further generation, since I just want to upgrade my KSP performance, which should easily be possible with something like a 3770, still very powerful ? Waiting 2 years could be a good idea, in order to get DDR4 when it's fully operationnal et affordable.

So... For 250€, what 1155 CPU to buy, specifically for KSP ? Maybe there is some particular reference, dealing with this kinda-monocore thing... Even with Unity 5, the multithread will only allow one core per vessel, soooo... Dunno, if you have advices, i'd like to learn about it !

A quick check of Amazon and Newegg shows only overpriced and refurbished CPUs for the 1155. So it's a bit hard to recommend buying something that can only be found overpriced or used.

If you can find a reputable place (like HardForum, but you can't just jump in and start buying from there, which is probably why it remains rebutable) to buy from, then a used CPU isn't so bad, but I would be nervous about buying hardware from something like Ebay.

DDR4 isn't really that expensive, it's pretty easy to find 8GB for $60-70, which isn't much more than DDR3. I'm not sure about motherboards, but there is probably no way to spend less than $350-400 or so for a CPU/MB/RAM combo, even if you wait a few months for the non-K CPUs and some more mid-range MBs.

Link to comment
Share on other sites

Okay, thanks, I totally forgot to say that I was, indeed, okay to buy a used CPU, it could even be a way to reach some high-end CPU at reasonable cost, in order to wait a year or two and change the whole configuration without losing lot of money.

I'm gonna see what's available and what could be reliable, in France, for such used CPU. Thanks for the answer and sorry for the slight off-topic !

Link to comment
Share on other sites

Kind of off topic, but kind of not. Some of us can't afford "the latest and greatest" systems.. I've just recently made an "upgrade" from a 4.4 ghz i7-920 to a 3.6 ghz i3-2130, with plans to go up to a 3770K some time next year. I've noticed a significant boost (nearly +60%) in KSP performance vs my ollllld 2008 i7 system, even with lower clock speeds on this newer platform.

Just thought I would note that out there to others.

Link to comment
Share on other sites

  • 4 months later...

I just purchased the Core I7 6700k. Eagerly awaiting its arrival. It will find a home in an Asus Ranger VIII Mobo with 16gb of 2133 ddr4 ram.

SOO EFFING EXCITED

Coming from a core I7 920 oc'd to 3.3.

Has anyone tried to run this benchmark test in 1.0.5 yet? 

 

Link to comment
Share on other sites

It doesn't work in 1.0; changes made to how parts unload while in the atmosphere make it so the majority of the flight occurs with all of the parts still loaded.

I'm waiting for 1.1 to make a new test craft and deal with the part unloading problem.

Link to comment
Share on other sites

12 hours ago, DMagic said:

It doesn't work in 1.0; changes made to how parts unload while in the atmosphere make it so the majority of the flight occurs with all of the parts still loaded.

I'm waiting for 1.1 to make a new test craft and deal with the part unloading problem.

Well it *works*, the rocket flies stable-ish straight up though a couple of stages want swapping (9 and 10 I think), and the results are informative for comparing different game settings and mods. But they aren't directly comparable with 0.90 results.

Though maybe they should be plotted anyway. The increased unload distance is one of the factors that degrades performance in 1.0 for "real" launches after all.

Link to comment
Share on other sites

6 hours ago, cantab said:

The increased unload distance is one of the factors that degrades performance in 1.0 for "real" launches after all.

Indeed, what matters to most players is how well (or not) the game runs, irrespective of why. If increasing the unload distance tanks performance, it's only fair to show this.

On 3/08/2015 at 11:48 AM, DMagic said:

I reverted the vessel unload/unpack ranges to their 0.90 values and ran the test again.

Mind sharing how?

Link to comment
Share on other sites

14 hours ago, cantab said:

Well it *works*, the rocket flies stable-ish straight up though a couple of stages want swapping (9 and 10 I think), and the results are informative for comparing different game settings and mods. But they aren't directly comparable with 0.90 results.

Though maybe they should be plotted anyway. The increased unload distance is one of the factors that degrades performance in 1.0 for "real" launches after all.

Well, yes. But you could really use any big vessel to do that.

7 hours ago, scribbleheli said:

What was the windows 10 friendly FPS logger? Fraps?

FRAPs works for Windows, I don't think I've tried it in 10, but it should still work.

7 hours ago, steve_v said:

Indeed, what matters to most players is how well (or not) the game runs, irrespective of why. If increasing the unload distance tanks performance, it's only fair to show this.

Mind sharing how?

The test is to show how performance varies with part count, with the new unload distances most of the run occurs with almost all of the parts, then most of them unload all at once, which isn't as useful.

 

I haven't taken the time to go through this and figure out again how to set the unload distances to be exactly the same as pre 1.0 (I think the "beta ranges" values should be the same as before), and there is a bunch of stuff that keeps this limited to specific vessels, but it's fairly simple:

using System;
using System.Collections;
using System.Collections.Generic;
using UnityEngine;

namespace CPU_Test
{

	[KSPAddon(KSPAddon.Startup.Flight, false)]
	public class CPULogger : MonoBehaviour
	{
		private const string vesselName = "CPU Bunga Bunga Plane";
		private const string vesselNamePlane = "CPU Bunga Bunga Plane Ship";
		private Vessel mainVessel;
		private VesselRanges.Situation flyingRanges = new VesselRanges.Situation(200, 250, 750, 200);
		private VesselRanges.Situation defaultRanges = new VesselRanges.Situation(200, 250, 50, 40);
		private VesselRanges.Situation suborbitalRanges = new VesselRanges.Situation(200, 500, 250, 40);
		private VesselRanges.Situation planeRanges = new VesselRanges.Situation(2250, 22500, 25000, 200);
		private VesselRanges.Situation betaRanges = new VesselRanges.Situation(2250, 2500, 5000, 200);
		private VesselRanges.Situation betaLandedRanges = new VesselRanges.Situation(2250, 2500, 350, 200);

		private void Start()
		{
			if (FlightGlobals.ActiveVessel.vesselName != vesselName)
				Destroy(this);

			initialize();
		}

		private void initialize()
		{
			print("[CPU Logger] Starting Up CPU Test Run...");

			GameEvents.onNewVesselCreated.Add(vesselCreated);
			GameEvents.onVesselCreate.Add(vesselCreated);

			mainVessel = FlightGlobals.ActiveVessel;

			mainVessel.vesselRanges = newRanges(true);
		}

		private void OnDestroy()
		{
			GameEvents.onNewVesselCreated.Remove(vesselCreated);
			GameEvents.onVesselCreate.Remove(vesselCreated);
		}

		private VesselRanges newRanges(bool plane)
		{
			VesselRanges range = new VesselRanges();

			VesselRanges.Situation flight = plane ? planeRanges : flyingRanges;

			range.flying = flight;
			range.subOrbital = suborbitalRanges;
			range.landed = defaultRanges;
			range.escaping = defaultRanges;
			range.orbit = defaultRanges;
			range.prelaunch = defaultRanges;
			range.splashed = defaultRanges;

			return range;
		}

		private void vesselCreated(Vessel v)
		{
			StartCoroutine(unloadRange(v));
		}

		IEnumerator unloadRange(Vessel V)
		{
			int timer = 0;

			while (timer < 20)
			{
				timer++;
				yield return null;
			}

			if (V.vesselName == vesselNamePlane)
			{
				V.vesselRanges = newRanges(true);
				print("[CPU Logger] New Plane Created - Setting Plane Unload Ranges : [" + V.vesselName + "]");
			}
			else
			{
				V.vesselRanges = newRanges(false);
				print("[CPU Logger] New Vessel Created - Setting Debris Unload Ranges : [" + V.vesselName + "]");
			}
		}
	}
}

 

Link to comment
Share on other sites

Well not a comparable test, But I finished my new build, loaded up KSP and the test rocket. And saw double the performance.

Since the unload distances are all super huge, I just visually watched my FPS as the rocket launched. With the Core i7 920 I got about 8-9 FPS at launch. With  the  6700k it was 18-19 ish until  things unloaded then just climbed to my 180FPS limit very rapidly.  

So that's pretty sweet.

Link to comment
Share on other sites

  • 2 months later...

I will update properly tomorrow; many changes are needed for this to be a more useful benchmark.

Don't submit results for now.

@sparkybear I'll check these results and add a few of my own when I get a chance.

Edited by DMagic
Link to comment
Share on other sites

I also ran the test just for fun. 

I was getting same frame rates as last time. little higher at 22-25 FPS at launch.

What was amazing is that It was in green physics time scale with a .4 delta t slider.

 

Completely uneventful all the way to orbit.

Edited by scribbleheli
Link to comment
Share on other sites

OK I have a few initial results with the old CPU rocket. All tests were run with the vessel unload distances set to their pre-1.0 levels (2.25km, instead of 22.5km in flight and 15km in suborbital):

DpfJHNU.png

There are a few things of interest here.

The first is the 1.0.5 to 1.1 x64 comparison, in blue and yellow. There is a significant increase in performance.

Another is that there is a significant difference between the 32- and 64-bit versions, in orange and yellow.

And I swear that the results in green are what I got when I ran it in 0.25.

I'll have to run it again when the final version is released, but that result is a little disappointing. It's likely that this rocket doesn't accurately show some of the performance increases, since the debris are quickly unloaded and it's mostly just one vessel. My new CPU rocket might give more representative results when it is ready. If anything this is just an indicator of how badly performance has degraded in pre-1.1 versions. :(

I should have some results on a weaker machine later today, too.

Link to comment
Share on other sites

Here are some more results from a lower powered machine.

The full specs are:

Surface Pro 4 - m3 6Y30 - 2.2GHz | 4GB DDR3L RAM | HD 515

alCPqqL.png

Results for both 1.1 and 1.0.5 include stock and pre-1.0 unload distances. All 1.1 tests were run in the x64 version.

The results here show a much bigger relative improvement. It goes from essentially unplayable in 1.0.5 to sort-of reasonable in 1.1.

 

I have two ideas in mind for new 1.1 tests. One is the new rocket that was shown a little earlier in the thread. It should better demonstrate improvements when multiple vessels are loaded. Another is to put a station in orbit where it would be away from PQS lag, and away from any aero calculations. The station could just fire off several vessels with varying part counts over the course of the test.

These will all have to wait until the full release though.

Edited by DMagic
Link to comment
Share on other sites

I'm leaning towards using a station, rather than launching a rocket, for the next version of the test vessel.

I think launches of big rockets are always going to be slow, and they are slow for many reasons, not just part physics. And since they basically consist of one big vessel continuously dropping off a few parts at a time, I don't think they will be the best indicator for how CPUs perform in KSP 1.1+. I also don't like the idea of needing a plugin to make the debris unload faster. I'll probably get the new version of the launch rocket working, but just keep it for myself for testing performance between different KSP versions.

For the actual CPU test, comparing performance of different CPUs, rather than different versions of KSP, I'm thinking I'll make a big station. Something like a 300-400 part station core. Then launch 8 small vessels, then 4 medium vessels, then 2 large, and finally 1 very large vessel. I'm thinking each set of vessels should have the same total part count.

It would be setup so you that all of the engines would be off, then you throttle up, and stage one set at a time. Then just wait for each set of vessels to get to the unload distance and fire off the next set. It should be simple and it should avoid other performance impacting elements, like aero effects, drag calculations, PQS ocean lag, and so on.

Link to comment
Share on other sites

This thread is quite old. Please consider starting a new thread rather than reviving this one.

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

 Share

×
×
  • Create New...