AMD will launch AM4 platform in March 2016 says industry source

[cbn]

It turns out that a lot of FM2+ chips are better host for mid-to-high end dGPUs than the G3258.

I would say with Nvidia dGPU that is true, but less so for AMD dGPU.

See the GTA V benches in post #273 as an example.

And some have speculated the reason the AMD CPU pairs better with Nvidia has to do with the Nvidia's multi-threaded graphics driver. The AMD graphics driver, in contrast, is single threaded.

 

[tential]

Many chose the pentium because the platform has farm more upgrade potential...

 

[cbn]

Many chose the pentium because the platform has farm more upgrade potential...

And I don't think I ever saw even a heavily overclocked Athlon x 4 860K beat a Haswell Core i3 in all those reviews that have gone by.

 

[cbn]

Regarding the idea of AM4 being brought in so early, I do hope this does not affect AMD's effort in mobile.

35W Bristol Ridge (with DDR4 2133/2400) in a laptop is probably the best use for the processor die.

If an early launch of AM4 end up affecting AMD's Bristol Ridge effort in mobile, then I would rather see AM4 delayed till Zen 8C/16T (and derivatives) are released sometime end of 2016/early 2017. (if necessary, AMD can survive with the combo of FM2+ and AM3+ on desktop till then)

 

[DrMrLordX]

R7 250X is a good combo with G3258.

And? That is not the point. The point is that the G3258 was expected to host a mid-to-high end GPU "just fine", which it obviously does not in many newer PC games.

And if you are going to say that is too weak for G3258 then we need to compare it to the AMD A10 APUs as well.

I didn't say that. People who actually bought the G3258 said that. And they didn't even give FM2+ a second look. Please stop trying to change the subject. We're already horribly off-topic.

I would say with Nvidia dGPU that is true, but less so for AMD dGPU.

So the HD 7950 + G3258 combo stuttering tells you what?

Many chose the pentium because the platform has farm more upgrade potential...

This is true, and they probably paid out more for their GPUs when initially building the system to "test the waters" with the G3258. Those that were not satisfied, eventually paid out more money for a better CPU. They might have done better to buy the better CPU in the first place.

And I don't think I ever saw even a heavily overclocked Athlon x 4 860K beat a Haswell Core i3 in all those reviews that have gone by.

How many "heavily overclocked" 860ks (or other Kaveris) have you seen in any review, period?

 

[cbn]

The point is that the G3258 was expected to host a mid-to-high end GPU "just fine", which it obviously does not in many newer PC games.

No, It is just a cheap little CPU with two cores and two threads.

Nobody expected it to play games well, let alone be paired with a an expensive GPU.

P.S. Athlon x 4 860K is not a good match for a mid-to-high end GPU either.

I didn't say that. People who actually bought the G3258 said that. And they didn't even give FM2+ a second look.

I don't remember seeing recommendations for pairing mid or high end dGPU with G3258. The most I saw recommended was R9 270 (and maybe R9 270X).

So the HD 7950 + G3258 combo stuttering tells you what?

AtenRA tested a Core i3 with HD7950 (OC to 1000 Mhz) processor earlier in this thread. I didn't see any G3258 with HD7950 test.

With that mentioned, honestly I believe a HD7950 is just too much GPU for either G3258 or a Athlon x 4 860K (for that matter).

 

[cbn]

And I don't think I ever saw even a heavily overclocked Athlon x 4 860K beat a Haswell Core i3 in all those reviews that have gone by.

How many "heavily overclocked" 860ks (or other Kaveris) have you seen in any review, period?

PC Lab reviews (like the one I used as an example in post #273) where they test one game with a whole bunch of different CPUs (and GPUs).

Not one time in each of the different game specific reviews can I remember the overclocked Athlon x 4 860K beating even the lowest Haswell Core i3.

If you can find one game where the overclocked Athlon x 4 860K did beat Haswell Core i3 let me know.

 

[AtenRa]

R7 250X is a good combo with G3258.

And if you are going to say that is too weak for G3258 then we need to compare it to the AMD A10 APUs as well.

Fact is dollar for dollar that G3258 + R7 250X was a much better value than any A10 desktop APU (and one reason was the throttling CPU cores onthe APU for the average user who didn't know how to manipulate AMDMrsTweaker and the other reason was lack of bandwidth on the A10 APU).

A10-7850K with CPU cores throttled down to 3.0 Ghz and 512sp iGPU @ 720 Mhz with cost adding DDR3 2400?

vs.

G3258 @ 4.0+ GHz and 640sp dGPU @ 1000 Mhz with 128 bit GDDR5?

That is no contest. (Even at stock CPU clocks I'll bet G3258 wins)

Hell, why drag the APUs in this again ?

For the same price as the Pentium G3258 + dGPU you had the choice of the Athlon 860K + dGPU.

And I don't think I ever saw even a heavily overclocked Athlon x 4 860K beat a Haswell Core i3 in all those reviews that have gone by.

You do know that the Core i3 is more expensive than the Core i3, and yet when OverClocked the Athlon 860K is faster in many workloads and extremely close or better than the Haswell Core i3s in games.

Anyway the thing is that dual cores are not for 2015 gaming and later unless you really want to stay with a entry-level dGPU. And as things are going im sure DX-12 games will only make the stuttering problems on dual cores even worse.

 

[AtenRa]

PC Lab reviews (like the one I used as an example in post #273) where they test one game with a whole bunch of different CPUs (and GPUs).

Not one time in each of the different game specific reviews can I remember the overclocked Athlon x 4 860K beating even the lowest Haswell Core i3.

If you can find one game where the overclocked Athlon x 4 860K did beat Haswell Core i3 let me know.

You dont have to beat it in Average fps, you also have to see the frame times each cpu produce in order to understand if the Core i3 will stutter or not.
Like the G3258 that has higher average fps but stutters to hell, the Core i3 may have a few average fps more than the OC Athlon but it may have higher stuttering. PClabs reviews never show frametimes and thus people only see the high average fps.

Im not saying the Athlon is better here because i havent tested my self, but as explained before higher average fps is not always the only metric you have to look.

 

[cbn]

For the same price as the Pentium G3258 + dGPU you had the choice of the Athlon 860K + dGPU.

Yes, I own both processors (G3258 and Athlon x 4 860K), and here was an informal test I did with Battlefield 4 64 player:

http://forums.anandtech.com/showpost.php?p=37501304&postcount=2

When I did informal testing of 4.3 Ghz Pentium G3258 vs stock speed Athlon x 4 860K in Batlefield 4 using R7 250X, the OC Pentium was about 10 FPS faster than the Athlon in the leveled building scene in 60-64 player Seige of Shanghai.

http://forums.anandtech.com/showpost.php?p=37322379&postcount=7

In my testing of the following leveled building scene in BF4:

https://www.youtube.com/watch?v=ZGiHqR61O4I&t=278

My overclocked G3258 @ 4.3 Ghz actually beat my stock clocked Athlon x4 860K with 60 to 64 players on the map using the same R7 250X card and detail settings.

I made numerous runs over and over again during the span of several games using both processors. I would say the OC G3258 was probably 10 FPS faster on average in that scene.

And here is the record of the frame rates I got for OC G3258 (for reference):

http://forums.anandtech.com/showpost.php?p=37279088&postcount=173

Regarding the informal BF4 64 player testing I mentioned back in post #164, Here is one of the areas I am using for comparisons:

https://www.youtube.com/watch?v=ZGiHqR61O4I&t=278

Walking thru the leveled building in Seige of Shanghai the Xbox One (which runs the game at 1280 x 720) stays mostly in the 40's/low 50's for FPS, but does drop as low as 30 FPS at one point.

With my G3258 overclocked to 4.3 Ghz, MSI Z97 U3 Plus, 2 x 4GB DDR3 1600 (which is actually faster than a Non-Z board would run the RAM), R7 250X running 1280 x 720 High setting (using Mantle API) I am getting 45 to 65 FPS ~99% of the time in that very same area with 60 to 64 players on the map. This over the span of several games. FPS did drop to a low of 31 FPS, but this was during an explosion and heavy shooting among multiple players.

The G3258 felt a lot better in the game.

 

[cbn]

Hell, why drag the APUs in this again ?

To me it just seems so ironic that if R7 250X is so low for such a budget build as G3258, then how does someone even begin classify desktop APUs?

And should these APUs (like A10) even be on the desktop?

For example with Bristol Ridge I believe allocating those 512sp dies to mobile is most likely a much better idea. And at 35W (with DDR4 2400) it would be nicely balanced all the way around.

But on the desktop? Do we really need that?

 

[AtenRa]

So you compering an OC Pentium vs stock Athlon with a entry level R7 250X in BF4 and again you dont mention anything about frame times.

But what happens when you go to a faster dGPU ???

 

[cbn]

So you compering an OC Pentium vs stock Athlon with a entry level R7 250X in BF4

My Gigabyte GA-F2A58M-DS2 doesn't let me OC the Athlon x 4 860K, otherwise I would have done it.

But yes, I was I was comparing 4.3 Ghz G3258 to stock Athlon x 4 860K (both using R7 250X @ 720p high settings on the same part of the map on Seige of Shanghai. This with 60 to 64 players in the game......for multiple rounds)

and again you dont mention anything about frame times.

Yes, it was just informal testing....but the G3258 felt a lot stronger.

In fact, I was so frustrated by the Athlon x 4 860K that I even made the following thread to check and see if my experience was atypical:

http://forums.anandtech.com/showthread.php?t=2424881&highlight=

 

[AtenRa]

To me it just seems so ironic that if R7 250X is so low for such a budget build as G3258, then how does someone even begin classify desktop APUs?

And should these APUs (like A10) even be on the desktop?

For example with Bristol Ridge I believe allocating those 512sp dies to mobile is most likely a much better idea. And at 35W (with DDR4 2400) it would be nicely balanced all the way around.

But on the desktop? Do we really need that?

The Athlon 860K + 250X is still low budget for dGPU gaming, also both G3258 and Athlon 860K paired with the 250X are more expensive than any APU today.

Also the A10 APUs like the 7850K is better than Athlon/Pentium paired with very low-end dGPUs up to R7 250 with DDR-3(see my review bellow). So yes they have a place in desktop as well.

 

[AtenRa]

what ram did you use with the Athlon ??

 

[cbn]

what ram did you use with the Athlon ??

Patriot Viper DDR3 1866 Cas 9 (2 x 4GB).

 

[cbn]

both G3258 and Athlon 860K paired with the 250X are more expensive than any APU today.

Not (in the United States) back when the Power Color R7 250X constantly cycled back and forth between $59.99 AR and $79.99 AR for well over a year.

 

[AtenRa]

Well, kaveris memory controller latency is higher than Trinity and you need to install a higher speed ram 2133MHz or more to lower the memory latency as much as possible. This is a problem the Kaveri has when you pair it with a dGPU, for iGPU is not that big deal.

But again without OC the Athlon to 4.3GHz or higher then you cannot make a fair comparison to the OC Pentium G3258.

A10-5800K with 1866MHz

A10-7850K with 2133MHz

yes in the USA there was a time when you could get the Athlon + 250X at lower/equal price as the A10 APUs.

 

[frozentundra123456]

The Athlon 860K + 250X is still low budget for dGPU gaming, also both G3258 and Athlon 860K paired with the 250X are more expensive than any APU today.

Also the A10 APUs like the 7850K is better than Athlon/Pentium paired with very low-end dGPUs up to R7 250 with DDR-3(see my review bellow). So yes they have a place in desktop as well.

Wow, if you go low enough, you can actually find a dgpu worse than an APU. Now there is a ringing endorsement.

 

[cbn]

Wow, if you go low enough, you can actually find a dgpu worse than an APU. Now there is a ringing endorsement.

There are also Nvidia dGPUs to consider.

I believe GT 730 GDDR5 is faster than R7 250 DDR3 due to having higher bandwidth (40 GB/s vs 25.6 GB/s).

But the main reason I am mentioning 512sp on desktop because I really hope AMD can make something of the 35W Bristol Ridge in a laptop.

And at this point I wonder if they would need every Bristol Ridge 512sp die to make that happen?

If so, maybe 384sp should be the top APU for Bristol Ridge desktop for an unspecified time.

Another reason, of course, is that 512sp Bristol Ridge on desktop will still be limited by bandwidth (although not as bad as Kaveri/Godavari due to the GCN 1.2 architecture).

 

[frozentundra123456]

Yea, considering the current state of APUs, mobile is the only place they make sense for gaming. On the desktop, there are just too many superior options with a cheap cpu like the 860k and a dgpu. I just don't buy the "I have to save ten or 20 dollars or I cant have a system" argument,when the cost of a tank of gas or a couple of cases of beer or one months cell phone bill, or even one game, will give a 50 to 100 percent increase in performance. Just seems like an artificially constrained scenario to justify an APU.

 

[Dresdenboy]

Yea, considering the current state of APUs, mobile is the only place they make sense for gaming. On the desktop, there are just too many superior options with a cheap cpu like the 860k and a dgpu. I just don't buy the "I have to save ten or 20 dollars or I cant have a system" argument,when the cost of a tank of gas or a couple of cases of beer or one months cell phone bill, or even one game, will give a 50 to 100 percent increase in performance. Just seems like an artificially constrained scenario to justify an APU.

How about this scenario: cheap pre-built "gaming" PCs and saved labour costs?

 

[monstercameron]

How about this scenario: cheap pre-built "gaming" PCs and saved labour costs?

Brought up that point years ago...

 

[cbn]

How about this scenario: cheap pre-built "gaming" PCs and saved labour costs?

1. The full 512sp desktop APU is not cheap.

2. These days a halfway decent dGPU is needed for AAA gaming.

 

[DrMrLordX]

PC Lab reviews (like the one I used as an example in post #273) where they test one game with a whole bunch of different CPUs (and GPUs).

There's only one linked test in that post showing an overclocking 860k with no indicator of what is the NB speed or memory speed, though it should be noted that in that test (and the one you cite in text), the 860k puts up a better minimum FPS score than the G3258. So it does just as well with Nvidia and AMD video cards in the minimum FPS department.

Not one time in each of the different game specific reviews can I remember the overclocked Athlon x 4 860K beating even the lowest Haswell Core i3.

Okay! Enjoy your i3 then . . .

Hell, why drag the APUs in this again ?

Wellll AM4's first slew of CPUs will be APUs, so discussing them is sort-of on-topic.

For the same price as the Pentium G3258 + dGPU you had the choice of the Athlon 860K + dGPU.

Yep, pretty much.

You do know that the Core i3 is more expensive than the Core i3

I know what you meant to say here, but it was still funny.

Anyway the thing is that dual cores are not for 2015 gaming and later unless you really want to stay with a entry-level dGPU. And as things are going im sure DX-12 games will only make the stuttering problems on dual cores even worse.

I'm not so sure it'll beat up on the dual cores that badly. At least in the early DX12 demos that were nothing but draw calls, the duals had no real improvement in max FPS but did see gains in min FPS, or so I recall . . .