Athlon 5350 Initial Impressions

Madpacket

Platinum Member
Nov 15, 2005
2,068
326
126
(April 19) - UPDATE #4 - Managed to get a few minutes to test Mantle on the 7970. Hope this helps some of you.


Battlefield 4:

Mantle = 25-45 FPS, holds steady lows around 25 fps on operation locker in the insane firefights and on siege of shanghai on 64 player madness. Average closer to 35 FPS. Definitely playable and a big difference in feeling from the Geforce 750 TI.

D3D = 20-30 fps, it quite a bit slower than with Mantle enabled but it was not a direct comparison as it was a different (64 player) map. Felt more like the 750 TI.


StarSwarm Results:

D3D

===========================================================
Oxide Games
Star Swarm Stress Test - �2013
C:\Users\minipacket\Documents\Star Swarm\Output_14_04_19_1512.txt
Version 1.10
04/19/2014 15:12
===========================================================

== Hardware Configuration =================================
GPU: AMD Radeon HD 7900 Series
CPU: AuthenticAMD
AMD Athlon(tm) 5350 APU with Radeon(tm) R3
Physical Cores: 4
Logical Cores: 4
Physical Memory: 8522588160
Allocatable Memory: 140737488224256
===========================================================


== Configuration ==========================================
API: DirectX
Scenario: ScenarioFollow.csv
User Input: Disabled
Resolution: 1920x1080
Fullscreen: True
GameCore Update: 16.6 ms
Bloom Quality: High
PointLight Quality: High
ToneCurve Quality: High
Glare Overdraw: 16
Shading Samples: 64
Shade Quality: Mid
Deferred Contexts: Disabled
Temporal AA Duration: 16
Temporal AA Time Slice: 2
Detailed Frame Info: Off
===========================================================


== Results ================================================
Test Duration: 360 Seconds
Total Frames: 5066

Average FPS: 14.07
Average Unit Count: 4145
Maximum Unit Count: 5587
Average Batches/MS: 187.35
Maximum Batches/MS: 356.38
Average Batch Count: 14179
Maximum Batch Count: 61072
===========================================================


Mantle:


===========================================================
Oxide Games
Star Swarm Stress Test - �2013
C:\Users\minipacket\Documents\Star Swarm\Output_14_04_19_1520.txt
Version 1.10
04/19/2014 15:20
===========================================================

== Hardware Configuration =================================
GPU: AMD Radeon HD 7900 Series
CPU: AuthenticAMD
AMD Athlon(tm) 5350 APU with Radeon(tm) R3
Physical Cores: 4
Logical Cores: 4
Physical Memory: 8522588160
Allocatable Memory: 140737488224256
===========================================================


== Configuration ==========================================
API: Mantle
Scenario: ScenarioFollow.csv
User Input: Disabled
Resolution: 1920x1080
Fullscreen: True
GameCore Update: 16.6 ms
Bloom Quality: High
PointLight Quality: High
ToneCurve Quality: High
Glare Overdraw: 16
Shading Samples: 64
Shade Quality: Mid
Deferred Contexts: Disabled
Temporal AA Duration: 16
Temporal AA Time Slice: 2
Detailed Frame Info: Off
===========================================================


== Results ================================================
Test Duration: 360 Seconds
Total Frames: 13027

Average FPS: 36.18
Average Unit Count: 3930
Maximum Unit Count: 5488
Average Batches/MS: 656.85
Maximum Batches/MS: 2311.84
Average Batch Count: 21514
Maximum Batch Count: 132474
===========================================================

That's an insane difference and shows a best case scenario for Mantle.

END

--

(April 19) - UPDATE #3 - Over the past few evenings I have tested additional games and finally discreet Geforce 750 TI testing. Next update (hopefully testing completed tonight) I will finally test out Mantle.

Gaming (APU):

Trials Evolution:

40-60+ FPS at 800x600, low settings and playable at 1280x800 but a little choppy in spots.

Doom 3:

35-60FPS at 1024x768, medium settings, perfectly playable.

Sleeping Dogs:

Fully playable at 640x480 / 800x600 (30-35 FPS). 1024x768 (22-25 FPS) was playable but not very comfortable.

Crysis:

Crashes shortly after launch, not sure why, skipped to next game.

Renegade Ops:

1680X1050, plays perfect at default settings, runs around 30-40 FPS, looks great.

Bioshock Infinite:

This game was a little too demanding for the APU. Average FPS = 23.89, Min FPS = 14.87, Max FPS = 66.72

---

Gaming and Power (Geforce 750 TI)

Idle power = 22 Watts!

Furmark + p95 = 99 Watts.

Furmark = 88 Watts.



FFXI scored 6220 and uses 42-47 watts. Results in "Very High".

BF4 Multiplayer Team DM and Conquest 32 players large ran 25-40 FPS at 1280x800 on low settings. Upping the resolution and or details dropped the FPS even more therefore this game while playable with reduced settings unfortunately is generally too taxing for the APU. Hopefully mantle can help here on the Radeon cards.

BF3 Multiplayer, 64 player conquest. 30-45 FPS at 1280x800 low settings. Although the fps were not a lot higher than BF4 the game actually felt much better so the 5350 seems to be enough for casual game or two.

Both BF4 and BF3 only use around 45Watts therefore you can tell the videocard isn't able to stretch it legs with this 5350 driving it.

Crysis 3:

1920x1200 on high with FXAA enabled = 25-70 FPS. Some dips in spots but totally playable and looks amazing. This game consumes most of what the Geforce 750 TI can dish out as it uses up to 85 Watts.

Sleeping Dogs:

1920x1200, high details, 28-45 FPS and used between 80-87 Watts! Perfect example of maxing out the video card, the 5350 is fast enough here to fully let the 7750 TI stretch its legs. This game looks amazing at higher resolutions.

Fallout New Vegas:

35-68 FPS, Ultra details, 1920x1200, perfectly playable and only consumed around 50 Watts which points to a CPU limitation however the game was perfectly playable and looks amazing.

Trials Evolution:

2304x1440 running highest details the game runs a solid 60FPS and consumes 72 watts from the wall. This is another game where the 5350 is plenty fast to drive a card like the 750 TI.

Rengade Ops:

2304x1440, 85-110 FPS! Crazy fast and consumes 75W. Same conclusion as Trials Evo.

QuakeLive:

COM_MAXFPS set to '250' - online CA was sitting between 160-250FPS in a 4 on 4 match (dm6 & trinity).

Offline practice on Asylum was also now consistently above 100fps, no more lag! 35Watts @ 1920x1200

Although overkill, the 750 TI and 5350 is fast enough to drive this game for competitive reasons.

That's it for this update.

END

---



(April 16) - UPDATE #2 - More Pico power and APU gaming results added. Next update will be to run the remaining APU game fps tests (have about 6-7 more games) and then move into discreet graphics card gaming.


Power:

Prime 95 + Furmark = 36W

Gaming:


Borderlands 2, all graphic settings to low except textures on high:

640x480 = 30-40 fps
800x600 = 25-35 fps (played an hour or so, a little slow in spots but totally playable, comparable to last gen console)
1024x768 = 20-25 fps, playable but not enjoyable.

BF4 (multiplayer) - Large Conquest - 1024x768, all set to low:

D3D = 13 FPS
Mantle = 15 FPS

Unfortunately this game is too demanding on the APU. It will be interesting to test with the discreet cards and if Mantle helps out.

Skyrim:

Playable at medium settings - 1024x768, 4x aniso, FXAA 24-30 FPS. Huge improvement over e350.

Defense Grid:

40+ FPS at 1680x1050 with medium detail settings. Older indy game but has good graphics and one of the most fun tower defence games. Much faster than e350.

XCOM Enemy Unknown: 25fps at default settings, 1024x768. Due to the nature of the game it's totally comfortable playing at these settings and the game looks amazing.

Trine 2: 20-22 FPS with Max details set at 1024x768. I didn't test this game very long but lowering the detail would easily make it more playable fps wise.

Left 4 Dead 2: High settings @ 1280x800 = 25-45 FPS on average, totally playable but may want to turn down resolution or details to get above 30 fps minimum.

End

--

160W Pico Killawatt Results:


50% load = 22W
Idle = 14/15W
Prime 95 all cores loaded = 30W
Various games = 22-32W



Either this AsRock board isn't as efficient as the others or the power supply is still to large for this combination. That being said these results are not too far off from my old Zotac e350 which I believe used a 60W adapter.

When I have more time I will test with a 60W 19v DC notebook adapter and compare the results.
--

Gaming on the APU:

I'm starting with some lighter games as I kicked off a download of 150+GB of games which will take some time to complete.

Super Meat Boy:

Played perfectly at default resolution, no dropped frames or weird dips. Compared to my old Zotac e350 this is a huge improvement.

Jamestown:

Also played great, no random slowdowns or dips in the frames (beyond what the game engine produces to avoid bullet hell) and was a great improvement over my older Zotac e350.

Quake Live:

Plays great. Ran tests at 1024/768 - the FPS dipped as low as 45 on the more stressful maps (Asylum) but it mostly sat between 85-120 FPS (120 FPS cap). This should be more than good enough for the casual game of quake but I feel the CPU is still holding it back for serious competitive gaming. Again compared to the older Zotac this was night and day faster.

--

Canned Benchmarks:

I'm not going to run many of these but they're good for comparison reasons.

FFXIV A Realm Reborn: 1280x800 with standard settings scored 1993 points - "slightly low".

--





*I'm not sure if I should start a new thread or piggyback off someone else's so please move or delete OP if I'm in violation.*

I received my Athlon 5350 and AsRock AM1H-ITX motherboard last night and wanted to share my initial impressions. I will list in a bullet point format my thoughts and hope to use this thread to perform additional testing based on any feedback I get.

- 5350 - Wow look how small this is! the APU box is roughly half the size/weight of the FM1/2 APU's.

- AsRock motherboard is pretty standard , nothing special other than the special SATA power cable that connect into a motherboard header to power 2 SATA devices. Considering the board has 4 SATA ports in total this is a bit of a let down. However if used in combination with a Pico power supply this works out perfectly as you can combine the two on board headers with 2 headers from the Pico power supply which in turn will power up to 4 SATA devices. Think of a low powered NAS powered by only a Pico power supply.

-Instead of using an external DC power adapter I'm currently powering the chip and board with a Pico 160W Pico ATX adapter. The Pico units are much more efficient than standard power supplies so this should work out well efficiency wise when I add in a low powered GPU for testing. Depending on what I end up using this chip and board combo in I may use the external DC power instead and it must be stated that it's really nice to have the option and represents the flexibility you get when only having to deal with such low powered systems.

-One last observation about the motherboard I thought was strange, there is no +12v motherboard power adapter. I guess the 24 pin ATX power adapter sends enough power to board where it's not longer needed.

-added in 2*4GB of DDR3 2133Mhz memory (don't have the brand off the top of my head, will look when I get home tonight).

-added a 180GB Intel 520 SSD to the first native port.

-Power the unit on and checked the voltages were correct, disabled third party SATA controller, adjusted memory voltage to correct voltage and then set motherboard to boot UEFI, attached external USB-DVD and then I proceeded to install Windows 8.1. Windows installed a little slower than I'm normally used to but not by much, in all around 25 minutes.

-Installed device drivers from motherboard DVD, installed windows updates. Rebooted several times, noted the boot up times were basically indistinguishable from my SSD equipped A10-6800K system which was a nice surprise.

-Observed clock speeds while installing device drivers and updates, noticed under normal load of the CPU speed liked to bounce between 1.0-1.8 Ghz or so.

- Ran some basic seat of the pants browsing tests and noticed how quick everything loaded again essentially indistinguishable from my 6800K system I last built. Ran a few 1080P youtube videos, observed about 20-30% CPU usage but no jitter or weird stuttering happens while playback running.

-Went back to BIOS to see if I could up the bclk but was disappointed to find out there are no controls available to do this. I was able to adjust the CAS timing from 2 to 1 under the RAM settings but I didn't bother tightening the rest of the timings yet. Note this board has the ability adjust memory voltage but nothing for the APU.

And that was about it for the testing for an evening. Not much was done however given what little I have done with the unit there were no real road blocks or surprises. Everything felt snappy enough from the seat of the pants.

Time permitting I plan on performing the following tests this week:

-Hook up to Killawatt and record wattage use. I'll measure both external 60w or 90w power brick as well as the existing PICO power supply and 600w EVGA bronze PS.
-Install steam and handful of older games from library for APU testing
-Install Radeon 7970GHz for more demanding game testing (BF4, Crysis 3 etc).
-Test mantle with/without multiplayer BF4
-Unigine benchmarks.
-Focus on lower power gaming, install Geforce 750 TI (ordered yesterday, should arrive before end of week), re-run demanding game tests and compare to 7970 results to see if there is any appreciable difference. See if Pico power supply is enough to power this card.
-Play with RAM timings to see if any speed improvements can be made.
-Netflix high quality HD playback tests via traditional browser, check for dropped frames etc.
-Winrar/7zip decompression speed testing.
-Install in a Radeon 290 to see if GCN 1.1 makes a difference for mantle tests.
-Check for DPC latency issues (Not sure what to use on Windows 8.1, if anyone knows please let me know).
-Install and run a couple VM's, observe if the host system bogs down whole VM's under load.

If there are any specific test you want run and it's not too time consuming please let me know. Given the hardware I have on hand I'm in a good position to test multiple different combinations for this AM1 platform.
 
Last edited:

Insert_Nickname

Diamond Member
May 6, 2012
4,971
1,691
136
If there are any specific test you want run and it's not too time consuming please let me know. Given the hardware I have on hand I'm in a good position to test multiple different combinations for this AM1 platform.

No tests, but just a quick question: exactly how tall is the included heatsink/fan assembly?

I've been thinking about upgrading my XPbox with this board, but there is quite a hard limit to how tall the cooler can be... :D
 

Abwx

Lifer
Apr 2, 2011
10,847
3,297
136
AsRock AM1H-ITX

- AsRock motherboard is pretty standard , nothing special other than the special SATA power cable that connect into a motherboard header to power 2 SATA devices. Considering the board has 4 SATA ports in total this is a bit of a let down.

4 sata but also 4 usb3 instead of two in other boards, 7.1 audio with both analog and optical output , said 19V DC connector, 4 video outputs, quite a few desirable features that made me choose this MB for a next build despite its anormaly higher price tag.

Other than that that s quite a lot of tests to come , would be much interested by the general impression and will follow the thread closely since i ll do my build only in a week or two, would be great to have some power numbers as this MB wasnt used in the AM1 reviews.
 

Madpacket

Platinum Member
Nov 15, 2005
2,068
326
126
4 sata but also 4 usb3 instead of two in other boards, 7.1 audio with both analog and optical output , said 19V DC connector, 4 video outputs, quite a few desirable features that made me choose this MB for a next build despite its anormaly higher price tag.

Other than that that s quite a lot of tests to come , would be much interested by the general impression and will follow the thread closely since i ll do my build only in a week or two, would be great to have some power numbers as this MB wasnt used in the AM1 reviews.

Good point, yes this board comes with a lot compared to the other AM1 boards available, and that's why I picked it up despite the price premium.

Yes a lot of testing planned, I'm not sure how long it'll take but I'll update the results in the main thread once I have numbers. I will get power numbers up soon as well no worries.
 

Gikaseixas

Platinum Member
Jul 1, 2004
2,836
218
106
I have a A8 3850 paired with a 5400rpm slow a$$ drive and it's killing me. Reading your post made me buy a small SSD drive. The APU powered PC is also the only system in the house without SSD.
 

Madpacket

Platinum Member
Nov 15, 2005
2,068
326
126
I have a A8 3850 paired with a 5400rpm slow a$$ drive and it's killing me. Reading your post made me buy a small SSD drive. The APU powered PC is also the only system in the house without SSD.

Good call, the A8-3850 is still decently fast (much faster than this Athlon 5350) which will seriously benefit from an SSD. I had a similar APU, the A6-3670K which was using an older 7200 RPM drive and after I upgraded to an SSD and saw an immediate boost in speed.

I no longer build PC's without an SSD, it's just not worth it :D
 

Insert_Nickname

Diamond Member
May 6, 2012
4,971
1,691
136
Not at home to measure but the dimensions are listed in this thread here:

http://forums.anandtech.com/showthread.php?t=2374795

Thanks. Just what I needed. Its notoriously difficult to judge height from just photos... :)

If its only 4cm tall it'll fit the case. Coupled with a 5350 it should be a nice upgrade from the C60 that system is currently running. As a bonus it can even drive my main monitor at full 1440p resolution.

(come to think of it, the four core Athlon at 2.05GHz is likely faster then the majority of hardware I've run XP on. Food for thought)
 

Madpacket

Platinum Member
Nov 15, 2005
2,068
326
126
Compared to a C60 this 5350 will feel much faster. I've been comparing it to an e350 which is the same architecture as a C60 but a few hundred megahertz faster and the 5350 makes the the e350 feel mostly antiquated.
 
Last edited:

davie jambo

Senior member
Feb 13, 2014
380
1
0
-Install Radeon 7970GHz for more demanding game testing (BF4, Crysis 3 etc).


We need to see how it does , could be an awesome budget chip this
 

bgt

Senior member
Oct 6, 2007
573
3
81
Just testing the same setup,
idle=11W on 19V jack(using a FSP Twinkle 65 mini adapter)
P95=33W
Furmark=28W
P95+Furmark=41W

Now with a 12V/6,5A LC-Power adapter(bric) and a pico 80W psu:

Idle=10W
P95=40W
P95+Furmark=54W

Fan is off(speedfan) in idle mode, temp then is 47C
Using P95 and Furmark raises the fan speed a bit because temp rising to 54C. Fan is very quiet.
 
Last edited:

strata8

Member
Mar 5, 2013
135
0
76
It's much better than the C60. 2.5x faster in single thread and anywhere between 4-5x faster in multi thread. The GPU is 3-4x faster as well.
 

bgt

Senior member
Oct 6, 2007
573
3
81
Now using the 12V/6,5A LC-Power adapter on the jack input:
Idle=9W
P95=32W
P95+Furmark=40W
 

bgt

Senior member
Oct 6, 2007
573
3
81
It's much better than the C60. 2.5x faster in single thread and anywhere between 4-5x faster in multi thread. The GPU is 3-4x faster as well.
It truly is a very snappy CPU. Comparing it to my Pentiums and Celerons it has the same oomph.
 

Madpacket

Platinum Member
Nov 15, 2005
2,068
326
126
Now using the 12V/6,5A LC-Power adapter on the jack input:
Idle=9W
P95=32W
P95+Furmark=40W

Thanks for adding your power usage results. Nice to see sub 10W idle. I'll have to retest tonight on my 160W power supply to see what furmark + p95 push together. I have a bunch of 65W 20 volt Lenovo Thinkpad power adapters but they have the wrong ends on them. I wonder if I can get an end converter to fit the AsRock, if that would work or would I risk damaging the motherboard as it's 20 volt instead of 19? Hmm...

For direct comparison using a game can you please run the FFXIV benchmark at the same settings and tell me what your power usage results are?
 

Abwx

Lifer
Apr 2, 2011
10,847
3,297
136
Looking at the board it looks like there s two DC-DC converters to generate the 5V rail as well as a 12V rail that is used to supply the hard drives but not to cater the APU DC-DC converter as this latter is either connected to an ATX 12V rail or directly to the +19V rail when the jack input is used, this would explain why 12V still work well as a voltage reducer DC converter wouldnt work well once the input voltage is too close of the fixed output voltage, other than that it s highly unlikely that using 20V instead of 19V would damage the board , there s inherently a 20% at least margin in respect of over voltage and much more tolerance for lower voltages than specs but it can happen that a too low input voltage can damage a switching mode PSU/DC-DC converter, depending of the design.
 
Last edited:

bgt

Senior member
Oct 6, 2007
573
3
81
Thanks for adding your power usage results. Nice to see sub 10W idle. I'll have to retest tonight on my 160W power supply to see what furmark + p95 push together. I have a bunch of 65W 20 volt Lenovo Thinkpad power adapters but they have the wrong ends on them. I wonder if I can get an end converter to fit the AsRock, if that would work or would I risk damaging the motherboard as it's 20 volt instead of 19? Hmm...

For direct comparison using a game can you please run the FFXIV benchmark at the same settings and tell me what your power usage results are?

FFXIV benchmark?
 

bgt

Senior member
Oct 6, 2007
573
3
81
Looking at the board it looks like there s two DC-DC converters to generate the 5V rail as well as a 12V rail that is used to supply the hard drives but not to cater the APU DC-DC converter as this latter is either connected to an ATX 12V rail or directly to the +19V rail when the jack input is used, this would explain why 12V still work well as a voltage reducer DC converter wouldnt work well once the input voltage is too close of the fixed output voltage, other than that it s highly unlikely that using 20V instead of 19V would damage the board , there s inherently a 20% at least margin in respect of over voltage and much more tolerance for lower voltages than specs but it can happen that a too low input voltage can damage a switching mode PSU/DC-DC converter, depending of the design.

@12V the jack converter, the 2 power mosfets and coil, stay much cooler. It only has to pass the current through the semiconductors. No switching losses.
Just a bit more loss on the external 12V bric. @19v and running P95 the 2 mosfets and coil reach 70C. Not a lot of cooling there. Ofcourse this is an extreme condition but still.......maybe a small cooling bloc on top of the semi's?
 

bgt

Senior member
Oct 6, 2007
573
3
81
Glued a cooler to the semi's.
The arrows show the hot chips(and coil)..................when u use P95.
AM1heat.jpg

AM1heat2.jpg
 

DigDog

Lifer
Jun 3, 2011
13,444
2,084
126
Quake Live:

Plays great. Ran tests at 1024/768 - the FPS dipped as low as 45 on the more stressful maps (Asylum) but it mostly sat between 85-120 FPS (120 FPS cap).
console-> maxfps 250

if you play EU i'm iSuckz (i play CA)
 

Abwx

Lifer
Apr 2, 2011
10,847
3,297
136
@12V the jack converter, the 2 power mosfets and coil, stay much cooler. It only has to pass the current through the semiconductors. No switching losses.
Just a bit more loss on the external 12V bric. @19v and running P95 the 2 mosfets and coil reach 70C. Not a lot of cooling there. Ofcourse this is an extreme condition but still.......maybe a small cooling bloc on top of the semi's?

If the components in your shots are effectively hot then the APU DC converter is likely in serial with the primary 19V-->12V DC converter, this would limit the overclocking capabilities, if ever there is some, unless an ATX coventional PSU is used, the HDD onboard dedicated 12V supply is indeed power limited according to Asrock when using the 19V DC input with two 3.5" HDDs being the maximum.

Cooling the mosfets is a good idea but beware that coils do not like piece of metals being very close, fortunately this is aluminium wich is not magnetism bounded and has little influence over the coils inductance.
 

Madpacket

Platinum Member
Nov 15, 2005
2,068
326
126
FFXIV benchmark?

FFXIV = Final Fantasy 14 which is an online MMORPG. Square basically rebuilt the game using an updated engine so it runs better on lesser hardware like the 5350. I think the engine makes good use of the available hardware.

You can grab it here:


http://na.finalfantasyxiv.com/benchmark/


console-> maxfps 250

if you play EU i'm iSuckz (i play CA)

I'm in NA and still play CA on occasion. QL is still the best FPS imho. I can set the cap to 250 but given the fps measured already I don't really see the point. If there's a specific scenario you want tested let me know.