2600K or i7 970

Coz you had too much traffick that's why. The raid controller with a pair of samsung hdd that's cheap can even give you a huge boost as the traffick is not over the qpi like you would do with a onboard that's a bottleneck plus more traffick over the qpi or DMI. Again it shows you the importance of a disk setup and balance pc in video editing coz the better cpu got beaten by a lower clocked cpu with a better disk setup and due to its bigger bandwidth. The reason why the 2600k don't have that card coz of its limitation it carried over.

The problem here is that you are correct - on a theoretical basis. On a practical basis, however, even the very fastest of the current GeForce GTX series GPUs do not take anywhere near full advantage of even PCI-e x8 bandwidth, let alone x16. In fact, a single GTX 580 barely saturates PCI-e 2.0 x4 bandwidth - and Premiere Pro CS5 does not support SLI at all (only one of the two GPUs is used). Hence, one can use a hardware RAID card in an LGA 1155 platform without losing any performance because of this fact about current GPUs as long as the secondary PCI-e slot runs in x8 mode.
 
The problem here is that you are correct - on a theoretical basis. On a practical basis, however, even the very fastest of the current GeForce GTX series GPUs do not take anywhere near full advantage of even PCI-e x8 bandwidth, let alone x16. In fact, a single GTX 580 barely saturates PCI-e 2.0 x4 bandwidth - and Premiere Pro CS5 does not support SLI at all (only one of the two GPUs is used). Hence, one can use a hardware RAID card in an LGA 1155 platform without losing any performance because of this fact about current GPUs as long as the secondary PCI-e slot runs in x8 mode.

Spooony has a point when he stated that my particular disk subsystem was over-trafficked: The Intel ICH or PCH simply could not handle the burden of doing double-duty as a RAID controller regardless of the platform. I tested RAID 5 on the 2600K / P67 system to test this, and discovered that the AVI time skyrocketed from about 65 seconds in RAID 0 or RAID 10 to about 480 seconds in RAID 5. This tells you that the ICH's/PCH's integrated SATA RAID controller sucked as a true RAID controller. In addition, with RAID enabled, there is too much back-and-forth switching going on within the ICH/PCH, which significantly affects performance in the H.264 (and to a lesser extent the MPEG-2) portions of the PPBM5 benchmarks.

In addition, those two top-scoring i7-920/950 systems have their OS drive running in IDE mode: They are not even running in AHCI mode, let alone RAID mode, in their ICH10R controllers. There is a reason: AHCI, like RAID, requires the use of software drivers for the integrated Intel RAID, which made the controller improperly utilize the CPU even without RAID. IDE mode, on the other hand, does not require software drivers other than the default Microsoft ATAPI driver.

Which goes back to this: The system, regardless of the CPU platform, really requires a $900+ RAID card in addition to multiple (12 or more) drives just to achieve such top performance results. A cheaper $500 card is just not good enough because it does not have enough cache. As a result, the performance improvement with the cheaper RAID cards does not justify their cost; in fact, several results from systems with the cheaper RAID cards show little to no improvement over the integrated Intel ICH/PCH RAID.
 
Last edited:
ard advice, it would seem that the 970 hex might be quicker but the 300 dollar difference does not justify the expense.

For almost everything out there, not likely. The problem is there is just precious little that uses 6 cores. Now if you do 3D rendering or maybe heavy multi-track audio work, then ok you could see a benefit. However for pretty much anything else, game in particular, nope.

You have to remember that clock-for-clock the 2600 is faster. So even if it was clocked the same speed as a 970, it would still be faster at executing things. Of course it is clocked a bit faster, so that helps too. The net effect of that is anything that uses 4 cores or less, which is almost everything at this point, will run faster on the SB. Only if something can use 5 or 6 cores would the 970 even have the possibility of being faster.

None of this factors in AVX either. How much that'll help and how widely it'll get used is very much up in the air right now, but Valve is excited and it has the potential to widen the performance gap.

So basically for anything you are likely to do, the 2600 is faster. The 970-990 are only useful for people who do very high end pro work: Lots of virtualization, 3D rendering, complex audio projects (really complex, audio doesn't hit the CPU all that hard and I had no problem on my Q9550) and so on.

Even if they were the same price, I'd still take the 2600 because it'll give better performance in the apps I actually run.
 
For almost everything out there, not likely. The problem is there is just precious little that uses 6 cores. Now if you do 3D rendering or maybe heavy multi-track audio work, then ok you could see a benefit. However for pretty much anything else, game in particular, nope.

You have to remember that clock-for-clock the 2600 is faster. So even if it was clocked the same speed as a 970, it would still be faster at executing things. Of course it is clocked a bit faster, so that helps too. The net effect of that is anything that uses 4 cores or less, which is almost everything at this point, will run faster on the SB. Only if something can use 5 or 6 cores would the 970 even have the possibility of being faster.

None of this factors in AVX either. How much that'll help and how widely it'll get used is very much up in the air right now, but Valve is excited and it has the potential to widen the performance gap.

So basically for anything you are likely to do, the 2600 is faster. The 970-990 are only useful for people who do very high end pro work: Lots of virtualization, 3D rendering, complex audio projects (really complex, audio doesn't hit the CPU all that hard and I had no problem on my Q9550) and so on.

Even if they were the same price, I'd still take the 2600 because it'll give better performance in the apps I actually run.

Very likely to have a few things running at once, however I do well understand your point.
Processor efficiency and possible higher overclocks is what the SB chip brings to the table. The six core would help in transcoding (in cases where the GPU is of less assistance), Photoshop, video work, occasional folding and the general mess for several programs actually running at once.... e-peen not withstanding :) Regardless, I am pretty much sold on the SB.
The discussion here is helpful, though it should be comparable, no consideration for subsystems. Assume that they will be the same for both systems.
 
Very likely to have a few things running at once, however I do well understand your point.
Processor efficiency and possible higher overclocks is what the SB chip brings to the table. The six core would help in transcoding (in cases where the GPU is of less assistance), Photoshop, video work, occasional folding and the general mess for several programs actually running at once.... e-peen not withstanding :) Regardless, I am pretty much sold on the SB.
The discussion here is helpful, though it should be comparable, no consideration for subsystems. Assume that they will be the same for both systems.

Well just having many things running doesn't call for many cores. Most things use precious little power. Right now I have 68 processes totaling 978 threads on my system. However CPU utilization is so low the C1E state is engaged and it is clocked down to 1.6GHz. Steam, IM, Firefox, ESET, Outlook, etc they just don't use much CPU.

In terms of Photoshop, not sure you'd see a real improvement. These days processors tend to be faster than PS needs. Been a long time since I saw anyone waiting on the processor for PS work, if they wait on anything it is the disk.

Video work it depends on what you are doing. The extra cores could help, but only if you are doing stuff that'd use them. For one, you'd need a video editor that made use of them. While all NLEs I've tried are multi-threaded, the specific subsystems are only threaded to a certain extent. Like Vegas only seems to use 2 cores for Windows Media encoding (uses 4 for AVC). Also if you are doing anything that needs that kind of power. An old Core i7-860, what I have at work, is enough to decode 24mbps 1080p AVCHD in realtime and cut it and all that. Only if you start doing some effects does playback drop frames.

Transcoding it would probably help, but then given time AVX may help more. That is one of the areas it is supposed to shine in. Plus you have to ask if you'd really care. So I use handbrake on pretty insane settings: CABAC, multi-hexagon search, QPRD in all frames, and so on. Ran at like 8-10fps on my Q9550. On my 2600k (stock speeds) it goes at above realtime speeds. A 970 might be faster, I'm not sure, but would I really care? That is already blazing and that is rather insane settings. Knock it down to a more reasonable amount of analysis and a smaller size for mobile devices and it flies.

Not saying the 970 is worthless, just that I see very few compelling uses for 6 cores, and not many on the horizon. I figure by the time it is needed, well I'll be wanting a new CPU anyhow.

As you noted, on top of all this there's the $300 price difference too. That really seals the deal. I probably wouldn't do it even for a dead equal price but for $300 more? No way.
 
By multiple programs running, I do mean active. Burning a disk while transcoding and viewing, however rare. The arguments for the 2600K are compelling enough for it to be my next purchase.

thanks again
 
Back
Top