Intel Core i3-2100 vs. AMD Phenom II X2 565

[ces]

MikeC, could you be talked into do a supplemental review and add an AMD Phenom II X4 84 to it?

Can SPCR lay their hands on a "fast" AMD CPU that costs either the same as the "slow" CPU reviewed, or an even "faster" AMD CPU so that the overall system cost is comparable to the Intel system, and then bench and review it and simply add it into the original review.

Its simply an unfair review as it stands, as the AMD buyer has more to spend on their CPU for the same build cost, and a very poor choice of CPU was made for the cost.

http://www.ebuyer.com/search?page=2&sto ... w&limit=10

As you see here, the "AMD Phenom II X2 565" is actually more expensive than the "AMD Phenom II X4 840" which will destroy the former in any multi-threaded test. I know it wont do as well as the chosen CPU in others, likely the ones that rely purely on clock speed, or even rely heavily on cache, and it certainly wont do well on the power consumption.

[NeilBlanchard]

Typo Patrol; 3rd sentence: "These dual core chips certainly don't elicit as [munch] excitement..."

[accord1999]

Going by Anandtech's bench comparison, the i3 has arguably more throughput than the X4 810:

http://www.anandtech.com/bench/Product/82?vs=289

[Mats]

Going by Anandtech's bench comparison, the i3 has arguably more throughput than the X4 810:

http://www.anandtech.com/bench/Product/82?vs=289

Yeah you need a 955 to match the i3: http://www.anandtech.com/bench/Product/289?vs=88

It costs $10 more, but the 1155 motherboards are usually more expensive than AM3 motherboards.

[jubrany]

Seeing as how the 565 was a BE, it should have made for some super easy overclocking... and the SB Core i3 is essentially not overclockable at all... I wonder how the X2 would have performed against the i3 after bumping up the multiplier.

[accord1999]

Seeing as how the 565 was a BE, it should have made for some super easy overclocking... and the SB Core i3 is essentially not overclockable at all... I wonder how the X2 would have performed against the i3 after bumping up the multiplier.

The X2 needs at least a 1GHz overclock to realistically compete with the i3.

[Hangfire]

Phenom II X2 has never been a good buy.

For typical prices and stock clocks, I agree.

For MicroCenter prices and overclocked, it used to make sense.

If cores are unlocked, it was a screaming value, and might still make sense, with a low-cost M/B.

I bought one at MC last year, a 550BE that unlocked and overclocked great, for $87 at MicroCenter. Now, that makes a lot of sense, but I know I got lucky that both cores unlocked OK.

AMD has to fix the instructions per clock cycle ratio on bulldozer to be competitive. Unfortunately, I have heard nothing about clock efficiency from their pre-release material. Its all about more threading and power. They've been beating that drum with more cores and less power while Intel cleans their clock since C2D was introduced. If Bulldozer doesn't fix the instructions per cycle deficit, I think it's over for AMD in the consumer realm, no matter how threaded their processors are. Only cloud server vendors will care.

[BillyBuerger]

MikeC, could you be talked into do a supplemental review and add an AMD Phenom II X4 84 to it?

There's plenty of performance reviews out there that you should be able to determine how the X2 compares to the X4s and then extrapolate that to the Core i3-2100. It seems a little unnecessary for SPRC to do it. I like the SPCR reviews because they give you reliable power and noise measurements that you can't get from other sites. There's way more then enough performance tests out there.

[ces]

I like the SPCR reviews because they give you reliable power and noise measurements that you can't get from other sites.

That's exactly why there is no substitute for SPCR doing it. The results would be reliable... and also comparable to the rest of the test.

[andyb]

That's exactly why there is no substitute for SPCR doing it. The results would be reliable... and also comparable to the rest of the test.

Exactly.

To extrapolate slightly, I just tried googling for some reviews of the "i3-2100" to compare the same benchmarks from different reviewers, I hope that you have better luck than me, I found that there is very little in way of comparisons between differet reviewers in what programs they use at what settings, and the system setups. Basically a meaningful comparison cannot be made, the only points of interest that anyone should come away with from comparing reviews is the overall performance, or specific benchmarks that are of interest to them.

I outright skip any numbers that show X beating Y if I never use that program - it is meaningless, and likewise I alsways skip anything like SiSoft SANDRA, as it rarely has any meaningful bearing on who wins what, its just a bunch of numbers.

Here is what I looked at and what I skipped in the Anandtech review of the i3-2100.

Sysmark 2007 - looked at.
Photoshop - skipped.
6x media encoding tests - skipped.
5x 3D Rendering tests - skipped.
PAR2 decompression - skipped.
7-ZIP benchmark - looked at.
7-ZIP compression - looked at.
7-ZIP decompression - WTF is this test, I want to look at it, I decompress 1,000 times as much as I compress.
Visual Studio - skipped.
Sorenson Squeeze - skipped.
Excel - skipped, not because I dont use it, but because no-one cares about a second here or there, I waste more time looking out of the window.
Game tests - all looked at bar WoW.
Power consumption looked at for a reference to how much cooling is needed and therefore how easy a CPU is to keep quiet.

As you can see the vast majority of testing done by Anandtech is worthless to me, and likewise to the average user, most people would be happier to save themselves some cash rather than to say "my system loads web pages 1-ms faster than yours does". The system power draw is more important in many respects, however when you look at how many years of lower electricity bills it will take to be equal to the extra money spent on the less power hungry system you will often find that it is a false economy.

Finally on the point of the gaming tests, what kind of moron would seriously spend that kind of money on that CPU, and then pair it with graphics cards that are that expensive and then run all of the tests at average resolutions. Its simply not realistic at all. I understand whay they do that, but no-one will actually do that with their system at home, people will pick a far more ballanced solution in reality - I hope.

I would finally like to draw your attention to SPCR's final score.

http://www.silentpcreview.com/article1179-page6.html

Note that the i3-2100 has only just beaten an old slow CPU that currently costs $66 less than the AMD CPU reviewed, or $77 les than the i3-2100, not to mention the board cost difference.! On that merit, one of the cheap X4's would likely do very well also, and so I suspect would the E-350 if it was added to that graph, even though it would come dead last in many of the performance tests.


Andy

[jeffy1021]

I am trying to compare the 2100 power consumption on this review to a previous one done for H67 motherboards and was wondering if a direct comparison can (or should) be made.

http://www.silentpcreview.com/article1143-page5.html

In the above review, the 2100 idles at around 20w DC (~30w at the wall).

In this review, the 2100 idles at close to 40w DC (~50w at the wall).

From what I can gather, here are the differences in the test setups:

on-chip graphics vs. discrete graphics
Notebook hard drive vs. Higher performance hard drive
H67 vs. P67 chipset
low power RAM vs. standard RAM (difference is probably negligible)

Do the above component differences really bring up the power consumption that much?

[Jordan]

Would have loved to see how the new i3 compared with older CPUs that people might be considering upgrading from. My e4300 is getting a bit long in the tooth and it would be nice to know that I can get a massive performance boost whilst also not taking a heat/power consumption hit.

Obviously I understand the test must have its limits though

[ilovejedd]

7-ZIP decompression - WTF is this test, I want to look at it, I decompress 1,000 times as much as I compress.

Probably because it's unnecessary. With decompression, I've noticed you're pretty much limited by HDD throughput.

[ces]

As you can see the vast majority of testing done by Anandtech is worthless to me, and likewise to the average user,

Yeah. I agree. They can't be unaware of that. Why do you think that is?

[ces]

Would have loved to see how the new i3 compared with older CPUs that people might be considering upgrading from.

That is a well taken point.

[andyb]

Probably because it's unnecessary. With decompression, I've noticed you're pretty much limited by HDD throughput.

Even on the SSD's they were testing with.? Although point taken in general terms, most people dont have SSD's, and even then most people who do probably dont decompress on them as they would often have those compressed files on a HDD (I am the example here, alternatively I decompress accross the network to the server, as its much quicker than decompressing locally and then moving the file).

Yeah. I agree. They can't be unaware of that. Why do you think that is?

I suspect its because 1.) They scale very well, so the end result is a pretty graph with numbers that can be compared to dozens of others. 2.) Because they were asked to do it years ago, and they keep on doing it for the sake of it. 3.) Who cares about the standard apps that show differences that are so tiny they are not worth noting e.g. Word, Excel, Outlook, Firefox.

So in summary, they go for things that show a real difference even if few people ever use them, its a test for the few and not the masses, yet they show it to the masses as an example of the performance "difference" between CPU's simply because the programs that the masses actually use show no differences. Would the average user notice the difference between the fastest and the slowest CPU in their system.? Maybe they would, but would they be prepared to pay the massive difference, no they would not. The vast majority of systems that are sold use cheaper components, not anywhere near the top end. Those who buy the (near) top end stuff know what they are buying, and why.

For me personally I look at the system that I want for the price I am prepared to pay, or for the performance I need e.g. a new game etc. For other people its a case of giving them something that is a bit better than they need now, at a sensible price.

6 months ago I built myself a system for work, I got myself an AMD tri-core system because it was a lump cheaper than a quad-core system. Clock speed was not really an issue for me due to the fact that the machine will spend most of its working time virus scanning HDD's, and 3-cores is plenty for that task - essentially I weighed up the benefits of me spending extra money on something that would in reality make no difference - I probably spent the money I saved on beer instead - now that is a real benefit


Andy

[Olle P]

What suddenly strikes me is that this review is actually fairly flawed against the Core i3!
The use of a P67 motherboard prevents the Quick Sync feature from stepping in. This makes some/many/all? of the multimedia tests provide worse results for the i3 than they really should with a proper (H67) motherboard.

Cheers
Olle

[leem]

http://www.silentpcreview.com/article1143-page5.html

In the above review, the 2100 idles at around 20w DC (~30w at the wall).

In this review, the 2100 idles at close to 40w DC (~50w at the wall).

jeffy1021, I agree and raised this in the post above, but the discussion went other ways. These are big differences, which relate to the test system and no doubt to the efficiency of the motherboard... so much so that it's hard to know the relative benefits of the CPU in terms of power draw

I'd appreciate more info...

[ces]

http://www.silentpcreview.com/article1143-page5.html

In the above review, the 2100 idles at around 20w DC (~30w at the wall).

In this review, the 2100 idles at close to 40w DC (~50w at the wall).

jeffy1021, I agree and raised this in the post above,

How did you get your link to go to a specific post?

[leem]

At top right of each post there is a small symbol next to "Posted: date...". Right-click on that and "Copy Link Location"

[jeffy1021]

http://www.silentpcreview.com/article1143-page5.html

In the above review, the 2100 idles at around 20w DC (~30w at the wall).

In this review, the 2100 idles at close to 40w DC (~50w at the wall).

jeffy1021, I agree and raised this in the post above, but the discussion went other ways. These are big differences, which relate to the test system and no doubt to the efficiency of the motherboard... so much so that it's hard to know the relative benefits of the CPU in terms of power draw

I'd appreciate more info...

Ultimately, I am trying to decide if getting a i3-2100 will save me enough power to warrant an upgrade. Over the past few months, I have been able to reduce the idle consumption of my current Q9550 system to a little over 50w at the wall. The discrepancies between the reviews are a little disconcerting. Saving 20-30w is probably enough for me to upgrade, 10w, not so much.

[lodestar]

The dual core 2100T is a 35W TDP part. But as the slowest of the Sandy Bridge Core i3s at 2.5Ghz they could be a concern about sacrificing too much performance to the extent that it might not even match your existing quad core. There are two further low TDP Sandy Bridge CPUs which have been announced by Intel, but not yet released. The first is the 35W TDP 2.7Ghz Core i5 2390T which despite being an i5 is a dual core part, but has a turbo mode. And the second is the 45W TDP Core 2.3Ghz i5 2500T which is a quad core, again with a turbo mode. I would guess at some stage in the future someone will do a group 'model T' Sandy Bridge CPU test to establish just how low system power draw could be for these three. Currently the only figure I have seen is for a 2100T/DH67CF Windows 7 system where an idle power draw of 14.5W was claimed.

[jeffy1021]

I built a 1090T system that I use for encoding now. After modifying P-states, it idles at a little more than 50w at the wall. In short, I am not looking to match the performance of my Q9550 since that system is only used for ripping movies and recording and playing back TV. If anything, I will probably sell my Q9550 system if I replace it with a 2100 or 2100T. The Q9550 has excellent resale value apparently.

[CA_Steve]

You know, all of the sandy bridge based systems (dual or quad core, low tdp or not) should consume about the same IDLE power, given the same mobo, psu and peripheral components. What we are seeing in the various reviews is the difference in power consumed by all of the non-CPU components, plus the efficiency of the mobo design's VRM circuitry, plus the efficiency of the PSU...

[leem]

CA_Steve, I agree, but the review does not make it clear how this compares to other CPUs on a like-for-like basis. Or to put it another way, it does not clarify the energy efficiency between boards, which is an important consideration for people looking at low-power systems.

Sandy Bridge represents a step-change in terms of efficiency under load, but does not appear to be significantly different from Clarkdale in terms of idle power. If we assume that many PCs sit for long periods at idle, or at low load, then the difference in energy efficiency between boards becomes an interesting story. If this was a difference of just a couple of watts then it could be ignored, but we're seeing big differences in energy consumption between similar boards, with little explanation as to why.

[ces]

CA_Steve, I agree, but the review does not make it clear how this compares to other CPUs on a like-for-like basis. Or to put it another way, it does not clarify the energy efficiency between boards, which is an important consideration for people looking at low-power systems.

Yes. It appears that with Sandy Bridge, the CPUs have gone as low as they can go at idle... to about 4 watts at idle. How close can you get to zero. What is Ivy bridge going to do, 2 watts.... no big deal. If you are chasing watts, the action has shifted from the CPU to the motherboard.

See thread here:
viewtopic.php?p=533209#p533209

[leem]

It appears that with Sandy Bridge, the CPUs have gone as low as they can go at idle... If you are chasing watts, the action has shifted from the CPU to the motherboard.

I agree, and so perhaps this needs more careful analysis in reviews. Intel boards appear to be setting the benchmark for efficiency, and MSI also have a good reputation. But it's often hard comparing between boards, rather than between CPUs, because the variety of test set-ups produces such a range of power consumption results.

Presenting an overall SPCR score is a great idea, although people may weight the criteria differently. For instance,

For the energy efficiency portion of score, we assumed that half of its operating life would be at low load (the average of idle and video playback consumption) and the other half at high load (the average consumption during our timed tests)

That's a fair assumption. But based on my PC usage, I'd work on the basis of 90% time at low load and 10% at high load, hence I give more weight to low idle power and efficiency... and therefore an ability to use an efficient pico-PSU, and therefore less noise etc.

[wtouist]

Hey you got a current system optimized for what you want, doing what you need , it is difficult to see the benefits. Always will be.

Realisticly more gains are likely in peripherals than CPU Motherboard change.
If your current system does not meet your needs, or forseeable needs, intel has the strongest cards, if only on likely resale plunge.
Where the money goes..

[jeffy1021]

I just wanted to give an update stating that I just built an i5-2400 machine and it idles at about 30 watts at the wall. This makes sense given the estimated DC wattage in the article. If I got a picoPSU, I could probably save an additional 10 watts.

i5-2400
Intel DH67BL motherboard
OCZ Vertex 2 SSD
Seagate 500GB 2.5" 5400RPM drive
BD/HD-DVD drive
Seasonic SS-300 80 Plus Bronze

[Mats]

If I got a picoPSU, I could probably save an additional 10 watts.

More like 5 W. It all depends on the components used, of course.

http://www.silentpcreview.com/article1179-page2.html
http://www.silentpcreview.com/article601-page3.html