TPU's Core i7 Overclocking and Feedback

[Chicken Patty]

Thanks Binge, so that just shows that more memoery speed and looser timings is not necessarily better. Better to run lower speed but tighter timings, am I correct?

 

[Binge]

thats what i been demonstrating for a few months now...

expensive parts are just that.... EXPENSIVE. no added bonus for the extra money.

i bet i could easily match anyones system with my low budget parts.

Thanks Binge, so that just shows that more memoery speed and looser timings is not necessarily better. Better to run lower speed but tighter timings, am I correct?

Actually CP it's what Fit is hinting at. The cas 8 1600Mhz memory performs within less than 1% of the high dollar 2000Mhz and even at slower speeds with tighter timings you won't see any significant performance increase save a "feel" that it's a little more responsive.

 

[Chicken Patty]

Actually CP it's what Fit is hinting at. The cas 8 1600Mhz memory performs within less than 1% of the high dollar 2000Mhz and even at slower speeds with tighter timings you won't see any significant performance increase save a "feel" that it's a little more responsive.

then this fucker had a point the whole way here

Yeah I always thought so too, thanks to FIT, i know have my eyes wide open on RAM at least.

 

[Binge]

then this fucker had a point the whole way here

Yeah I always thought so too, thanks to FIT, i know have my eyes wide open on RAM at least.

Well since we've been talking a lot of smack on motherboards too it seems that the real key to the i7 is... well the individual i7 processor. How's that for obvious?

I'm interested to hear about more overclocking results, but I hear a number of people claim difficulty overclocking the i7. I also believe that once someone understands the basics of the bios and relation between the voltages to parts of the CPU then overclocking becomes even simpler than any previous chipset. The only real exception would be getting a dead or failing component, but the overclocker can hardly be blamed for that happening.

 

[Thrackan]

Well since we've been talking a lot of smack on motherboards too it seems that the real key to the i7 is... well the individual i7 processor. How's that for obvious?

...

Which makes perfect sense since the chip has the memory controller built in.

Basically i7 has 2 major points that make it differ from earlier chips (at least up until today):
- 1 chipset (X58) which is the same for every board.
This means that, to a certain extent of course, every motherboard should be the same in performance. Obvious reasons like bad heatsinks or other components could be the exception.
Basically motherboard choice is up to taste and features.

- the on-chip memory controller which negates the effects of "bad" north- or southbridges on memory speed.
Because the mem controller is Intel spec and the chipset is too, the only thing that might be any difference is really bad RAM sticks.

My conclusion here is that Intel finally decided to make a complete backbone in the shape of X58+i7 in order to make sure they sell you a performance set of hardware.

 

[Fitseries3]

now they just need to fix the BCLK problem so we can see 250+ bclk.

 

[kid41212003]

Any idea why we can't adjust the QPI link speed below x36? I think that's the major problem why getting high BCLK so hard...

 

[Fitseries3]

they forgot to put a lower multi.

 

[mudkip]

Any idea why we can't adjust the QPI link speed below x36?

to make us buy cpu's with unlocked multipliers

 

[Binge]

Any idea why we can't adjust the QPI link speed below x36? I think that's the major problem why getting high BCLK so hard...

That's not a problem with the D0 as I proved earlier with a 4.2GHz overclock with x10 memory. My qpi is working at 4GHz, so close to the maximum for the 920.

to make us buy cpu's with unlocked multipliers

You can unlock the core multi but the qpi multi doesn't go any below x36 even for the extreme chips.

 

[Fitseries3]

if they added a 34x and 32x multi to the bios i bet we would see 240-260bclk

 

[PaulieG]

Soooo.... you're all going to laugh Super tight timings on ram? Pshhhh.... uber high frequency? Meh.... So I chose to test a CPU benchmark and a gaming benchmark to see if there was much of a difference between ram timings and frequency and if it really made as much or more difference than the frequency of the i7 itself. Here's the results!

i7 920 @ 4.2GHz Kingston HyperX 1600MHz 6-7-6-15
http://img.techpowerup.org/090517/ramtimingsvscoreclock1.png

Processor: Intel(R) Core(TM) i7 CPU 920 @ 2.67GHz @ 4200 Mhz
CPU ID: Intel64 Family 6 Model 26 Stepping 5
Operating System: Microsoft® Windows Vista™ Ultimate 64-bit Service Pack 1
Physical memory: 6.00 GB
Display adapter: NVIDIA GeForce GTX 295 0 MB
Video Processor: NVxx
Driver version: 8.15.11.8585 (20090501020200.000000-000)
Motherboard: DFI DFI LP UT X58 1.0

==============================================================
TimeDemo Play Started , (Total Frames: 1882, Recorded Time: 56.90s)
!TimeDemo Run 0 Finished.
Play Time: 43.65s, Average FPS: 43.12
Min FPS: 33.31 at frame 1638, Max FPS: 52.76 at frame 1560
Average Tri/Sec: 24930622, Tri/Frame: 578174
Recorded/Played Tris ratio: -0.32
!TimeDemo Run 1 Finished.
Play Time: 41.36s, Average FPS: 45.50
Min FPS: 33.31 at frame 1638, Max FPS: 53.60 at frame 1525
Average Tri/Sec: 26294232, Tri/Frame: 577846
Recorded/Played Tris ratio: -0.32
!TimeDemo Run 2 Finished.
Play Time: 41.36s, Average FPS: 45.50
Min FPS: 33.31 at frame 1638, Max FPS: 54.86 at frame 1570
Average Tri/Sec: 26403556, Tri/Frame: 580313
Recorded/Played Tris ratio: -0.32
TimeDemo Play Ended, (3 Runs Performed)
==============================================================

//////////// Summary \\\\\\\\\\\\\

05/17/2009 07:29:24 - Microsoft® Windows Vista™ Ultimate 64-bit Service Pack 1

DirectX 10 ENTHUSIAST 3X @ Map: frost @ 0 1920 x 1080 AA 0x
==> Framerate [ Min: 33.31 Max: 54.23 Avg: 45.50 ]


i7 920 @ 4.2GHz Kingston HyperX 1600MHz 8-8-8-24
http://img.techpowerup.org/090517/ramtimingsvscoreclock2.png

Processor: Intel(R) Core(TM) i7 CPU 920 @ 2.67GHz @ 4200 Mhz
CPU ID: Intel64 Family 6 Model 26 Stepping 5
Operating System: Microsoft® Windows Vista™ Ultimate 64-bit Service Pack 1
Physical memory: 6.00 GB
Display adapter: NVIDIA GeForce GTX 295 0 MB
Video Processor: NVxx
Driver version: 8.15.11.8585 (20090501020200.000000-000)
Motherboard: DFI DFI LP UT X58 1.0

==============================================================
TimeDemo Play Started , (Total Frames: 1882, Recorded Time: 56.90s)
!TimeDemo Run 0 Finished.
Play Time: 43.41s, Average FPS: 43.35
Min FPS: 35.27 at frame 1367, Max FPS: 53.49 at frame 1565
Average Tri/Sec: 24598810, Tri/Frame: 567448
Recorded/Played Tris ratio: -0.33
!TimeDemo Run 1 Finished.
Play Time: 41.30s, Average FPS: 45.56
Min FPS: 35.16 at frame 1655, Max FPS: 53.59 at frame 1524
Average Tri/Sec: 26746064, Tri/Frame: 586997
Recorded/Played Tris ratio: -0.32
!TimeDemo Run 2 Finished.
Play Time: 41.35s, Average FPS: 45.51
Min FPS: 35.16 at frame 1655, Max FPS: 53.93 at frame 1571
Average Tri/Sec: 26130090, Tri/Frame: 574143
Recorded/Played Tris ratio: -0.33
TimeDemo Play Ended, (3 Runs Performed)
==============================================================

//////////// Summary \\\\\\\\\\\\\

05/17/2009 07:07:11 - Microsoft® Windows Vista™ Ultimate 64-bit Service Pack 1

DirectX 10 ENTHUSIAST 3X @ Map: frost @ 0 1920 x 1080 AA 0x
==> Framerate [ Min: 35.16 Max: 53.76 Avg: 45.54 ]

i7 920 @ 4.2GHz Kingston HyperX 2000MHz 8-8-8-24
http://img.techpowerup.org/090517/ramtimingsvscoreclock4.png

Processor: Intel(R) Core(TM) i7 CPU 920 @ 2.67GHz @ 4200 Mhz
CPU ID: Intel64 Family 6 Model 26 Stepping 5
Operating System: Microsoft® Windows Vista™ Ultimate 64-bit Service Pack 1
Physical memory: 6.00 GB
Display adapter: NVIDIA GeForce GTX 295 0 MB
Video Processor: NVxx
Driver version: 8.15.11.8585 (20090501020200.000000-000)
Motherboard: DFI DFI LP UT X58 1.0

==============================================================
TimeDemo Play Started , (Total Frames: 1882, Recorded Time: 56.90s)
!TimeDemo Run 0 Finished.
Play Time: 42.89s, Average FPS: 43.88
Min FPS: 33.91 at frame 1635, Max FPS: 53.80 at frame 1553
Average Tri/Sec: 25446648, Tri/Frame: 579873
Recorded/Played Tris ratio: -0.32
!TimeDemo Run 1 Finished.
Play Time: 41.20s, Average FPS: 45.68
Min FPS: 33.91 at frame 1635, Max FPS: 53.80 at frame 1553
Average Tri/Sec: 26388626, Tri/Frame: 577728
Recorded/Played Tris ratio: -0.32
!TimeDemo Run 2 Finished.
Play Time: 41.35s, Average FPS: 45.51
Min FPS: 33.91 at frame 1635, Max FPS: 54.97 at frame 1563
Average Tri/Sec: 26367188, Tri/Frame: 579353
Recorded/Played Tris ratio: -0.32
TimeDemo Play Ended, (3 Runs Performed)
==============================================================

//////////// Summary \\\\\\\\\\\\\

05/17/2009 07:49:14 - Microsoft® Windows Vista™ Ultimate 64-bit Service Pack 1

DirectX 10 ENTHUSIAST 3X @ Map: frost @ 0 1920 x 1080 AA 0x
==> Framerate [ Min: 33.91 Max: 54.39 Avg: 45.60 ]

i7 920 @ 4.0GHz Kingston HyperX 2000MHz 8-8-8-24
http://img.techpowerup.org/090517/ramtimingsvscoreclock3.png

Processor: Intel(R) Core(TM) i7 CPU 920 @ 2.67GHz @ 4000 Mhz
CPU ID: Intel64 Family 6 Model 26 Stepping 5
Operating System: Microsoft® Windows Vista™ Ultimate 64-bit Service Pack 1
Physical memory: 6.00 GB
Display adapter: NVIDIA GeForce GTX 295 0 MB
Video Processor: NVxx
Driver version: 8.15.11.8585 (20090501020200.000000-000)
Motherboard: DFI DFI LP UT X58 1.0

==============================================================
TimeDemo Play Started , (Total Frames: 1882, Recorded Time: 56.90s)
!TimeDemo Run 0 Finished.
Play Time: 42.77s, Average FPS: 44.00
Min FPS: 33.55 at frame 1631, Max FPS: 54.44 at frame 1548
Average Tri/Sec: 25262982, Tri/Frame: 574157
Recorded/Played Tris ratio: -0.33
!TimeDemo Run 1 Finished.
Play Time: 41.61s, Average FPS: 45.23
Min FPS: 33.55 at frame 1631, Max FPS: 54.69 at frame 1568
Average Tri/Sec: 26344398, Tri/Frame: 582456
Recorded/Played Tris ratio: -0.32
!TimeDemo Run 2 Finished.
Play Time: 41.51s, Average FPS: 45.34
Min FPS: 33.55 at frame 1631, Max FPS: 54.69 at frame 1568
Average Tri/Sec: 26135890, Tri/Frame: 576426
Recorded/Played Tris ratio: -0.32
TimeDemo Play Ended, (3 Runs Performed)
==============================================================

//////////// Summary \\\\\\\\\\\\\

05/17/2009 07:40:19 - Microsoft® Windows Vista™ Ultimate 64-bit Service Pack 1

DirectX 10 ENTHUSIAST 3X @ Map: frost @ 0 1920 x 1080 AA 0x
==> Framerate [ Min: 33.55 Max: 54.69 Avg: 45.29 ]


Video card was a GTX295 at stock settings. The only changes to the bios were to change the timing and frequency of the ram and CPU on the last test. The last test reveals that most benchmarks will be affected by CPU frequency instead of ram frequency. I am more surprised that this D0 can run at 4.2GHz with 2000Mhz ram That alone is a bright shiny beacon of niceness that came out of this abyssmal line of testing. Why am I less enthusiastic about my 50nm Elpida HyperX ram modules? Well I think the results speak for themselves Happy overclocking!

I did a similar test a couple of weeks back, and I wish I had posted it. It proves that there is no need for expensive DDR3 for this platform. I do think that motherboard choice is important. I have noticed some disparity in various boards BCLK when using the same chip. This does not mean that more expensive is better. Hell, my Asus gene was $240, and schooled my $280 Blood Rage in stable BCLK.

 

[n-ster]

I would like to know... how much difference can a X58 board make? say I run an i7 (c0/c1) at 3.8ghz or 4ghz... from board to board, I'm guessing you would need different vcores... would this make much of a difference?

So say the ECS is board will do fine?

Also, for i7s life span, what temps/vcore max would you need so the i7 lasts you 5 yrs? Say using a S1283V C0/C1 stepping

Is it safe to have a 4ghz i7 at 80C full load? willl it last at least 5 yrs?

 

[mudkip]

I would like to know... how much difference can a X58 board make? say I run an i7 (c0/c1) at 3.8ghz or 4ghz... from board to board, I'm guessing you would need different vcores... would this make much of a difference?

So say the ECS is board will do fine?

Also, for i7s life span, what temps/vcore max would you need so the i7 lasts you 5 yrs? Say using a S1283V C0/C1 stepping

Is it safe to have a 4ghz i7 at 80C full load? willl it last at least 5 yrs?

My i7 gets 78-80 degrees on full load.

will it last 5 years? i don't care because by that time I'll have another CPU anyway

 

[Tatty_Two]

hey tatty have you noticed 3gb on vista is not snappy unless you close some programs

I find it OK but Win 7 feels a little better, do you have 32 or 64 bit?

As an aside, I have ordered the UD4P, it will be here on Tuesday, I am ready for a whole new overclocking lesson

 

[aquax]

need help please, I don't know which board to choose for my i7 build

1- Gigabyte EX58-UD4P
2- ASUS P6T DELUXE V1

 

[mudkip]

need help please, I don't know which board to choose for my i7 build

1- Gigabyte EX58-UD4P
2- ASUS P6T DELUXE V1

Do some research, but my opinion : UD4P

 

[Thrackan]

need help please, I don't know which board to choose for my i7 build

1- Gigabyte EX58-UD4P
2- ASUS P6T DELUXE V1

If both boards have the features you want, go with the cheaper one.

But, as mentioned, it is important you do your own research. For i7 boards, ask yourself the following:
- Does the board have all the features I want? PCI-E slots? SATA RAID? Fan connectors? Etc etc...
- Does the board look ugly enough not to buy it?

If both boards are still in your list and have about the same price tag, double check the layout of the boards. Little things like IDE connector placement can be annoying sometimes.

And if you still have more than 1 board on the list, either flip a coin or go with the cheapest one. Or the colour that matches your case. Or the one that's best available at your favourite store. Maybe even the one that delivers the best extras in the box...

 

[PaulieG]

From what I've heard from a trusted member, the UD4 has a high max BCLK, and since it's built similar to the UD5 (which I have), it's a very well built tank.

 

[aquax]

thanks guys, I'm going with the UD4P

 

[Fitseries3]

see...... some can do it.... http://www.xtremesystems.org/forums/showthread.php?t=224581

i want ALL to do it.

 

[kid41212003]

That's not a problem with the D0 as I proved earlier with a 4.2GHz overclock with x10 memory. My qpi is working at 4GHz, so close to the maximum for the 920.

You can unlock the core multi but the qpi multi doesn't go any below x36 even for the extreme chips.

If i'm not wrong,
QPI is the point-to-point technology that connects your cpu to your memory, northbridge, and southbridge. Uncore is the part of the cpu that has the L3 cache and memory controller.

And, Uncore is adjustable freely, but not QPI (not below x36). At 200 BCLK and x36 QPI multi, we have QPI running at 7.2GT/s or 3600MHz, a really high speed compare to default clock.

I can't find any other reason why the Extreme chip can clock alot higher except that its QPI link speed always alot lower.

 

[Tatty_Two]

Which makes perfect sense since the chip has the memory controller built in.

Basically i7 has 2 major points that make it differ from earlier chips (at least up until today):
- 1 chipset (X58) which is the same for every board.
This means that, to a certain extent of course, every motherboard should be the same in performance. Obvious reasons like bad heatsinks or other components could be the exception.
Basically motherboard choice is up to taste and features.

- the on-chip memory controller which negates the effects of "bad" north- or southbridges on memory speed.
Because the mem controller is Intel spec and the chipset is too, the only thing that might be any difference is really bad RAM sticks.

My conclusion here is that Intel finally decided to make a complete backbone in the shape of X58+i7 in order to make sure they sell you a performance set of hardware.

I agree completely....... however, once you get past the cooling solutions for a particular motherboard, throw out the "accessories" that come with the various models which pretty much determine the price point, such as Esata, Wifi, dual or even Tri Lan, what it really comes down to is BIOS and BIOS coding which = Stability which of course equals overclocking success.

 

[Fitseries3]

i have something for you guys.......

 

[Tatty_Two]

From what I've heard from a trusted member, the UD4 has a high max BCLK, and since it's built similar to the UD5 (which I have), it's a very well built tank.

Is it not actually identical apart from the fact it has less accesories such as Esata etc? The BIOSes I know are identical as too is the cooling I think.