Best value "Entry Level" gaming PC.

[Seba]

That is a really good point you are making about the 720p TV considering an iGPU machine is most likely to be used as a gaming HTPC.

What I think I will probably end up doing is test that resolution in cases where increased detail settings (or graphically demanding games) are causing a strong frame rate drop off between 1024 x 768 and 1280 x 1024.

1280 x 1024: 67% more pixels than 1024 x 768
1280 x 720 (aka 720p): 17% more pixels than 1024 x 768

Most 720p LCD TVs ("HD Ready") are 1366x768. So adjust your tests at least for that resolution (if not for Full HD/1920x1080).

 

[Insert_Nickname]

I would definitely expect Haswell Celeron to be stronger than either AM1 or FM2 or FM2+ dual core at low resolutions.

As resolution and/or detail settings rise, a FM2 or FM2 dual core with a strong enough GPU should start to win at some point. However, with that mentioned I have two questions:

1. Does AMD actually provide enough GPU on their low end FM2 or FM2+ SKUs to allow this to happen?

2. If a FM2 or FM2+ dual core does surpass a Haswell Celeron as resolution and/or detail settings increase does it happen with a playable frame rate? (Example: I have seen cases of AMD APU beating Intel processors in gaming benchmarks.....but only at levels of GPU stress that result in both processors being unplayable. While this is interesting data for academic reasons, it says little for practical usage utility)

Xbitlabs did a comparison of Haswell and Trinity/Richland, the 192SP's (5400K/6400K) are about equal to Haswell GT2. At and above 256SP Trinity/Richland wins hands-down. I'd expect Kaveri to behave quite similarly, but a bit faster due to new architecture (GCN). And yes, this happens at playable framerates (~30) even at 1366x768 at medium quality. If you turn down a notch to 720p and lower quality, it would properly stay above 30.

http://www.xbitlabs.com/articles/cpu/display/amd-richland-intel-haswell-gpu.html

 

[Seba]

And if you pair a Pentium Haswell with a GTX 750, then you get playable framerates at 1920x1080. Even better with the slightly more expensive GTX 750 Ti.

 

[NTMBK]

And if you pair a Pentium Haswell with a GTX 750, then you get playable framerates at 1920x1080. Even better with the slightly more expensive GTX 750 Ti.

Spending more money gets higher framerates? Shocker.

 

[cbn]

Xbitlabs did a comparison of Haswell and Trinity/Richland, the 192SP's (5400K/6400K) are about equal to Haswell GT2. At and above 256SP Trinity/Richland wins hands-down. I'd expect Kaveri to behave quite similarly, but a bit faster due to new architecture (GCN). And yes, this happens at playable framerates (~30) even at 1366x768 at medium quality. If you turn down a notch to 720p and lower quality, it would properly stay above 30.

http://www.xbitlabs.com/articles/cpu/display/amd-richland-intel-haswell-gpu.html

Here are the gaming results from that article:

(results continued in next post)

 

[cbn]

 

[cbn]

Looking at those results, A6-5400K and A6-6400K are non-playable in Battlefield 3, Bioshock Infinite, Crysis 3, Metro Last Light and Tomb Raider at the resolutions tested.

In Grid 2, A6-5400K and A6-6400k achieve frame rates in the mid 30's at 1366 x 768 High Quality and low 30's at 1920 x 1080p Medium Quality. But is this playable considering it is a racing game?*

In War Thunder, both A6-5400K and A6-6400K achieve playability at 1366 x 768 High Quality. But at 1920 x 1080p Medium quality, both processors just barely fail to reach playability.

Based on the 30 FPS cut off rule, the resolutions used are too high for the AMD dual cores (with the exception of War Thunder and GRID 2)

* I need to revisit the racing games to see how bad 30 FPS is with that type of gameplay. My suspicion is that it should probably follow (or at least attempt to follow) the 60 FPS rule like the First person shooters do.

 

[Seba]

Spending more money gets higher framerates? Shocker.

Gets playable framerates. Notice the difference?

 

[ytsoc]

Hello everybody. Sorry to barge in, but i'm not sure about a decision i need to make.

Basically i need to replace a dying pc , so the new components must be cheap.
Light gaming(full hd monitor) meaning older games (2010 era) but mostly web browsing and movies.

So my choices are: intel G1820 and amd A4-6300/7300. I'll be relying on their IGPU's or maybe i can find an old cheap nvidia 8600GT for the intel build(i can find an used cpu and spend the extra on the 8600gt)

I'm unsure on what i should chose. Will the intel+8600gt be better than the A4-7300 with it's igp?

Thanks

 

[cbn]

Hello everybody. Sorry to barge in, but i'm not sure about a decision i need to make.

Basically i need to replace a dying pc , so the new components must be cheap.
Light gaming(full hd monitor) meaning older games (2010 era) but mostly web browsing and movies.

So my choices are: intel G1820 and amd A4-6300/7300. I'll be relying on their IGPU's or maybe i can find an old cheap nvidia 8600GT for the intel build(i can find an used cpu and spend the extra on the 8600gt)

I'm unsure on what i should chose. Will the intel+8600gt be better than the A4-7300 with it's igp?

Thanks

If you are choosing a Haswell Celeron, that 8600 GT is not that much of a gain (even if it is the GDDR3 version):

http://www.tomshardware.com/reviews/gaming-graphics-card-review,3107-7.html

According to Tom's GPU heirarchy chart, 8600 GT GDDR3 is only equal to Intel HD 4000 (Gen 7, 16 EUs.) This compared to Haswell Celeron's iGPU (Gen 7.5, 10 EUs).

So take this into consideration when determining how much you will pay for the 8600GT.

 

[ytsoc]

so the amd apu will beat the intel with or without the 8600gt.
I guess i'll go with the A4-7300

 

[cbn]

That is a really good point you are making about the 720p TV considering an iGPU machine is most likely to be used as a gaming HTPC.

What I think I will probably end up doing is test that resolution in cases where increased detail settings (or graphically demanding games) are causing a strong frame rate drop off between 1024 x 768 and 1280 x 1024.

1280 x 1024: 67% more pixels than 1024 x 768
1280 x 720 (aka 720p): 17% more pixels than 1024 x 768

Most 720p LCD TVs ("HD Ready") are 1366x768. So adjust your tests at least for that resolution (if not for Full HD/1920x1080).

Thanks for bringing that up.

Here is the comparison once more, with 1366 x 768 added:

1280 x 1024: 67% more pixels than 1024 x 768
1366 x 768: 33% more pixels than 1024 x 768
1280 x 720 (aka 720p): 17% more pixels than 1024 x 768

 

[Yuriman]

so the amd apu will beat the intel with or without the 8600gt.
I guess i'll go with the A4-7300

If they're the same price, I'd go with the Celeron. AMD's graphics are faster on the A4, but the CPU performance difference is even bigger in favor of the Celeron.

Depends on the game, though.

 

[frozentundra123456]

If they're the same price, I'd go with the Celeron. AMD's graphics are faster on the A4, but the CPU performance difference is even bigger in favor of the Celeron.

Depends on the game, though.

I agree. I would not even consider a single module APU. Cpu performance is just too anemic. I would either go for an Athlon X4 or pentium with a discrete card or at least a two module APU like the A8-7600, which is decently priced.

 

[cbn]

I agree. I would not even consider a single module APU. Cpu performance is just too anemic. I would either go for an Athlon X4 or pentium with a discrete card or at least a two module APU like the A8-7600, which is decently priced.

But see A8-7600 or Athlon X4/Pentium + discrete card is not a fair comparison to a $40 A4-7300 or $43 Celeron G1820.

If the person just wants a cheap processor for a spare parts build secondary desktop or even a brand new build secondary desktop they may not want to pay a lot for extra performance (that is what their primary Windows Tower is for).

 

[Madpacket]

Been busy with RL lately, just got caught up on the thread. Looks like some good discussion happening on the low end side of things! The high end stuff get's all the attention so this is fun for a change

I finally ordered a motherboard - MSI ITX H81I to pair up with a G3258 so I'll be doing some GT1 testing myself against the 5350 soon. I could have gone cheaper with mATX but I decided to go ITX as it gives the best flexibility case wise. Besides, it was around the same cost as my AsRock AM1H-ITX and it compares pretty even features wise.

As for game testing sure let's start compare Skyrim. The game is a little old but still very popular so it makes sense and is ideal for these types of systems. From memory I know Intel does well here as this game is still fairly CPU intensive so perhaps we should balance it out with another game that isn't as CPU demanding? Maybe Dirt 3 as it has a built in benchmark?

As for what part of Skyrim to test. Why not right area when you first get out into the real world after the initial cave area, you know when the dragon flies by?

**Edit

Let's choose some standard low resolutions for comparison.

How about 800x600, 1024x768 and finally 1366x768?

 

[cbn]

As for game testing sure let's start compare Skyrim. The game is a little old but still very popular so it makes sense and is ideal for these types of systems. From memory I know Intel does well here as this game is still fairly CPU intensive so perhaps we should balance it out with another game that isn't as CPU demanding? Maybe Dirt 3 as it has a built in benchmark?

Dirt 3 would be awesome.

I'm actually already downloading the game right now (I got it from the Never Settle forever Bronze Award when I bought my R7 250X). I'm interested to see how it will play with my Pentium iGPU.

]As for what part of Skyrim to test. Why not right area when you first get out into the real world after the initial cave area, you know when the dragon flies by?

You must mean after escaping Helgen Right? That may works well enough for a simple comparison. Maybe travel to a certain destination a short distance away from Helgen?

 

[cbn]

**Edit

Let's choose some standard low resolutions for comparison.

How about 800x600, 1024x768 and finally 1366x768?

Those work for me.

 

[frozentundra123456]

But see A8-7600 or Athlon X4/Pentium + discrete card is not a fair comparison to a $40 A4-7300 or $43 Celeron G1820.

If the person just wants a cheap processor for a spare parts build secondary desktop or even a brand new build secondary desktop they may not want to pay a lot for extra performance (that is what their primary Windows Tower is for).

The use case in discussion was light gaming at full HD, I assume 1080p, so that is more intense than a simple box for web browsing and playing videos.

Edit: but no matter what the use case I would not buy a kabini, atom, or single module APU for a desktop. Not when so little extra cost buys a lot more performance which might be needed as internet web sites and other uses potentially get more demanding.

 

[cbn]

The use case in discussion was light gaming at full HD, I assume 1080p, so that is more intense than a simple box for web browsing and playing videos.

Yes, he did mention light gaming at full HD (aka 1080p) and specified this as 2010 era games (I am not sure what games he is using from 2010, some were more demanding on hardware than others).

At some point (when I have time) I will do some testing with some of the older games I have and see how well they scale to 1080p with the Pentium iGPU.

 

[NTMBK]

Gets playable framerates. Notice the difference?

The difference being that you made a completely arbitrary distinction? Yeah, I see that.

 

[ytsoc]

Ups apparently i made a mistake
While looking at cpu benchmarks i didn't notice i was actually looking at the A10-7300 results wich were slightly better than the g1820, i can't seem to find the A4-7300 results on the cpubenchmark chart but i guess it will be a bit better than the 6300 (appx 500points difference between intel and amd according to cpubenchmark.com)

This pc is supose to replace an ancient Amd x2 4400+ and a 7300gt

As for the games, i asked my brother what he plays : FIFA 08 and CS 1.6/GO
so nothing fancy

Why am i still hung up on the amd APU? because of these results

 

[cbn]

Why am i still hung up on the amd APU? because of these results

Here are the charts from your link:

At least two thing of mention (that I can think of at the moment):

1.) For the ARMA results, I can't figure out how a Haswell Core i3-4130 with GT2 scores so much lower at both resolutions tested than Haswell Pentium G3220 with GT1?

2.) In ARMA, All the AM1 processors plus the FM2 and Haswell Dual cores have unplayable frame rates. (the A4-4000 just barely fails at the lower resolution though).

So anyway.... what happens if the resolution and/or detail settings are lowered to achieve playable frame rates? How much does the pecking order change and how quickly?

 

[MeldarthX]

Frozen and cbn brought up some interesting ideas. Dual core piledriver isn't that bad; but again I think with the prices coming down on 6600k that could be the most viable bottom choice on APUs.......

Intel has come a long way with the igpu; but their drivers still need a lot of work when you compare texture quality to AMD's apu; but at least things are becoming playable

I have seen A4 laptop that either had trinity or richland in them also play everything thrown at it at 768 range strangely better than my I5 laptop. As I said I don't think people realize just how good these apus are; looks like Intel dual cores are also becoming viable.

I have access to 6600k; 7600, and will need to check on laptop A4...I do have to be careful as the 7600 is work pc; so is my I5 laptop but I don't worry about I'll see if I can run some tests...

I do have a broad range of games...and list what I find

 

[cbn]

Here are my Dirt 3 results (using the game's built-in benchmark tool) for the following set-up:

Processor: G3258 @ 3.0 GHz
Memory: 2 x 2GB @ 1333 (iGPU shared memory at 512 MB)

Multisampling: OFF
ultra low preset

800 x 600:
102.41 avg FPS
76.03 min FPS

1024 x 768:
82.41 avg FPS
63.97 min FPS

1366 x 768:
67.1 avg FPS
53.57 min FPS

1920 x 1080:
39.44 avg FPS
32.74 min FPS

I can run some more benchmarks with higher presets and/or multisampling as well.