Battle At $140: Can An APU Beat An Intel CPU And Add-In Graphics?

Page 4 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Toggle sidebar Toggle sidebar
Status
Not open for further replies.
I don't understand these comparative articles. As far as upgrade path, you can stick a $70 6670 card in the AMD system and you now have a powerful crossfire setup. You can't beat that as far as a low-cost upgrade path. That's really one of the major points, so I don't understand why Tom's keeps missing it.
 
[citation][nom]ddelrio[/nom]I don't understand these comparative articles. As far as upgrade path, you can stick a $70 6670 card in the AMD system and you now have a powerful crossfire setup. You can't beat that as far as a low-cost upgrade path. That's really one of the major points, so I don't understand why Tom's keeps missing it.[/citation]

Because There's no point. A $140 APU plus a $70 graphics card leaves the Intel system with room for a $140 graphics card such as a 6790 with GDDR5, and it would utterly obliterate the APU/CrossFire setup.
 


Then why did not you use a 1800Mhz ram on the A8 and show whether there is any performance difference?
 
Cleeve, i think you have missed the point, we want to know if the ram controller was overclocked, and if this would close the gap, maybe 8GB would help, but does the system need more ram? we dont know u did the test, would faster ram help the AMD system? not more just faster? intel can't take advanteg of this faster ram at this price, but maybe with a $80 board it could, and the extro needed for the fast board could give the extra money for more ram in the amd system,

IE ddr3 1600 4gb with better MB in the intel set up
with ddr 1800 4/8gb(cost being the decided) with the standard MB

the main issue i have is the lack of clarity in the article, which doesn't go into how AMD has stated the memory timing in the APU's is crucial to performance as the GPU uses the system memory so picking the best u can can make a massive difference.
 
Nevermind AMD vs. Intel; anyone who is even half awake knows the story. What I'd really like to see is [standard version] vs. [low power version].
I recall testing a 4850e and finding it too weak for my uses, bottlenecking even a HD4670. I've been trying to interpolate / extrapolate three or four recent articles pairing various CPUs and GPUs, and still not getting a clear picture.
What I'd really like to see is something like i3-2120 vs. i3-2120T, X3 440 vs. X3 415e, G620 vs. G620T, i5-2400 vs i5-2400S.
As you move up the GPU hierarchy chart, where does an i3-2120T become a bottleneck, vs. an i3-2120? Can you build a viable low-power gamer with an i3-2120T and a HD7750? Does an i3-2120T really use less power than an i3-2120 given it takes longer to complete the same tasks? These low-power versions are often more expensive than their thirstier siblings. Why? Is it justified?
 


To me, the point was seeing what would happen with the exact same $$ invested into the CPU and graphics subsystem. We said that, we did that.

If you guys want the point to be optimizing the AMD system for best possible performance for the price cap of the system, then we should do that for the Intel system too. No?

You guys are nitpicking. If you think 1866 RAM and a cheapo Intel board will change the results significantly, I think you're stretching.

I wish I had the time to redo the tests with faster RAM and a cheapo Intel board but I work on tight deadlines. But if you guys want to chime in and support the idea I can lobby to retest the whole thing in a future article.

 
I don't know why people are getting so worked up over this.

What we have is an APU vs a CPU/GPU combo. It is not surprising the CPU/GPU combo won out in the end when it comes to graphics. Of course the dedicated GPU did better in games.. that is why we all have them, right? That said, the APU didn't do that badly and came pretty close. In applications, the AMD offering did very well, which is not surprising, considering you have a 3.0GHz quad vs a 2.6GHz dual.

What to walk away from this article? The newer APUs have nice computing power and the graphics performance is much better then past motherboard integrated solutions. But if gaming is your forte, then you are better served going the dedicated GPU route (which is how it has always been).

 


True, As they said in conclusion this is very ideal for laptops
 
Earlier i had some faith in APUs as cheap gaming alternatives, but Intel clearly has the lead w.r.t. gaming on this one.

Sandy Bridge ROX!
 
[citation][nom]Cleeve[/nom]So you guys want motherboard and RAM included in the cost comparison?And you want me to do the best I can in the same price total for both sides?If the rest of the platform gets included in the cost you're shooting the AMD system in the foot. If your argument is that Llano *needs* 8 GB of superfast ram to make a decent budget system, then we should optimize the Intel system for cost, too. That means cheap RAM, and 4 GB will do because the graphics card has its own dedicated graphics RAM on the card.Here's the realistic scenario:OPTIMIZED AMD SYSTEM:-------------------------------Cheapest FM1 board: ECS A55F-M3 - $55Cheapest 8GB 1866 RAM: Team Xtreem LV 8GB (2 x 4GB) - $56A8-3870K: $140-----------------$251(actually, this might be underpriced because I believe FM1 might require the more expensive A75 chipset for dual-channel 1866 RAM. So the board cost would go up another $30. But for now, let's give AMD the benefit of the doubt)OPTIMIZED INTEL SYSTEM:---------------------------------Cheapest 1155 board: Jetway GTI61AG3: - $50Cheapest 4GB 1600 RAM: Crucial Ballistix Smart Tracer 4GB (2 x 2GB) - $25Pentium G620 - $70HIS H577FL1GD Radeon HD 5770 1GB 128-bit GDDR5 - $105-----------------$250So do you guys really want me to go ahead and build this comparo? I don't know what you think it will prove, because app results will be the same and Llano's graphics will get an insignificant boon from the faster memory, while the pentium system will get even more of a graphics boost from the 5570.Or do you suggest I completely optimize the AMD system and then cripple the Intel's by not making use of it's strengths when it comes to cost savings? DUMB INTEL SYSTEM:--------------------------Cheapest 1155 board: Jetway GTI61AG3: - $50Cheapest 8GB 1866 RAM: Team Xtreem LV 8GB (2 x 4GB) - $56Pentium G620 - $70HIS H667FS1G Radeon HD 6670 1GB 128-bit DDR3 - $75-----------------$251But who in his right mind would choose this over the first option? And the sad part is, *Intel results won't change one iota from the original test because motherboard cost does not impact performance*. If anything it'll probably increase a bit in apps, because despite the fact that Intel doesn't scale as quick with faster memory it *will* scale a little.On a side note, nobody seems to be commenting how AMD got the benefit of the doubt BECAUSE I NEVER OVERCLOCKED THE DISCRETE CARD ON THE INTEL SYSTEM, by the way. So to be fair, shouldn't I overclock the discrete Radeon also?Or do you guys want me to avoid that? Because it seems to me you're making every little nitpick you can about the AMD system while you're happy to leave the Intel system with unrealistic handicaps.[/citation]

Hi Cleeve, first of all we respect your work, but there is missing fragments.

And as I say before, before a week i buy 3870K for my friend with 4GB 2000mhz Corsair ram. I will say that 3,5gb, which is left is more than enough and 8GB with embedded graphics is useless. So you can get a pair of 1866 or 2000mhz 4gb kit and retest again! And when you show overclocked performance, PLEASE overclock memory controller, because its crucial for the GPU. And please stop saying that we whant to cripple Intel, because AMD is now crippled.

So if you want to see how DDR3 scales on this APU - http://www.anandtech.com/show/4476/amd-a83850-review/4, check this one.

We are on other side and we have systems like this - this system DEFENETLY can more (15% or more)!
 


You're in the minority, the last time I used 4 GB in an Intel/AMD comparo everyone screamed that I wasn't using 8 GB.



I call BS on that. The Intel system was never overclocked at all, and the 6670 has headroom. Hell, the G620 has headroom, too, if you know how to push it.

I just didn't bother because I thought the point has been made. I think to reasonable human beings it has. But if we're going to strive for the ultimate, impeccable perfection in fairness I will have to overclock the Intel system as well.

 
[citation][nom]Cleeve[/nom]To me, the point was seeing what would happen with the exact same $$ invested into the CPU and graphics subsystem. We said that, we did that.If you guys want the point to be optimizing the AMD system for best possible performance for the price cap of the system, then we should do that for the Intel system too. No?You guys are nitpicking. If you think 1866 RAM and a cheapo Intel board will change the results significantly, I think you're stretching.I wish I had the time to redo the tests with faster RAM and a cheapo Intel board but I work on tight deadlines. But if you guys want to chime in and support the idea I can lobby to retest the whole thing in a future article.[/citation]

And yes we will highly appreciate if there is way to redo the test, to say in same money gap, with the suggestions that i make in previous post...No one knows everything...
I can say that i'm real overclocking enthusiast, and i'm favoring to AMD for many years, and i truly know how to squeeze maximum of their CPUs. And this APU truly can do MORE. On PII 555(x4) and 1333mhz CL7 i got 16gb/s in Sandra, for which you may be say "its impossible"?? Just do your best next time with ram and mem cntrlr, and this will pay-back to you.

Thanks in advance for your efforts, and we wait for your come back!
 
[citation][nom]Cleeve[/nom]You're in the minority, the last time I used 4 GB in an Intel/AMD comparo everyone screamed that I wasn't using 8 GB.I call BS on that. The Intel system was never overclocked at all, and the 6670 has headroom. Hell, the G620 has headroom, too, if you know how to push it. I just didn't bother because I thought the point has been made. I think to reasonable human beings it has. But if we're going to strive for the ultimate, impeccable perfection I will have to overclock the Intel system as well.[/citation]

Not a problem at all to squeeze Intel too :) This will show that low-end PCs can do the Gaming at 720p and even 1080p (Wow for sure), and can open questions about high, unreasonable prices of some components... :) May the force be with you :)
 
Cleeve pls don't waste your time on these teenage fanbois.
Your article just proved what I've been saying since the launch of Llano: Llano only makes sence in a notebook. We've all seen in Anands article that 1866 memory improves performance a little bit but even he concluded that DDR3-1600 seems to be the sweet spot because of the performce/cost ratio.
Time by time it amazes me how strongly some people want to support Llano no matter it's deficiencies. When OC'd, power usage is just through the roof. Nobody in their right mind will use a $50 motherboard for a CPU that pushes close to 300W. So forget about that option. (and let's not forget the expensive cooler you'll need to dissipate +250W)
What's more amazing is that all major websites are now being called biased because of their conclusions about Llano.
I saw one guy here say that power figures for the Intel/GPU combo can't be right because he's counting the TDP numbers! He almost went as far to call you a cheater. When someone makes comments like that, why do you even bother replying. It think we've all seen enough review where a 65W Pentium barely breaks the 45W barrier.
And people suggesting that in this time and age where the memory controller is in the CPU, motherboard 'performance' is something that matters. Huh? I haven't seen more than 5% difference between motherboards for a long time now. Specially on the Intel side.
People saying that you're not comparing APU's. Well that's the whole point of this article: to see if it makes sense to stick with an APU on a desktop system or use a cheaper cpu+gpu. On a desktop, why wouldn't you make use of a discrete videocard? Specially in this case where the Intel+GPU combo is faster and uses less power. What's there to think about? Even ITX cases allow you to install low profile videocards like the HD6570.
 


Wow and MMOs maybe, but not demanding games like Metro 2033 or Battlefield 3. You'd need a 6670 GDDR5 for that at the very least.


I won't use 4 GB because we'll end up with the same situation. Everyone will complain that AMD's integrated graphics isn't given enough RAM.
 
Some of the AMD chipset motherboards have onboard video that can Crossfire with discrete graphics. I would assume those onboard chips could Crossfire with an APU as well. Is this true? How would this affect performance in, say, an HTPC setup?

Edit: And of course, I meant Crossfire, not SLI. My bad.
 
There's just no winning Cleeve. :/

You did this article (I'm assuming?) specifically to address issues raised in the sub-$200 CPU roundup by AMD fans, and they still can't accept it, even though the results are almost exactly what most un-biased people would have predicted.

I think they are under the impression for some reason that everyone who's not a crazy AMD fan it rooting for them to fail, when the reality is more along the lines of just being disappointed because we with they'd step it up, but they just don't.

Actually with as much application testing as you did, I think you painted the APU in a fairly favorable light in a decent number of tests, comparative to how often you'd actually likely see it excel on a daily basis compared to the Intel build.

Unfortunately no matter how many times it's stated that motherboard price has virtually zero performance impact it's constantly mentioned, maybe perform the next few tests with a $60 Intel board and a $200 AMD board just to drive it home? :/

I thought you did a good job answering the legitimate questions at least. It's pretty much just AMD fan excuses at this point. To bad they can't direct that energy toward pressuring AMD to get competative again.
 
[citation][nom]Cleeve[/nom]The Intel system was never overclocked at all, and the 6670 has headroom. Hell, the G620 has headroom, too, if you know how to push it. I just didn't bother because I thought the point has been made. I think to reasonable human beings it has. But if we're going to strive for the ultimate, impeccable perfection in fairness I will have to overclock the Intel system as well.[/citation]

I'd really like to see how much you're going to pull out of the Intel system, with a completely stable OC. If we're getting into changing the bus speeds and full stability isn't a concern, the AMD system could function on many FM1 boards with 133 Mhz, bringing the "potential" to nearly 5Ghz (An unrealistic potential on air, granted) while also OCing the iGPU. No no, in terms of OC potential, the winner is obvious.

I think people just wanted to see the results of how the better memory would have performed, even if it was a separate test bar and you had to note that config cost slightly more. It isn't as if using better memory would double the results. It IS worth noting however, that your APU results are reduced by using that memory.

I mean, as far as value goes, you could also get an A8 3850 for 10 dollars less and get nearly the same performance as the stock 3870. One of the graphics disabled models like the 631 might have been interesting paired with a discrete too, I don't know if anyone ran that comparison yet. How would an A6 3500 fare in these waters for $80, considering the graphics side is fairly close to an A8 and the CPU is still a triple core?
 
The reality is you have to look at the total system price. Not just the price of select components. This is a good step to see toms actually looking for performance @ dollar value but this needs to take into account all components, not just the ones they think should matter.
 
[citation][nom]cangelini[/nom]Here's our own take on memory performance and A8: http://www.tomshardware.com/review [...] 975-6.htmlThere is bandwidth to be had from going to 1866 MT/s. However, game performance does not scale linearly. In fact, 1600 MT/s is indeed the sweet spot. Hope that helps to answer your question peroludiarom! All the best guys,Chris[/citation]

I and can't stay and say that is OK again - did you guys clock the memory controller - because this is one of most-bottlenecking part in Pehnom-based(also Liano) processors ? Did you try to clock by base clock?

With my full of software windows, 3 mounts old (including visual studio with all .net patches), with ddr3 1333mhz CL7, and just by rising the multiplier of my PII 555, i just got 15.2gb/s Sandra Mem band. In which of your tests you met this bandwidth with such ram speed, because i can't remember? And this is possible only because of meme controller overclock to 2800mhz.
So please show what settings do you you or use one that we suggest.

Just for you i upload some screens (2 old, and 2 from about 10 minutes), to show you how memory controller scales at 2800mhz (we all know that Liano uses PII architecture):

PII 555 (x4)
Corsair 1600mhz CL9(clocked to 1333 CL7) 8GB
5850 Toxic at 900mhz/1210mhz
Asus M5A99X
Old :
http://i40.tinypic.com/34863qh.png
http://i39.tinypic.com/2z5ksv5.png

New:
http://i43.tinypic.com/2zi1vk2.jpg
http://i42.tinypic.com/i4ec08.png
 
[citation][nom]cangelini[/nom]Here's our own take on memory performance and A8: http://www.tomshardware.com/review [...] 975-6.htmlThere is bandwidth to be had from going to 1866 MT/s. However, game performance does not scale linearly. In fact, 1600 MT/s is indeed the sweet spot. Hope that helps to answer your question peroludiarom! All the best guys,Chris[/citation]

And once again - i wake up every morning with your site - i love it for years, but when you miss something i have to say it.
No one knows everything - i know that is hard to write articles in tight deadlines, but we want to see that when comes to overclocking, you done it write. For Intel platform you do this for years, but for Phenom family i always watch how you missing the one of biggest bottlenecking parts - memory controller. :)
 
[citation][nom]peroludiarom[/nom]And once again - i wake up every morning with your site - i love it for years, but when you miss something i have to say it. No one knows everything - i know that is hard to write articles in tight deadlines, but we want to see that when comes to overclocking, you done it write. For Intel platform you do this for years, but for Phenom family i always watch how you missing the one of biggest bottlenecking parts - memory controller.[/citation]

For sure--in my particular case (for the A8 launch story), the motherboards had massive teething problems. I did *not* do any IMC overclocking. This page tells that story more completely: http://www.tomshardware.com/reviews/amd-a8-3850-llano,2975-7.html.

However, given the fact that Sandra shows a notable gain in bandwidth moving to 1866 MT/s suggests we're not seeing a bandwidth bottneck per se. The iGPU simply isn't taking as effective advantage of the move from 1600 to 1866 as it did from 1333 to 1600.

With all of that said, Thomas is currently working on a memory round-up for FX. If you think an exploration of memory controller overclocking might be useful to the community, I can suggest that he do a comparison of several different architectures afterward?

All the best, and thanks for reading peroludiarom. The feedback is always read and appreciated =)

Best,
Chris
 
[citation][nom]cangelini[/nom]For sure--in my particular case (for the A8 launch story), the motherboards had massive teething problems. I did *not* do any IMC overclocking. This page tells that story more completely: http://www.tomshardware.com/review [...] 75-7.html.However, given the fact that Sandra shows a notable gain in bandwidth moving to 1866 MT/s suggests we're not seeing a bandwidth bottneck per se. The iGPU simply isn't taking as effective advantage of the move from 1600 to 1866 as it did from 1333 to 1600. With all of that said, Thomas is currently working on a memory round-up for FX. If you think an exploration of memory controller overclocking might be useful to the community, I can suggest that he do a comparison of several different architectures afterward?All the best, and thanks for reading peroludiarom. The feedback is always read and appreciated =)Best,Chris[/citation]

:) Hi again, for FX i can't say that there is too much to explore :) The first wave is obvious with a lot of problems. But i can suggest to include Phenom and Liano family (i know that Phenom is near its end) with memory scaling by memory controller and by bus overclock. And you can use fast RAMs as you done for high end Intel processors (i see speeds around 2400mhz on Intel benches). This will breath some live to this old architecture.

One interesting path for upgrade (which will pop up after above) is what i done - buy good MB and good ram (mine is peace of crap, but i understand it too later) and cheapest BE Phenom processor. Do some good overclock of all parts (here i will note that in order to achieve good speeds of ram and mem controller you should up voltages of path between CPU and RAM also), and wait for second gen FX.

Also i again will state that with mature MBs for Liano, if you have time - make some more tests with high RAM speeds and high Mem controller.

And I' stop writing because i will go for spammer :) If you have any further questions, i will love to answer you questions by email. Bye :)
 
Status
Not open for further replies.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom