What are the disadvantages of using an Core i5 system build over an I7? I heard something about it using only x8 / x8 lanes instead of x16? would this affect GPU performance?
Core i5 does not have hyperthreading so will not perform as well in multithreaded applications. In terms of splitting the 16 PCIe into 8 PCIe lanes on a P55 motherboard, this is only the case if you are running two graphics cards. I believe 8 lanes is perfectly adequate to operate anything upto and including a Radeon 5870. This article from TechPowerUp will give you more information. However, as Bindi has found the X58 does perform better than the P55 with two graphics cards.
Technically, 8 lanes of PCI-E will power anything, it just won't be AS powerful as a full 16x slot. If you have definite plans to have multiple graphics cards, you should buy either AMD or X58 i7. Otherwise P55 will be fine.
Are you suggesting four cores are not enough? If anything, that's excessive for 90% of people. It's marketing that makes you want an "8-thread" CPU, not need. If anything, dual core-four thread is enough for the very vast majority of people and a quad core will see then through the next few years easily. For many, their Q6600 is still "good enough" - want and need are two different things. Unless you do a job that requires hours of waiting for CPU threads, then buy a Xeon MP workstation.
So with a Core i5 processor and P55 motherboard, I will still get the max performance out of a gtx 295 graphics card for example without any bottle necks. Also I like to do video encoding and sometimes can take upto 3/4 hours on a Intel Pentium D 2.8ghz, which is old, and your saying an Core i5 should be fine just for this?
if you do a shed load of encoding work that is multithreaded then id look at a i7 tbh that said a i5 will not be slow and will easily have enough power to run any graphics card.
it'd be fine, i think the 3/4 hours can be cut down to less than one and half hours with an overclocked i5, and around 1 hour with overclocked i7. apart from those differences, you won't notice anything else.
I think you'll see greater gains than that. If something took one hour to encode with a P4 2.8D, then a OCed i5 will probably munch through it in 20 minutes. Prior to this rig, which I built almost 12 months ago, I was using a S939 FX60 running at 3.0GHz. My encoding times have less than halved, sometimes completing set tasks like encoding a full single layer DVD in one third of the time it took my FX60... and my OCed FX60 was way faster than a 2.8D How much impact the HT and triple channel is having I don't know, but I still reckon a OCed i5 will MASSIVELY beat a 2.8D in encoding times.
i have been more than happy with my core i5 rig. its more than adequate for my uses, and i do the occasional dvd to h.264 m4v encodes using handbrake it takes about 15 - 20 minutes for a 1.5 -2 hrs standard def (current oc to 3.8ghz). as for pci-e bandwidth i have a single hd4850 at the moment and it is performing as expected. watch carefully when shopping for p55 boards though, i have seen some budget boards that have 16x physical slots but only 8x and 4x signal. also i agree with 4 cores being more than enough, 90% of the programs i use barely load 1 or 2 cores let alone all 4. there still aren't a huge amount of multi threaded applications out there.
Ask not what you will ask your CPU to do today, ask what you may be asking your CPU to do tomorrow... and try and buy accordingly. i7 920 ...every time.
In tommorrow's world a Core i7 920 will be as much use as a Pentium 4 is today. Buy what suits your needs today and worry about tomorrow when it comes.
i5...or LGA1156 won't have the option of upgrading to hexcore. As bindibadgi pointed out, what do you really need more cores for anyway? another difference, your wallet will be lighter if you go for i7
Can't give you a source but I am of the same understanding as javaman, that the Intel hex cores will only be released as LGA1366.
I think you'll find this was discussed in our i5 or i7 thread a few weeks ago. The lynnfield memory controller probably coudn't feed more than four cores. Something like that. Can't remeber which site it is, but there were figures and such like to go with it.
the dual-channel DDR3 controller is known to be only able to feed 3 cores when all 3 cores are access the memory at the same time. i have heard this form an intel interview video somewhere near lynnfield's release. but the point is that it's only the theory. you will never have that many cores accessing memory at the same time, and further optimisations can be done to compensate for the dual-channel, Intel just re-introduced the L3 cache, next iteration will hopefully be better (unless they get into the Netburst mess again) the way i see it, by looking at Intel's forcing EOL of cheap x58 motherboard, the LGA1156 will be around for a long time, going to be longer than LGA1366 for the affordable market. so unless you've bought i7 920 a year ago, might as well buy the lynnfield for now, cheapest way into high performance category. (and bit-tech January buyer's guide agrees)
if your doing media encodeing or rendering that can use 8 cores can get up to 20% more rate on an i7 or more (depends if it has SSE4.2 optimisation or not) things like folding at home linux SMP do not benerfit much due to only useing SSE2 or 1 main disavantage is no more then 4 core support (X58 should have 6-8 core support intel will screw the users over and force you to buy an new motherboard for 8core and maybe even 6core as well as they did with 45nm support for core2), and 16x max PCI-e support (may be an issue for Quad SLI or 5890 in CF X mode) in the hardware part i am going to go down the route of i7 920, UD5 , 6gb ram , heatsink, Asus D2X (785 in total)
