QUESTION: How to best use AnandTech Bench to make a CPU Decision

[Starbuck1975]

I've been debating a new build for the better part of a year. Budget has always been the limiting factor.

I've always built AMD based systems, but held out for Sandy Bridge before making a final decision this time around.

I am now debating between the following three processors, which represent different price points:

Intel Core i5 2500 or 2500K
AMD Phenom II X4 (would scale specific model to budget)
AMD Phenom II X6 (would scale specific model to budget)

I've looked at the AnandTech Bench, and the SYSMark 2007 -Overall scores range from the low to high 200s.


My first question would be...how to interpret the SYSMark scoring. For example, an expensive sports car may be able to do 140mph, or is capable of ridiculous acceleration, but that performance essentially goes to waste for driving on roads with posted speed limits. So what level or threshold of scoring represents adequate performance for today's games and applications, and say the next three to five years? Is 150 an adequate threshold, or maybe 200?

My second question. I don't feel a need to wait around for Bulldozer...there is always a new processor around the corner. What I've always liked about AMD is the stability in their socket architecture. The foundation for my current build is a MOBO for which I've swapped the CPU twice. My worry in building an AM3 based system now is that I am at the end of the lifecycle for that socket architecture. Should I be thinking about, and wait, for an AM3+ MOBO build, even if I go with an AM3 CPU now (I believe AM3+ is meant to be backwards compatible to AM3 CPUs)? Of course this assumes an AMD build. I almost pulled the trigger on a Sandy Bridge build, until the release problems hit the press. Granted, they've fixed things since then, but how they handled it did not leave a favorable impression for my switching from AMD.

Final thought. I tend to hold on to systems for a long time. I typically get five years out of a build, with minor upgrades along the way. I don't manually overclock, although I will use overclock utilities if offered by the MOBO manufacturer. I use my systems primarily for home office applications, gaming, photo editing and home movie video editing.

 

[cusideabelincoln]

It sounds like Sandy Bridge is the way you should go.

At best try to get a quad core like the 2300, 2400, or 2500. They are all better options for most of the things you do than even the Phenom X6.

And if you scale down to a Phenom X4, I'd consider the dual core i3 2100 a better option at that price point. Home office, photo editing, and gaming should pretty much be better. Also the upgrade path is not as uncertain as it is with current AMD boards.

 

[Sparky19692]

I was in the very same boat about 2~3 months ago, Not saying that I made the best choice however I am thinking it was the best one for me. Old system X2 3800 (939) 2gig 3850 APG 512mg. I kept my box and PSU OCZ powerstream 520 note they are both six years old.
New system BE560 Phenom II Unlocked from 2 to 4 cores OC to 3.8 stock voltage this CPU is tried and true No promise it will unlock but most all do. 4gig ram 2x2 ASUS board with used GTX260.

Now for gaming at 1920x1200 it runs great Bad Company 2 and such Runs smooth as glass. Converting HD home videos to DVD Grand kids, Sons fights basic stuff us average people do it is simply GREAT wish I would have upgraded earlier.

Initial cost about $185 plus used video Craigslist $60.

I recently change to a new 1 t Samsung f4 spin point and now use my old 400g as storage only for $59 Microcenter.

Final thoughts there are a lot of people with no budget, but many of us are just average Joe wanting the best upgradeable bang for the buck. You can spend more than I did for the whole PC on anyone component if you wish. I just needed something snappy for general usage I truly have that, there will always be a chip right around the corner you can always upgrade the CPU in a few years to a 6 core or more memory to 8 gig. Bottom line I had the same setup and am AMAZED at how fast this is.

 

[drizek]

Pretty much the only stressful activity you do is video editing/encoding.

Assuming you are using a highly threaded editing program, I think you should wait for Bulldozer. It's only a couple more months.

If you don't want to wait, then the i5 2500K is pretty easily the best CPU you can get right now.

 

[Starbuck1975]

I'm using Sony Vegas Movie Studio HD Platinum

 

[dma0991]

I suggest that you wait for Bulldozer to be released. It is just 40 days away from release if Computex on the 7th June is the real release date.

By then you would have the option to either get Bulldozer or Sandy Bridge which might reduce in price when it competes with Bulldozer. This applies only if you are not in a hurry. If you are in a hurry then get a Sandy Bridge with a Core i7 2600 now.

 

[betasub]

^ The i7-2600 isn't one of OP's options, based on price point/budget.

Budget has always been the limiting factor.

 

[Starbuck1975]

Appreciate the feedback, but would still like to address my original question around benchmarks, as that will ultimately drive my decision.

A new processor is always around the corner, with the promise of adjustments to pricing. I've already delayed this build once in anticipation of Sandy Bridge, and maybe it does make sense to wait for Bulldozer at this point, but ultimately the sweet spot of price to performance will drive my decision.

Having said that, can anyone enlighten me on how to interpret the SYSMark 2007 benchmarks? Is it a suitable benchmark for making a general but educated decision around overall bang for the buck? Similarly, what is the minimum score I should consider in terms of performance?

I've never purchased the latest and greatest, nor do I need to have the fastest machine on the block. If I can play current games, have an upgrade path to swap out say a CPU or GPU to get a performance boost in the foreseable future, and can run my applications with acceptable lag, I'm happy.

 

[evilspoons]

With a few exceptions (that I can't think of right now), not many benchmarks are actually used a a minimum threshold "pass/fail" kind of judgement. This is especially true due to the multitude of variables in each test (motherboard, RAM, hard drive speed, video card, ambient temperature, wind direction, star signs, etc.)... tests like this are more of a relative indicator - i.e. "I can see the i3 2100 is clearly better than blah blah because I trust that Anandtech did a good job of minimizing all the variations in testing while they were benchmarking."

Pretty much every processor you can buy today is good enough for office applications, web browsing, email, etc., and they can all do photo editing and video editing. Cheaper processors will just make you wait a longer for the video editing to complete its final rendering/encoding stage. How important is this to you?

A quick & dirty way to interpret benchmark results would be to run a test of your own (output your own video) and compare it to your processor in a similar benchmark online. You can then take the ratio of whatever you're looking to upgrade to and get a very rough estimate of how much faster that computer might be, assuming the benchmark scales linearly.

(For example: you render out a video in Adobe Premiere Pro and it takes 18 minutes. You find your processor in Bench doing something in Premiere Pro and it takes 5 minutes. You want to upgrade to another chip that takes 3 minutes, and scaling on the benchmark seems roughly linear... therefore your 18 minute test will probably take 10 minutes on this new system. This is by no means particularly accurate but it does give you a general guideline to plant the benchmark results in reality vs just looking at a big list of numbers. You can then think to yourself whether spending $xyz is worth it to have what you were doing complete 8 minutes faster.)

On the other hand, you mentioned gaming. If games are important enough to change your decisions, you're going to have to find all potential processor purchases, find the games you play, and pick whichever one is highest up on the most benchmarks. Again, actual numbers don't matter because there are so many variables that will change on your personal system versus the benchmark system, you just want the one with the most potential.

 

[Arzachel]

Imo, this is the worst time to upgrade a CPU. When Bulldozer comes out, Intel should either drop prices or up the clocks on SB and Phenom II prices should plumet. Currently, a Phenom II X4 should be enough for gaming, but that might change, when Battlefield 3 comes out.

 

[aigomorla]

In encoding Hyper threaded cpu's own...

Unless the encoding software was poorly designed not to use parallelism.

So id say get the i7 2600 /w Hyper Threading if your an encoder.

I dont know how bulldozer does... but in all honesty im not holding my breathe on them being faster then the 2600K or having an overclocking window as high as one.

 

[tweakboy]

That Phenom 6 core is slower then a 2500, the K stands for its unlocked and ready to be OCed. Alto AMD has 6 cores, the 8 threads you get with a 2500 or 2600K will beat the AMD easily..

I've been debating a new build for the better part of a year. Budget has always been the limiting factor.

I've always built AMD based systems, but held out for Sandy Bridge before making a final decision this time around.

I am now debating between the following three processors, which represent different price points:

Intel Core i5 2500 or 2500K
AMD Phenom II X4 (would scale specific model to budget)
AMD Phenom II X6 (would scale specific model to budget)

I've looked at the AnandTech Bench, and the SYSMark 2007 -Overall scores range from the low to high 200s.


My first question would be...how to interpret the SYSMark scoring. For example, an expensive sports car may be able to do 140mph, or is capable of ridiculous acceleration, but that performance essentially goes to waste for driving on roads with posted speed limits. So what level or threshold of scoring represents adequate performance for today's games and applications, and say the next three to five years? Is 150 an adequate threshold, or maybe 200?

My second question. I don't feel a need to wait around for Bulldozer...there is always a new processor around the corner. What I've always liked about AMD is the stability in their socket architecture. The foundation for my current build is a MOBO for which I've swapped the CPU twice. My worry in building an AM3 based system now is that I am at the end of the lifecycle for that socket architecture. Should I be thinking about, and wait, for an AM3+ MOBO build, even if I go with an AM3 CPU now (I believe AM3+ is meant to be backwards compatible to AM3 CPUs)? Of course this assumes an AMD build. I almost pulled the trigger on a Sandy Bridge build, until the release problems hit the press. Granted, they've fixed things since then, but how they handled it did not leave a favorable impression for my switching from AMD.

Final thought. I tend to hold on to systems for a long time. I typically get five years out of a build, with minor upgrades along the way. I don't manually overclock, although I will use overclock utilities if offered by the MOBO manufacturer. I use my systems primarily for home office applications, gaming, photo editing and home movie video editing.

 

[Syzygies]

I don't manually overclock, although I will use overclock utilities if offered by the MOBO manufacturer.

What keeps us all young is when the world changes enough that we have to rethink our policies from scratch. Overclocking used to be harder.

You're not going to fairly assess the 2500K if you don't consider benchmarks for a conservative, manual overclock. Compared to an automatic overclock, you can take longer to do a better job. If you don't, you're leaving money on the table.

I am stunned at how quiet my 2600K overclock is. Now I'm in a nine fan case that cost more than many people's first car, with a cooler that weighs more than I tend to eat in a day, but still, this is new for me.

I am also stunned at how easy it is. At first, it seemed one just needs to reset the Turbo Ratio and pray. At the other extreme, there are a dozen settings people recite that you should set on faith. Somewhere in the middle,

* Set Load Line Calibration to High
* Set a modest Offset Voltage
* Set a modest Turbo Ratio

and hammer the hell out of each choice, keeping a chart like you're playing minesweeper.

There are plenty of people who can read my chart (at top of post) and say I wasn't so lucky on my chip, but I'm still thrilled. This 2600K smokes a Q6600, which dimmed my street lights when I overclocked it.

Just don't buy into the macho game of maximum overclock. A conservative overclock is dead simple modulo some testing, and reveals exactly how much performance you can exact from your chip.