AMD Athlon 3000G Review with Overclocking

Brent_Justice

Administrator
Staff member
Joined
Apr 23, 2019
Messages
949
Points
93
banner.png




Introduction



It may not be as exciting as AMD’s Renoir announcement recently, but at least you can buy this APU in online and retail stores right now at an incredibly low price. We are talking about the AMD Athlon 3000G APU which was released in the Fall of 2019, almost now one year ago. You can buy this CPU with integrated Radeon Vega 3 graphics right now for only $49 on B&H. It was also
Continue reading...
Click to expand...
 
Last edited:
The question - for the ultimate in inexpensive system, get a cheap APU like the Athlon 3000G, or go buy something old, but cheap, like a Core2Quad 9650?
 
Paul_Johnson said:
APU
Click to expand...
Interesting - I somewhat recently advised a friend to pickup a Core2Quad for $75 (full system, complete with 8GB ram, and 500watt bronze power supply), and add a used GTX 980 to it for another $75. I assume, though may be wrong, that would end up both faster and cheaper than anything he could build new.
 
Endgame said:
Interesting - I somewhat recently advised a friend to pickup a Core2Quad for $75 (full system, complete with 8GB ram, and 500watt bronze power supply), and add a used GTX 980 to it for another $75. I assume, though may be wrong, that would end up both faster and cheaper than anything he could build new.
Click to expand...

Sure, it's cheaper but I definitely doubt it's nearly as powerful. I'd be really surprised if the C2Q didn't badly limit the video card. IPC has come a long way since the Core2 days as I can attest. A year ago I finally upgraded from a Q6600@3.6 to a Ryzen 5 2600x and the difference was like night and day. I only had a Radeon HD5770 which wasn't much newer than the Q6600 but in some games I was still CPU limited and the GTX980 is a hell of a lot more powerful than that HD5770.
 
Yeah that old C2Q system, I think I'd prefer something modern even if it were a bit slower. A lot of things have come a long way: DDR4, USB3, NVMe, SATA3, etc.

That's before you get into intangibles like security updates. warranty, software support, etc.

You may get the older system to benchmark a bit better, but I think the overall experience would be a lot better with a modern system (APU or otherwise). And you could always drop that 980 in on the APU as well.

Then again, if your budget is only $150... you don't really have a lot of room there to be too choosy. I think in that range, I'd just say save until you get the $300 or so it would take to build a bare bones budget new APU-based system, and then you could expand/upgrade from there pretty readily for the next few years.
 
SmokeRngs said:
Sure, it's cheaper but I definitely doubt it's nearly as powerful. I'd be really surprised if the C2Q didn't badly limit the video card. IPC has come a long way since the Core2 days as I can attest. A year ago I finally upgraded from a Q6600@3.6 to a Ryzen 5 2600x and the difference was like night and day. I only had a Radeon HD5770 which wasn't much newer than the Q6600 but in some games I was still CPU limited and the GTX980 is a hell of a lot more powerful than that HD5770.
Click to expand...
Are you sure the IPC is dramatically better from Core2Quad to today? Here at the house, I have a stock C2Q 9650 running Rosetta@home 24x7. I also have the following at the house running rosetta 24x7:
a 5 year newer A10-7870k OC'd to 4.44ghz,
a 8 year newer i3-6100
a 10 year newer Ryzen 2700X.

The C2Q is generating ~3200 average credit, while the A10 7870K is generating ~2000 average credit, the 6100 ~1800 average credit, and the 2700X ~13,000 Credit. Per core, outside of power usage, the C2Q seems to hold its own pretty well.

Edit to correct the model number on my A10.
 
Last edited:
I'm fairly sure I have a Q6600 in the closet and I can have Brent send the 3000G down to me. Should a showdown be set up between the 3000G @ 2c/4t against the Q6600 at 4c/4t? If so, what "tasks" would need to be compared? Gaming? Something else?

Not promising that I'll put this on the publication schedule, but can certainly consider it...
 
It would be very interesting to me, maybe not others. For used hardware, it might be best to see where the bang for the buck is? Can you reliably get, say, a used Ivy Bridge system for under $100?

Looks like the drop in GFlops / core is quite large between the Q6600 and the Q9650, but less than 10% between the 9650 and the 3500U:


CPU modelNumber of computersAvg. cores/computerGFLOPS/coreGFLOPs/computer

Intel(R) Core(TM)2 Quad CPU Q6600 @ 2.40GHz [Family 6 Model 15 Stepping 11]1124.002.5210.08

Intel(R) Core(TM)2 Quad CPU Q9650 @ 3.00GHz [Family 6 Model 23 Stepping 10]394.003.4913.95

AMD Ryzen 5 3500U with Radeon Vega Mobile Gfx [Family 23 Model 24 Stepping 1]508.003.6329.06

Source:
 
Endgame said:
Looks like the drop in GFlops / core is quite large between the Q6600 and the Q9650
Click to expand...

That seems to be more clock speed related than anything - bump the Q6600 up to 3GHz and it closes the gap a good bit.

So you're more focused on boinc performance than desktop tasks/gaming for this comparison?
 
David_Schroth said:
That seems to be more clock speed related than anything - bump the Q6600 up to 3GHz and it closes the gap a good bit.

So you're more focused on boinc performance than desktop tasks/gaming for this comparison?
Click to expand...
Its all I have to compare at this time in my home setup (edit - its all I have to compare old, cheap hardware to whatever newer stuff I have on hand or can read reviews for). I have a fair number of friends that ask me what to get, though (I'm "the IT" guy for all my friends). When a friend asked what the cheapest amount he could spend to get his kid a pc for League of Legends, Rocket League, and CS: GO, I based the decision on my personal boinc performance, as I didn't have much data for the games. I know my A10-7870K isn't great with the integrated GPU for some things - XCom 2 barely runs for example - but I figured that the 980 would more than cover plenty of FPS for the esports games.
 
Last edited:
Endgame said:
Are you sure the IPC is dramatically better from Core2Quad to today? Here at the house, I have a stock C2Q 9650 running Rosetta@home 24x7. I also have the following at the house running rosetta 24x7:
a 5 year newer A10-7870k OC'd to 4.44ghz,
a 8 year newer i3-6100
a 10 year newer Ryzen 2700X.

The C2Q is generating ~3200 average credit, while the A10 7870K is generating ~2000 average credit, the 6100 ~1800 average credit, and the 2700X ~13,000 Credit. Per core, outside of power usage, the C2Q seems to hold its own pretty well.

Edit to correct the model number on my A10.
Click to expand...

In the case of my Q6600@3.6Ghz I managed to get a multi-core score of 741 in CB20. My Ryzen 2600x pulled a multi-core score of 3004. It's not a perfect comparison because the Ryzen obviously has more cores/threads and a higher clockspeed but even by tripling the Q6600 score to "match" the thread count between the two you're still falling quite a bit short of the Ryzen at 2223. At that point the Q6600 is still 781 points short of the Ryzen which makes the Q6600 33% slower than the Ryzen.

I'm lucky to have those CB20 numbers for my Q6600 yet and I don't have any other numbers written down for anything else. I've done a lot of encoding for my Plex server and from memory (it's been a year since I swapped from Q6600 to Ryzen 2600x) I want to say that after normalizing the thread count between the processors my Ryzen is at least as much faster in encoding as it was in CB20.

I'm not running any DC projects at the moment due to heat but when I was I still had the Q6600@3.2 in my server going as well as my Ryzen 2600x. Keep in mind that the Ryzen as my main system wasn't always running 12 threads especially if I was doing anything with the system. In most cases I would drop the used thread count to 50% while gaming and around 84% while doing anything non-stressing to avoid latency and smoothness issues. The Q6600 was 100% all the time and rarely had any sort of workload. In this case the Ryzen averaged 6x-8x the points of the Q6600. Similar work units of the same project tended to take a minimum of 2x as long on the Q6600 and sometimes 3x as long.

When the first Core-i processors released there was a considerable performance increase compared to the C2Qs. That IPC increase has only increased since then. The increases between most generations wasn't large but it has added up to quite a bit over the years with the current Ryzens meeting or exceeding IPC of current Intel CPUs.
 
And now that I've finally read the review which I enjoyed quite a bit I can add an addendum to my previous post.

First of all, the DC project I was talking about is WCG. As usual I forgot to state specifically what I was talking about.

Second, from the review you can see that in CB20 the stock speed Athlon 3000g with only two cores but four threads outperformed my Q6600 and it did that with a 100mhz deficit in clock speed. There are other factors to take into account considering the Q6600 is only running DDR2 800 RAM and the review system at stock speeds and my own system are running DDR4 3200 speeds. The RAM and other differences likely have a positive impact but that's part of the whole package. It was also nice to see a sort of mini-review of the motherboard I have.
 
You must log in or register to reply here.
Become a Patron!
Back
Top