Hardware Intel Core i7 CPU and Platform Value

[Bindibadgi]

http://www.bit-tech.net/hardware/200...i7-cpu-value/1

With everything else tested, we put our math cap on and cracked out the calculators to work out which platform is the best punt for your pound. Does the supreme Core i7 performance warrant the extra cost it demands? Read on to find out!

[Teq]

I would have liked to have seen the gaming averages in the conclusion completely separate from the 2d/synth stuff, I should imagine that to most people that frequent bit-tech that is where their priorities lay

[Yardstick]

Agreed on separating out gaming from other 2d stuff. When I am gaming performance is everything. When I am decoding DVDs, zipping files etc it's usually running in the back ground and performance is not a major issue.

[Teq]

...instead of just bitching I decided to do something about it and have very quickly created an average normalised gaming performance chart ( top down ), nice to see the Q6600 still near the top

( 146.0 ) Core 2 Duo E8400 OC (2x4.0GHz, 1,760MHz FSB, DDR2)
( 138.3 ) Core 2 Quad Q6600 OC (4x3.6GHz, 1,600MHz FSB, DDR2)
( 134.6 ) AMD Phenom X4 9850 OC (4x3GHz, 2.0GHz HTT)
( 134.1 ) Core 2 Duo E8400 OC (2x4.0GHz, 1,760MHz FSB, DDR3)
( 131.8 ) Core 2 Duo E8500 (2x3.16GHz, 1,333MHz FSB, DDR2)
( 125.2 ) Core 2 Quad Q6600 OC (4x3.6GHz, 1,600MHz FSB, DDR3)
( 123.8 ) AMD Phenom X4 9950 BE (4x2.6GHz, 2.0GHz HTT)
( 123.3 ) AMD Phenom X4 9750 (4x2.4GHz, 1.8GHz HTT)
( 121.7 ) AMD Phenom X4 9850 BE (4x2.5GHz, 2.0GHz HTT)
( 118.9 ) Core 2 Duo E8500 (2x3.16GHz, 1,333MHz FSB, DDR3)
( 103.8 ) Core 2 Quad Q6700 (4x2.66GHz, 1,066MHz FSB, DDR2)
( 101.8 ) Core i7 920 OC (4x4GHz, 1,600MHz DDR3, 3.6GHz QPI)
( 96.3 ) Core i7 920 (4x2.66GHz, 4.8GHz QPI, SMT enabled)
( 96.3 ) Core 2 Quad Q9550 (4x2.83GHz, 1,333MHz FSB, DDR2)
( 95.4 ) Core 2 Quad Q9450 (4x2.66GHz, 1,333MHz FSB, DDR2)
( 90.0 ) Core 2 Quad Q9550 (4x2.83GHz, 1,333MHz FSB, DDR3)
( 88.1 ) Core 2 Quad Q9450 (4x2.66GHz, 1,333MHz FSB, DDR3)
( 84.9 ) Core 2 Quad Q6700 (4x2.66GHz, 1,066MHz FSB, DDR3)
( 71.5 ) Core i7 940 (4x2.93GHz, 4.8GHz QPI, SMT enabled)
( 58.2 ) Core i7 965 (4x3.2GHz, 6.4GHz QPI, SMT enabled)
( 57.4 ) Core 2 Quad Q9650 (4x3.0GHz, 1,333MHz FSB, DDR2)
( 56.1 ) Core 2 Quad Q9650 (4x3.0GHz, 1,333MHz FSB, DDR3)
( 55.1 ) Core 2 Extreme QX9770 (4x3.2GHz, 1,600MHz FSB, DDR2)
( 51.0 ) Core 2 Extreme QX9770 (4x3.2GHz, 1,600MHz FSB, DDR3)

[sandys]

Nice round up, pretty much came to the conclusion that the X4 was best buy for me on price when I upgraded despite not necessarily being a headline performer, especially as I had and SLI graphics setup, once thats factored in the Intel platform is mucho expensive.

In those benchmarks whilst you've included OC'ed results it seems to me at least that you may have not got the best performance out of the X4, for example you can up the NB speed as well as HTT and it appears you've done neither the jump in NB speed and HTT from 1.8/2Ghz to 2.2/2.4/2.6Ghz (depending on what you get) is quite beneficial, seems you've only upped the multiplier for clock speed though giving no additional cache memory bandwidth to the AMD platform so it could perform even better than it does in these charts whereas the Intel OCs do get that bump.

OK the increase is probably little more that a few % from the benchies I did when OC'ing but if you're going to include Oc'ed results from each platform at least exploit all the easy stuff, I'm running on a Nvidia chipset with x64 which is supposedly poor for Oc'ing and have no problems cranking these things up.

[Er-El]

Looks like the Q6600 is the biggest value quad core CPU out there after all.

[Tim S]

for gaming the CPU makes little-to-no difference once you are playing games at GPU limited settings. That's not to say you can't get away with a rubbish CPU and a flagship graphics card, but if you're a gamer, a q6600 at stock is generally plenty of performance. There's no difference in peed between an e8400 and an i7 920 in most games. That's why I'd be against separate gaming scores.

Modern games are, for the most part, NOT CPU-limited unless you play at 640x480 or without AA/AF

[Teq]

If there is no difference between CPUs when benchmarking games - why do they even appear in a value roundup? For the most part I agree with you Tim, but separation for gamers will at least show those of us who are uninterested in basic platform performance that there is little reason to upgrade from what we are currently using...

[Aterius Gmork]

Originally Posted by Er-El

Looks like the Q6600 is the biggest value quad core CPU out there after all.

I don't think you really can tell by this comparison. Don't you think, 3.6Ghz on a Q6600 is a bit over the top guys? I certainly wouldn't push my CPU this far, because it might be possible but probably not ideal for extensive daily use for a few years?

[Bindibadgi]

Originally Posted by Aterius Gmork

I don't think you really can tell by this comparison. Don't you think, 3.6Ghz on a Q6600 is a bit over the top guys? I certainly wouldn't push my CPU this far, because it might be possible but probably not ideal for extensive daily use for a few years?

I had someone complaining 3.6GHz was too slow and he ran his at 4Gs all day. Others thought 3.2 is too slow.. I just found the stable maximum for our CPU - you can never make a set of results to suit everyone, after all, we used X48 instead of P45... etcetc

If my 2.4 did 3.6, I'd run it at 3.6. If it died, by that time I'd upgrade.

Originally Posted by Teq

If there is no difference between CPUs when benchmarking games - why do they even appear in a value roundup? For the most part I agree with you Tim, but separation for gamers will at least show those of us who are uninterested in basic platform performance that there is little reason to upgrade from what we are currently using...

Because there is some difference, and we set out to ascertain the BS or non-BS about "of our new engine is multithreaded" - in which case, the Core i7s would be loving them. Crysis was apparently "made for quad cores" remember, yet the E8400 loves it.

Also, Teq, the people above you are complaining they ONLY want to see gaming benchmarks. Soooo... we try to build a platform for everyone.

[djDEATH]

this is very encouraging, what it means is that although my AMD machine is never gonna be at the top of the benchmarks, it shows that it still holds very strong value for money, and that a comparable Intel machine WOULD cost me more.

I like it, and i have decided against my initial decisions too get meself a 9850 BE X4. Nice

[barrkel]

IMO you should have plotted a scatter-graph of price-performance when doing these comparisons, rather than percentage-based bar-charts.

The thing about price/performance tradeoff is that different people are willing to make different compromises, particularly when they can see something better that's close in price. However, your percentage-based charts hide this fact, as something slow and very cheap will appear to be similar to something expensive and very fast.

Price/performance scatter graphs are easily read using a line of negative slope (i.e. like \). The slope (steepness) depends on personal preference (i.e. willingness to pay for performance). The line is slid out right as far as possible until the last point is on the line. This is then the optimal price/performance value for your preference.

The crucial advantage is that points close to the selected point are clearly visible and, because the graph hasn't lost information, the exact tradeoff - more or less absolute performance with less or more value for money - is simultaneously visible.

[Burnout21]

Q6600 is about as close as we can get to the old value for money performance that the AMD Athlon 2500+ Barton offered, i had both types of barton, the XP-M was a beast and out perfromed my friends brand new Athlon 3400+ SKT 754, ah those were the days!

Love my Quad, cant see upgrading it for a while yet, but every now and then i want to change it. Its the upgrade bug!

[fathazza]

OK humble readers, it's time to crack out those calculators and pop on that math hats

I think we need our English hats too

interesting article

[Ending Credits]

I'd say with the new Denebs coming 9850BE + AM2+ >= Q6600 S775.

3.0GHz on a Phenom 9850 is fairly commonplace.

[Bindibadgi]

Originally Posted by fathazza

I think we need our English hats too

interesting article

those hats

that hat

our hats

OR - you could wear multiple math hats for SLI or Hyperthreading hatting action

[Skill3d]

is that a hattrick?

/ot guess ill buy the q6600.... except if I win the lottery offcourse

[oasked]

Interesting to note that the Q6600 has gone up in price AGAIN! Its also out of stock by the looks of it on eBuyer.

[metarinka]

interesting article. I too put in my vote for a scatter plot graph.
I think this graphs over represents the 920 system as while the chip itself is fairly priced the mobo's are insane as well as memory, and it sounds like it will be awhile to come before we see <$200 mobos for the core i7
I've been doing the math and while I like the core i7 and the performance I think you can get beef up a system in many other ways if you include the extra cost of the mobo and RAM

[Bindibadgi]

Originally Posted by barrkel

IMO you should have plotted a scatter-graph of price-performance when doing these comparisons, rather than percentage-based bar-charts.

The thing about price/performance tradeoff is that different people are willing to make different compromises, particularly when they can see something better that's close in price. However, your percentage-based charts hide this fact, as something slow and very cheap will appear to be similar to something expensive and very fast.

Price/performance scatter graphs are easily read using a line of negative slope (i.e. like \). The slope (steepness) depends on personal preference (i.e. willingness to pay for performance). The line is slid out right as far as possible until the last point is on the line. This is then the optimal price/performance value for your preference.

The crucial advantage is that points close to the selected point are clearly visible and, because the graph hasn't lost information, the exact tradeoff - more or less absolute performance with less or more value for money - is simultaneously visible.

Nice idea! While impossible to do in our graphing engine I could screenshot Excel (like I do in PSU reviews).

So you're saying you have relative performance (or, absolute?) on the X axis, then price on the Y?

Originally Posted by Skill3d

is that a hattrick?

Hahaha