AMD AM1 Sempron 2650 1.5GHz dual-core or 3850 1.3GHz quad-core for single-thread?

Page 3 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

waltchan

Senior member
Feb 27, 2015
846
8
81
After careful decisions, I officially went with the Sempron 2650, and opened the box already. Thanks to everyone here for sharing good opinions and recommendations for me. :)

Here's some reasons why I went with Sempron 2650 instead of 3850:

  • Only $19.99, I'm not in the mood to wait for shipping and pay the extra money for 3850.
  • 13.33% faster single-thread speed than 3850.
  • Will be used for Windows XP Home Edition, Windows Vista Home Basic, and Windows 7 Starter 32-bit operating systems only (no 64-bit). Three different partitions installed into one 1TB hard drive.
  • It will look weird and inappropriate to put in a luxury quad-core processor for all the cheap 32-bit operating systems I'm installing.
  • I sometimes put in 2GB RAM to save on costs (right now I have 4GB). 3850 won't be happy with 2GB and looks weird. 2650, okay, it's more appropriate given the cheap price.
  • Lower power consumption. 2 cores missing and slower GPU speed are about 17% lower power consumption than 3850.
 
Last edited:

SPBHM

Diamond Member
Sep 12, 2012
5,058
410
126
I don't see any difference in using 32bit OS with a 2650 or 3850, It's "wrong" for both, also 2GB is to little for both dual and quad core and just look at all those Atom quad core tablets with 2GB.

I honestly think this power usage and ST performance difference will be a lot more difficult to notice than the advantage from the 2 extra cores, example a web browser like chrome under heavy load or when you happen to run a video without hardware acceleration working properly.. (it can happen with all the software and driver updates), or for example if you try to play HEVC, I think lower quality HEVC can run on the 3850, but not the 2650?

but I can understand your choice, it's a little cheaper, not inferior 100% of the time, and you think it will be enough for this usage...
 

waltchan

Senior member
Feb 27, 2015
846
8
81
but I can understand your choice, it's a little cheaper, not inferior 100% of the time, and you think it will be enough for this usage...
I have to rebuild and build 6 mATX PCs from scratch stacked on top each all the way up to the ceiling (plus earthquake brackets).

I do have a fast-one, actually. I finished my first build already, which is a Pentium G3258 LGA1150 overclocked at 4.3GHz with a Asus B85 Vanguard motherboard and 8GB RAM. Six different UEFI 64-bit partitions already installed, all the way up to Windows 13 in the future.

The second one is a Sempron 2650 I'm building right now. 32-bits are all about cheap processors and low memory, that's what we had in the past. Plus Sempron and Pentium names rhyme. I will be using a MSI AM1M motherboard. Only four non-UEFI 32-bit partitions can be installed max.

My goal is to buy one different socket each, one different motherboard brand, one different DVD drive maker, hard drive, and etc. This is a very complicated process. I plan to install FM2, AM3, LGA1155, upcoming LGA1151, and maybe Bay-Trail for the last four.

I can only choose Gigabyte, ASRock, Biostar, ECS, and Foxconn left each for the last 4 (not going to be easy as not all of them make that socket and specs I need).
 
Last edited:

Fayd

Diamond Member
Jun 28, 2001
7,970
2
76
www.manwhoring.com
Your C-50 single-thread score is only 263 :eek:, half the speed of Sempron 2650 at 504. I'm surprised you're still using it today, as the C-50 is rated by me one of the worst processors ever built in the "worst CPU" thread I mentioned.

Change the power plan to "high performance" mode, it may help.

then you have absolutely no idea what you are talking about. it's cpu performance is lackluster...who cares. it's comparison was an atom N230 to N270. and it blew them out of the water.

and yes, i used a c50 based netbook for 2 years as my primary laptop.
 

Abwx

Lifer
Apr 2, 2011
11,515
4,301
136
i like apus and all but even i don't think amd should be offering such a low level of performance.

Realisticaly that s about the equivalent of a 1.8/1.95 BT in Integer/FP and 1.67/1.75 for a Bobcat.

Now on the frequency requirement it s clear that someone did a very bad job at AMD, looking at AT AM1 review it is obvious that they could had clocked thoses 2C at 1.85-2.05 and still retain a power comsumption lower than a 4C Kabini, moreover given that the AM1 chips are GF fabbed and have very good specs.

Edit : At least, and contrary to BT, thoses build can later be upgraded when the Athlon 5350 will also be 15-20$ or so...
 
Last edited:

Shivansps

Diamond Member
Sep 11, 2013
3,875
1,530
136
Still looks awfull to me, i rather go G1820/G3220 for CPU/IGP power, A4-7300 for IGP power, J1800/1900 for lower power and passive, J1800 for lower price, J1900/2900 for HTPC or a 1037U for performance/price. I just dont see any room for AM1 anywhere, that thing exist just because FM2 failed on ITX, and AMD needed to have something in that sector.

BT are not gona be obsolete by next year because BT does not sell because it has a good igp, and you probably be able to get a better video by adding a $30 gpu intro the x1 pci-e than upgrading to CT.
 
Last edited:

cbn

Lifer
Mar 27, 2009
12,968
221
106
I decided to buy the 2650 since it was on sale via this combo deal:

http://forums.anandtech.com/showthread.php?t=2426489

Interestingly a complete Windows 7 SFF Core 2 duo (with a good CPU like the E6550) computer can be on sale for around that same price or less. Of course it is used hardware and will have higher idle power consumption, but it won't need memory, storage and OS like that Fry's combo deal does.

Regarding performance, I will be doing some informal testing over the next few days with the 2650 (mainly just browsing and playing videos on a dual monitor set-up just to see how liveable it is or isn't). Just to forewarned this is a slow cpu with a passmark score that is actually a hair lower than E2140 (which is the slowest Core 2 duo ever made), although the iGPU is certainly better than a GMA 3100.
 

cbn

Lifer
Mar 27, 2009
12,968
221
106
Now on the frequency requirement it s clear that someone did a very bad job at AMD, looking at AT AM1 review it is obvious that they could had clocked thoses 2C at 1.85-2.05 and still retain a power comsumption lower than a 4C Kabini, moreover given that the AM1 chips are GF fabbed and have very good specs.

I agree the cpu clocks should be higher.

It also too bad the memory bandwidth and iGPU clocks were not higher as well.

Max DDR3 speed on Athlons and Sempron 3850 is DDR3 1600, but Sempron 2650 can only run memory at DDR3 1333.

iGPU clocks on the Athlons are 600 Mhz while the Sempron 3850 runs 450 Mhz and the Sempron 2650 runs 400 Mhz.

P.S. I get the point about product segmentation, but in this particular instance I think AMD should have been generous with the technology on the dual core.
 

cbn

Lifer
Mar 27, 2009
12,968
221
106
Team Fortress 2 with Sempron 2650 (in Windows 8.1) is unplayable even at 800 x 600 low. (frame rates were in the single digits a good amount of the time in Saw Mill)

In contrast, my Athlon 5350 could play the game at 1080p low.

P.S. For a very cheap TF2 set-up, I would think E6550 with a sale priced $10 AR free shipping Power color HD5450 would be something to consider (though I have not tested this exact combo yet).
 

VirtualLarry

No Lifer
Aug 25, 2001
56,542
10,167
126
Team Fortress 2 with Sempron 2650 (in Windows 8.1) is unplayable even at 800 x 600 low. (frame rates were in the single digits a good amount of the time in Saw Mill)

In contrast, my Athlon 5350 could play the game at 1080p low.
But, is that because of the CPU cores, or the down-binned / down-clocked IGP? (I assume that you were playing on the IGP?)

Edit: As much as I rail against Intel ferociously castrating their CPUs, to create extreme market segmentation, I actually think that the AMD Sempron 2650 is the worse example of that. Lowering the clocks on both the CPU cores and the IGP, makes that chip basically nearly unusable. If the cores were faster, and the IGP speed at par with the other AM1 APUs, then it might have been viable.
 
Last edited:

coercitiv

Diamond Member
Jan 24, 2014
6,593
13,915
136
Lowering the clocks on both the CPU cores and the IGP, makes that chip basically nearly unusable. If the cores were faster, and the IGP speed at par with the other AM1 APUs, then it might have been viable.
Something tells me cat cores @TSMC did not go as well as planned. But that's history now.
 

cbn

Lifer
Mar 27, 2009
12,968
221
106
But, is that because of the CPU cores, or the down-binned / down-clocked IGP? (I assume that you were playing on the IGP?)

I didn't notice any increase in FPS dropping down to 640 x 480 low and the game remained unplayable even at this lowered setting.

According to MSI afterburner the CPU cores were pretty much pegged at all three test resolutions of 1024 x 768, 800 x600 and 640 x 480. GPU utilization (which looked very jagged) bounced around between 50 and 100% during gameplay at the highest resolution of 1024 x 768. When GPU utilization did hit 100% it didn't stay there for very long, maybe a moment or two.

So I think it is most definitely the CPU cores that are the limiting factor.
 
Apr 20, 2008
10,064
984
126
Team Fortress 2 with Sempron 2650 (in Windows 8.1) is unplayable even at 800 x 600 low. (frame rates were in the single digits a good amount of the time in Saw Mill)

In contrast, my Athlon 5350 could play the game at 1080p low.

P.S. For a very cheap TF2 set-up, I would think E6550 with a sale priced $10 AR free shipping Power color HD5450 would be something to consider (though I have not tested this exact combo yet).

Source can use 3-4 cores if they're clocked low. The 1.3GHZ quad was probably the better choice, even for gaming. When I lock the FX-8350 at 1.4Ghz, all source games are still very playable, if not 60fps with vsync.
 

waltchan

Senior member
Feb 27, 2015
846
8
81
Plans changed in my build rack. AM1 is not compatible with Windows Vista at all, so I was forced to replace with a A4-4000 FM2 processor and Gigabyte GA-F2A58M-HD2 motherboard. A used A4-4000 processor is insanely-cheap on eBay right now, selling for less than $20 shipped, way less than the Sempron 2650.

Sold A4-4000 listings: http://www.ebay.com/sch/i.html?_from=R40&_sacat=0&_nkw=a4-4000&LH_Complete=1&LH_Sold=1&rt=nc

This Gigabyte board can be purchased for FREE at Micro Center with combo purchase (then sell the costly FM2+ processor on eBay), and it can be overclocked up to 4.14 GHz easily using multiplier. Incredible deal I think, lowest-price (almost FREE A4-4000 combo) and fastest-performance all in one.
 
Last edited:

cbn

Lifer
Mar 27, 2009
12,968
221
106
Plans changed in my build rack. AM1 is not compatible with Windows Vista at all, so I was forced to replace with a A4-4000 FM2 processor and Gigabyte GA-F2A58M-HD2 motherboard. A used A4-4000 processor is insanely-cheap on eBay right now, selling for less than $20 shipped, way less than the Sempron 2650.

Sold A4-4000 listings: http://www.ebay.com/sch/i.html?_from=R40&_sacat=0&_nkw=a4-4000&LH_Complete=1&LH_Sold=1&rt=nc

This Gigabyte board can be purchased for FREE at Micro Center with combo purchase (then sell the costly FM2+ processor on eBay), and it can be overclocked up to 4.14 GHz easily using multiplier. Incredible deal I think, lowest-price (almost FREE A4-4000 combo) and fastest-performance all in one.

I couldn't find the passmark score for A4-4000, but here is the A4-4020 which scores 1753 CPU marks: http://www.cpubenchmark.net/cpu.php?cpu=AMD+A4-4020+APU&id=2305

P.S. I also have a A6-5400K Processor which I can test on Team Fortress 2. Hopefully I have some time available to do that soon. If all goes well I will test Team Fortress 2 in Linux as well (just keep in mind AMD graphics drivers for Linux are known to be not the greatest)
 

waltchan

Senior member
Feb 27, 2015
846
8
81
A4-4000 benchmark score is around 1650 area (200 MHz less than A4-4020), which describes its very-low resale value. It originally sold for $39.99 last year. Now only $15-$20 used. That's way less than Celeron G1820.

But overclocking it to 4.14 GHz brings it up to 2300 area.

I even did a post on A4-4000 at SlickDeals back in November:

http://slickdeals.net/f/7391302-amd...center-b-m-or-for-free-with-a4-4000-processor
 
Last edited:

escrow4

Diamond Member
Feb 4, 2013
3,339
122
106
A4-4000 benchmark score is around 1650 area (200 MHz less than A4-4020), which describes its very-low resale value. It originally sold for $39.99 last year. Now only $15-$20 used. That's way less than Celeron G1820.

But overclocking it to 4.14 GHz brings it up to 2300 area.

I even did a post on A4-4000 at SlickDeals back in November:

http://slickdeals.net/f/7391302-amd...center-b-m-or-for-free-with-a4-4000-processor

You do know how slow that A4 is?

http://www.techspot.com/review/681-amd-a10-6800k-a4-4000/page3.html

"The A4-4000 on the other hand scored just 4972pts, making it 7% slower than the A4-5300 and 31% slower than the Intel Celeron G1610."

"The A4-4000 was 4% slower than the A4-5300 and 30% slower than the Celeron G1610."

Next page:

"The A4-4000 was considerably slower at 33.80 seconds. This made it 20% slower than the A4-5300 and 40% slower than the Celeron G1610."

Power consumption:

"This means that the A4-4000 system consumed 46% more power than the Pentium G2020 and 53% more power than the Celeron G1610."

Low end AMD chips are just rubbish, both mobile and desktop. And that isn't even Haswell.
 

waltchan

Senior member
Feb 27, 2015
846
8
81
You do know how slow that A4 is?

http://www.techspot.com/review/681-amd-a10-6800k-a4-4000/page3.html

"The A4-4000 on the other hand scored just 4972pts, making it 7% slower than the A4-5300 and 31% slower than the Intel Celeron G1610."

"The A4-4000 was 4% slower than the A4-5300 and 30% slower than the Celeron G1610."

Next page:

"The A4-4000 was considerably slower at 33.80 seconds. This made it 20% slower than the A4-5300 and 40% slower than the Celeron G1610."

Power consumption:

"This means that the A4-4000 system consumed 46% more power than the Pentium G2020 and 53% more power than the Celeron G1610."

Low end AMD chips are just rubbish, both mobile and desktop. And that isn't even Haswell.
Well, I understand your point. That's why the A4-4000 has no selling value. But it can be overclocked up to 4.15 GHz on selected motherboards, which makes it great, cheap buy than A6-5400K and A6-6400K.
 

cbn

Lifer
Mar 27, 2009
12,968
221
106
Well, I understand your point. That's why the A4-4000 has no selling value. But it can be overclocked up to 4.15 GHz on selected motherboards, which makes it great, cheap buy than A6-5400K and A6-6400K.

A6-5400K does have a better iGPU though (192 VLIW stream processors vs. 128 VLIW stream processors on the A4-4000).
 

Shivansps

Diamond Member
Sep 11, 2013
3,875
1,530
136
Makes no sence, a A4-7300 cost $42 NEW and its already 4ghz.

Not even sure if you can overclock the A4-4000 at all, its not K.



I whould still pick up a G1820 for $45 instead, if no 3D is needed.
 

MiddleOfTheRoad

Golden Member
Aug 6, 2014
1,123
5
0
A6-5400K does have a better iGPU though (192 VLIW stream processors vs. 128 VLIW stream processors on the A4-4000).

Agreed -- I'd rather buy a used A6 5600K for $30 on eBay -- and get the superior GPU and totally unlocked chip.