A12-9800 Overclock with Firestrike & Timespy

Nightfire

2[H]4U
Joined
Sep 7, 2017
Messages
3,279
So I went aheaed and purchased a Bristol Ridge CPU. Yeah, I know, but it works fine until Raven Ridge comes out.
Core components list: 8 GB Patriot Viper 3200mhz, A12-9800, & ASUS B350m TUF - about $300 total. The remaining components I had.
Things started out bad when I misdiagnosed my PSU instead of a bad MSI Mortar M/b. I ended up getting a Corsair STX PSU first, only to discover that the system would still not boot.
I had discovered that it would only light up with the CPU power disconnected. Crap. RMA is difficult for me since I am APO, but we will see.

Here are stock results:
fs stock.PNG



Overclocking was much tougher than Godvari. I tried to raise the multiplier first with no luck. I later was successful with LLC, but the temps were INSANE!
I was also disappointed that I could not get 2400 Mhz Ram working, despite the timings or voltages. The only success so far has ben raising the base clock
which in return
raises the CPU freq, GPU freq, and Ram freq. However, it seems to hover around 4.0 GHz and idle much higher. I was also able to drop the timings considerably, which helped - It didn't seem to with Godvari, but those
were already pretty low. Lastly, there is now way to raise NB freq, which helped Godvar.Here is the o/c results:



fs 105 cas 12.PNG


Keep in mind, this is Windows 10. Windows 7 is generally 70 points lower on these APUs with all things equal.
Godvari results can be found here:
https://hardforum.com/threads/a10-7860k-overclock-with-firestrike.1900984/

I wasn't able to run Timespy stock, but o/c it was 788 points :p
Yeah, not amazing by today's standards, but still fun to mess around with until Raven Ridge.

I ordered an H80i v2 so I might try and play around some more :)
 
Admittedly, probably more about Bristol ridge her than anyone would care to know. But I like the homage " it is better to drive a slow car fast than a fast car fast".

Anyhow, I did some more tweaking with the new H80i v2. Cooling is ridiculously good but I can't get bast 4.3 GHZ no matter what. These are the results with 4.3 Ghz (105 x 41) and CAS 10: (2122 points)
fs 105 cas 10.PNG
I added Sandra soft information. As you can see, I am only getting about 11 GB/s. This is in line with Steve at Techspot who was running faster memory and slower timing. The thing is, the RD890 memory controller should be capable of 16 GB/s in even DDR3 mode.
I am going to make the assumption that this memory is not working in dual channel. I will try to see if there is any difference in SC configuration. Perhaps I can get 2400+ mhz working??

Here is some good info about single vs dual channel and how it may not be as necessary as we are led to believe:
https://www.gamersnexus.net/guides/1349-ram-how-dual-channel-works-vs-single-channel?showall=1
 
Well, I did try single channel with rather terrible results. I was still not able to load 2400 mhz no matter what timings or voltages I used. I even tried a lower baseclock.

All settings were the same as the previous run:

singel max.PNG

My memory bandwidth dropped from 10.66 Gb/s to 8.57 GB/s or a 20% loss.

However, the hit on Firestrike was HUGE!

My score fell from 2122 to 1237 or a 40% loss!
Graphics score went from 2416 to 1410or a 41.5% loss.
Even by Desktop graphics performance dropped from 8.0 to 7.0 on WEI.

It amazes me that they still Nerfed this APU with memory bandwidth. AMD did nothing going from DDR 3 to DDR 4. I would have been really curious to know what this APU could do with even 3000 MHZ of DDR4 dual channel.

Oh well, Raven Ridge will be here in a few months.
 
I remember from my Godaveri chip that the RAM memory speed had an insane performance impact. I hope when Raven Ridge comes out that it has a good support of higher RAM speeds. Thanks for posting your performance results. The integrated graphics market is gonna get a real shake up with AMD integrated graphics sitting on Intel CPUs in the future.
 
ARRGHHH so close to 2200! I will try again tomorrow:

2197.PNG


I couldn't get cas 10 working even when dropping down to 2400 mhz, but i was very happy with theses timings: 11,11,11,28,48 with a 352 trfc @2520 mhz. Strangely, it is still 352/1056 even with the Flarex - must be a MB thing.

The CPU portion is crap compared to Godvari. 41 x 105 Mhz is stable, but 42 x 104 Mhz (4368 mhz) took every last mV and even then only worked half the time. Godvari was good for 4700 mhz at least.
the Physics portion of FS consumed 180 watts @ 4.7 ghz with Godvari, but a whopping 210+ watts @ 4.3 ghz with this turd. I am really surprised the score is this high compared to my Godvari chip.
 
Finally!

2204.PNG

I tried Timespy with the same settings but it crashed. I dialed it down a bit and got a 794 but noticed temps hit 89*! The CPU portion of the test was running at 240 watts, peaking to 250 watts. My poor motherboard LOL. So I put the H80i fans on max and managed to hit an 831 :)
831.PNG

Keep in mind that the 2700u only managed to get a 915. Are you kidding me?? So a 4/8 Ryzen - Vega APU only manages 10% more performance in gaming than a 2/4 Bulldozer - R7?! Yeah I know the desktop versions will clock much higher, but I was hoping for a MINIMUM of 2x performance when comparing o/c to o/c. Raven Ridge is looking to be disappointing when compared to this glorified Kaveri from Jan 2014. That chip will be 4 years old by the time Raven Ridge desktop is for sale, and this is the performance boost we get?!
 
Finally!

View attachment 46009
I tried Timespy with the same settings but it crashed. I dialed it down a bit and got a 794 but noticed temps hit 89*! The CPU portion of the test was running at 240 watts, peaking to 250 watts. My poor motherboard LOL. So I put the H80i fans on max and managed to hit an 831 :)
View attachment 46010

Keep in mind that the 2700u only managed to get a 915. Are you kidding me?? So a 4/8 Ryzen - Vega APU only manages 10% more performance in gaming than a 2/4 Bulldozer - R7?! Yeah I know the desktop versions will clock much higher, but I was hoping for a MINIMUM of 2x performance when comparing o/c to o/c. Raven Ridge is looking to be disappointing when compared to this glorified Kaveri from Jan 2014. That chip will be 4 years old by the time Raven Ridge desktop is for sale, and this is the performance boost we get?!
I think the thermal limit is killing the potential performance here. These new chips dynamically throttle to maintain their TDP.
 
Back
Top