Overclock.net › Forums › Industry News › Hardware News › [Xbit]: Intel Core i7-3820 vs. Core i7-2700K and Core i7-3930K
New Posts  All Forums:Forum Nav:

[Xbit]: Intel Core i7-3820 vs. Core i7-2700K and Core i7-3930K - Page 7

post #61 of 86
Quote:
Originally Posted by tpi2007 View Post


You know pefectly what you said and what I meant. You said "Quad channel memory required", when it is not required. And don't pick the one example where it will take a performance hit. I was saying you can use single, dual and triple channel if you want, meaning you can use it in a dual channel configuration just like you can use Sandy Bridge right now. Use 2x 4GB sticks right now for example, upgrade with an additional 2 x 4 GB sticks later for a total of 16 GB.
The core i7 3820 does not have a GPU and the added PCIe lanes and additional memory channels don't add up to 130w. It's not the same but it's not the difference between a 95w and a 130w TDP CPU as you were implying.
You can read it on Anand's review:

3.2w at load difference between a 2600k and a 3820. That is hardly consistent with a 130w TDP and you know it. They could have labeled this as 95w TDP CPU if they wanted, but the fact is this is for the enthusiast platform, so they label it the same as the six core CPUs.
Quote:
and no Quicksync or iGPU @ 130W
What I said about QuickSync and the integrated GPU was because you phrased it like this. Or did you mean that the CPU could have an integrated GPU without Quicksync like the Celerons ? That would make even less sense.
The performance hit I'm talking about has to do with using QuickSync on a gaming PC. If you have a discrete card and use QuickSync with the Lucid Logix arrangement you have a performance penalty in gaming with the discrete card(s) because of how the system works, routing the graphics calls from the integrated GPU to the discrete GPU(s).
Adding a dummy plug like you said only further tells that Intel's implementation of QuickSync, which is a very useful feature, is less than optimal like I said.
Not all 2600k's can reach 5 Ghz, much less on air. The 2700k's yes and with lower voltages than the 2600k's that do, but not all 2600k's and especially not the ones that are being sold after the 2700k.
What are you saying about 125 Mhz being the limit for keeping the PCIe and DMI within spec ? This is not Sandy Bridge and that would actually be around 103-105 Mhz in that case.
Quote:
With Sandy Bridge E, overclocking changes a bit. The clock generator is still mostly impervious to significant bus clock changes, however you're now able to send a multiple of its frequency to the CPU if you so desire. The options available are 100MHz, 125MHz, 166MHz and 250MHz.
- Anand's Core i7 3960 review.
You can fine tune the frequency you want on the i7 3820 by changing the multiplier in addition to changing the BCLK multiplier. 125 Mhz is the one used by the reviewers because it's the one most people will realistically use with a higher CPU multiplier. Yes, you can use the 250 Mhz option, just use a lower multiplier for the CPU. It can go as low as 12.
Besides, Anand got it to work at 4.875 Ghz, just not stably. He didn't try any form of exotic cooling nor any high end non Intel motherboard. Besides, the overclocking potential of his sample is not indicative of 100% of the chips, especially since it may have been a C1 stepping, and C2 steppings may overclock a little better.

I do know exactly what I said. You can choose to use single channel if you want and take a performance hit along with it. Why is this even being debated? It supports up to quad channel and you have the choice to use whatever you want. I wasn't saying that you could ONLY use quad channel memory.

I wasn't aware that is was similar to the 2600K's TDP. The thread's source made it out to be like it was operating around 130W. That's obviously a plus.

What is so confusing about mentioning the lack of an iGPU and Quicksync? I never said anything about it behaving like a Celeron... If you don't want to use a dummy plug, you can just plug it into the other port on your monitor and never use it. You just need something physically in place for it to work. Quicksync is awesome. There is no performance hit like you are trying to make out since you don't use Lucid while gaming...

125Mhz is the maximum BCLK that can be used to keep the PCIe and DMI within their specifications. If you use 250Mhz and lower the multiplier, you're not going to have a stable system and will be limited in your OC frequency choices(I've yet to see a system boot at 250Mhz anyways). For example, even on the 125Mhz BCLK, you can't hit 4.5Ghz. The closest you can get to it is 4.625.

The increased price of a X79 mobo compared to Z68 makes it even worse. Why spend all that money on the mobo and only get a quad core with limited OCing? You get more for less money with Z68 boards. The only benefit would be in memory intensive applications that need quad-channel bandwidth.

Anyways, Ivy-Bridge will be out in 2 months and make the situation even worse.
Edited by PoopaScoopa - 1/29/12 at 6:10pm
post #62 of 86
Quote:
Originally Posted by PoopaScoopa View Post

I do know exactly what I said. You can choose to use single channel if you want and take a performance hit along with it. Why is this even being debated? It supports up to quad channel and you have the choice to use whatever you want. I wasn't saying that you could ONLY use quad channel memory.
I wasn't aware that is was similar to the 2600K's TDP. The thread's source made it out to be like it was operating around 130W. That's obviously a plus.
What is so confusing about mentioning the lack of an iGPU and Quicksync? I never said anything about it behaving like a Celeron... If you don't want to use a dummy plug, you can just plug it into the other port on your monitor and never use it. You just need something physically in place for it to work. Quicksync is awesome. There is no performance hit like you are trying to make out since you don't use Lucid while gaming...
125Mhz is the maximum BCLK that can be used to keep the PCIe and DMI within their specifications. If you use 250Mhz and lower the multiplier, you're not going to have a stable system and will be limited in your OC frequency choices. For example, even on the 125Mhz BCLK, you can't hit 4.5Ghz. The closest you can get to it is 4.625.
The increased price of a X79 mobo compared to Z68 makes it even worse. Why spend all that money on the mobo and only get a quad core with limited OCing? You get more for less money with Z68 boards. The only benefit would be in memory intensive applications that need quad-channel bandwidth.


Well, if you know exactly what you said, they you should know that when you said "required" you shoudn't have said required. More, why do you keep insisting on talking about single channel ? You can use dual channel, just like you use with Sandy Bridge, so there is no performance penalty.
Quote:
Originally Posted by PoopaScoopa View Post

Intel does list it as 130W though.
Quad channel memory required, limited multiplier and no Quicksync/iGPU? No thank you.
edit, it doesn't actually run near 130W. and quad-channel memory is the only benefit here.


There is nothing confusing about QuickSync or integrated GPU, I was just reading your sentence. When you say "or", I assume that when you talked about the integrated GPU after having talked about QuickSync, you meant an integrated GPU that doesn't have QuickSync, by exclusion. When you mentioned QuickSync in the first place, everybody has to assume the existence of an integrated GPU, but when you continue the sentence saying "or" an integrated GPU, I assume that you mean something different, otherwise the last part of the sentence is redundant. Get it now ?

The thing is, you just edited your original post and removed the "or", more than three hours after having originally written it, so you know perfectly well what I'm talking about.

Also, you can get QuickSync-like functionality with the new Radeon HD 7000 series, and I bet Kepler will have similar functionality too, so not having QuickSync is not going to be a problem for someone building a new system now.

Why can't you hit 4.5 Ghz with a 125 Mhz BCLK multiplier ? 36 x 125 = 4.5 Ghz.
Edited by tpi2007 - 1/29/12 at 6:26pm
 
Metro 2033 review
Metro 2033
CPUMotherboardGraphicsRAM
Core i7-3820 Asus Sabertooth X79 MSI GTX 750 Ti TF Gaming 16 GB Corsair DDR3 1866 Mhz Dominator 
Hard DriveOptical DriveCoolingOS
Samsung SSD 830 128GB + WD Caviar Black 1TB Sony Optiarc DVD-RW Corsair A70 + Noiseblocker M12-P Windows 7 Home Premium 64-bit 
MonitorKeyboardPowerCase
BenQ RL2455HM Cooler Master Octane Corsair AX750 Professional Modular 80 Plus Gold Cooler Master HAF 912 Plus 
Mouse
Cooler Master Octane 
  hide details  
Reply
 
Metro 2033 review
Metro 2033
CPUMotherboardGraphicsRAM
Core i7-3820 Asus Sabertooth X79 MSI GTX 750 Ti TF Gaming 16 GB Corsair DDR3 1866 Mhz Dominator 
Hard DriveOptical DriveCoolingOS
Samsung SSD 830 128GB + WD Caviar Black 1TB Sony Optiarc DVD-RW Corsair A70 + Noiseblocker M12-P Windows 7 Home Premium 64-bit 
MonitorKeyboardPowerCase
BenQ RL2455HM Cooler Master Octane Corsair AX750 Professional Modular 80 Plus Gold Cooler Master HAF 912 Plus 
Mouse
Cooler Master Octane 
  hide details  
Reply
post #63 of 86
Quote:
Originally Posted by PoopaScoopa View Post

...
125Mhz is the maximum BCLK that can be used to keep the PCIe and DMI within their specifications. If you use 250Mhz and lower the multiplier, you're not going to have a stable system and will be limited in your OC frequency choices(I've yet to see a system boot at 250Mhz anyways). For example, even on the 125Mhz BCLK, you can't hit 4.5Ghz. The closest you can get to it is 4.625.
...

I don't believe this is true. At least in theory the 1.66X and 2.5X strap ratios should keep PCI-E and DMI within spec at 100 MHz, according to Intel's Sandy Bridge-E press deck. Judging by the lackluster results at that strap ratios, and the non-functionality of the 2.5X ratio, something else is preventing them from working properly.
Quote:
Originally Posted by tpi2007 View Post

...
Why can't you hit 4.5 Ghz with a 125 Mhz BCLK multiplier ? 36 x 125 = 4.5 Ghz.

It's not known exactly why, but there seems to be a bug with the 36X multiplier and the 1.25X strap ratio, some reviewers have found that the system wont boot at that setting, but will boot using the 37X multiplier for 4.625 GHz. It's probably something like with the i7 920, where the 21X multiplier was better for overclocking that the 20X multiplier.
post #64 of 86
Quote:
Originally Posted by tpi2007 View Post

Well, if you know exactly what you said, they you should know that when you said "required" you shoudn't have said required. More, why do you keep insisting on talking about single channel ? You can use dual channel, just like you use with Sandy Bridge, so there is no performance penalty.
There is nothing confusing about QuickSync or integrated GPU, I was just reading your sentence. When you say "or", I assume that when you talked about the integrated GPU after having talked about QuickSync, you meant an integrated GPU that doesn't have QuickSync, by exclusion. When you mentioned QuickSync in the first place, everybody has to assume the existence of an integrated GPU, but when you continue the sentence saying "or" an integrated GPU, I assume that you mean something different, otherwise the last part of the sentence is redundant. Get it now ?
The thing is, you just edited your original post and removed the "or", more than three hours after having originally written it, so you know perfectly well what I'm talking about.
Also, you can get QuickSync-like functionality with the new Radeon HD 7000 series, and I bet Kepler will have similar functionality too, so not having QuickSync is not going to be a problem for someone building a new system now.
Why can't you hit 4.5 Ghz with a 125 Mhz BCLK multiplier ? 36 x 125 = 4.5 Ghz.

Yes, the only benefit is quad-channel memory. That's why I said "required". If you want to see the benefit of X79 on the 3820, you need to use quad channel memory. Everything else though, is worse than the Z68 platform. I removed the "or" because you go on and on talking about things like the Celeron having an iGPU but no Quicksync when I never said anything about that. You're not reading it properly and are trying to imply that I said you could have Quicksync without an iGPU, which I didn't.

The 3820 has no iGPU, no Quicksync, platform costs more and is limited in overclocking. You can't hit 4.5Ghz because it's unstable. Have you even tried it? Read the reviews.
Quote:
Originally Posted by gplnpsb View Post

I don't believe this is true. At least in theory the 1.66X and 2.5X strap ratios should keep PCI-E and DMI within spec at 100 MHz, according to Intel's Sandy Bridge-E press deck. Judging by the lackluster results at that strap ratios, and the non-functionality of the 2.5X ratio, something else is preventing them from working properly.
"In theory" but I've yet to see anyone boot with a 250Mhz BCLK and 166Mhz seems to have limited success. I don't think this is a BIOS problem but more of a chipset problem TBH.
Edited by PoopaScoopa - 1/29/12 at 6:42pm
post #65 of 86
Quote:
Originally Posted by PoopaScoopa View Post

Yes, the only benefit is quad-channel memory. That's why I said "required". If you want to see the benefit of X79 on the 3820, you need to use quad channel memory. Everything else though, is worse than the Z68 platform.

No, you said "required" as if it were a downside, as if you had to buy more memory sticks to populate four channels, which you don't. At least that is how a normal person would read your sentence.

And no, quad channel isn't the only benefit. With socket 2011 you get the upgrade path to Ivy Bridge-E. And that is a big benefit.
Edited by tpi2007 - 1/29/12 at 6:37pm
 
Metro 2033 review
Metro 2033
CPUMotherboardGraphicsRAM
Core i7-3820 Asus Sabertooth X79 MSI GTX 750 Ti TF Gaming 16 GB Corsair DDR3 1866 Mhz Dominator 
Hard DriveOptical DriveCoolingOS
Samsung SSD 830 128GB + WD Caviar Black 1TB Sony Optiarc DVD-RW Corsair A70 + Noiseblocker M12-P Windows 7 Home Premium 64-bit 
MonitorKeyboardPowerCase
BenQ RL2455HM Cooler Master Octane Corsair AX750 Professional Modular 80 Plus Gold Cooler Master HAF 912 Plus 
Mouse
Cooler Master Octane 
  hide details  
Reply
 
Metro 2033 review
Metro 2033
CPUMotherboardGraphicsRAM
Core i7-3820 Asus Sabertooth X79 MSI GTX 750 Ti TF Gaming 16 GB Corsair DDR3 1866 Mhz Dominator 
Hard DriveOptical DriveCoolingOS
Samsung SSD 830 128GB + WD Caviar Black 1TB Sony Optiarc DVD-RW Corsair A70 + Noiseblocker M12-P Windows 7 Home Premium 64-bit 
MonitorKeyboardPowerCase
BenQ RL2455HM Cooler Master Octane Corsair AX750 Professional Modular 80 Plus Gold Cooler Master HAF 912 Plus 
Mouse
Cooler Master Octane 
  hide details  
Reply
post #66 of 86
Quote:
Originally Posted by gplnpsb View Post

I don't believe this is true. At least in theory the 1.66X and 2.5X strap ratios should keep PCI-E and DMI within spec at 100 MHz, according to Intel's Sandy Bridge-E press deck. Judging by the lackluster results at that strap ratios, and the non-functionality of the 2.5X ratio, something else is preventing them from working properly.
It's not known exactly why, but there seems to be a bug with the 36X multiplier and the 1.25X strap ratio, some reviewers have found that the system wont boot at that setting, but will boot using the 37X multiplier for 4.625 GHz. It's probably something like with the i7 920, where the 21X multiplier was better for overclocking that the 20X multiplier.

Agreed with the first part of your post. Intrigued by the second. Let's see how it goes when we have more people testing more CPUs as the months go by. Maybe it's a bug in the Bios that will be corrected.
 
Metro 2033 review
Metro 2033
CPUMotherboardGraphicsRAM
Core i7-3820 Asus Sabertooth X79 MSI GTX 750 Ti TF Gaming 16 GB Corsair DDR3 1866 Mhz Dominator 
Hard DriveOptical DriveCoolingOS
Samsung SSD 830 128GB + WD Caviar Black 1TB Sony Optiarc DVD-RW Corsair A70 + Noiseblocker M12-P Windows 7 Home Premium 64-bit 
MonitorKeyboardPowerCase
BenQ RL2455HM Cooler Master Octane Corsair AX750 Professional Modular 80 Plus Gold Cooler Master HAF 912 Plus 
Mouse
Cooler Master Octane 
  hide details  
Reply
 
Metro 2033 review
Metro 2033
CPUMotherboardGraphicsRAM
Core i7-3820 Asus Sabertooth X79 MSI GTX 750 Ti TF Gaming 16 GB Corsair DDR3 1866 Mhz Dominator 
Hard DriveOptical DriveCoolingOS
Samsung SSD 830 128GB + WD Caviar Black 1TB Sony Optiarc DVD-RW Corsair A70 + Noiseblocker M12-P Windows 7 Home Premium 64-bit 
MonitorKeyboardPowerCase
BenQ RL2455HM Cooler Master Octane Corsair AX750 Professional Modular 80 Plus Gold Cooler Master HAF 912 Plus 
Mouse
Cooler Master Octane 
  hide details  
Reply
post #67 of 86
Quote:
Originally Posted by tpi2007 View Post

No, you said "required" as if it were a downside, as if you had to buy more memory sticks to populate four channels, which you don't. At least that is how a normal person would read your sentence.
And no, quad channel isn't the only benefit. With socket 2011 you get the upgrade path to Ivy Bridge-E. And that is a big benefit.

No, I agree with what Poopa said. If I went to the trouble to buy X79 I certainly wouldn't cheap out and buy less than quad-channel memory. What would be the point of that?
post #68 of 86
Quote:
Originally Posted by tpi2007 View Post

No, you said "required" as if it were a downside, as if you had to buy more memory sticks to populate four channels, which you don't. At least that is how a normal person would read your sentence.
And no, quad channel isn't the only benefit. With socket 2011 you get the upgrade path to Ivy Bridge-E. And that is a big benefit.

If you're buying X79 for IVY-Bridge-E, I think you're buying it for the wrong reasons. There will be better mobos out when that comes around. I don't know how else to get it across to you but the benefit of the 3820 is the increased memory bandwidth. If you're going to just use dual-channel, why even buy it?
post #69 of 86
Quote:
Originally Posted by tpi2007 View Post

Also, you can get QuickSync-like functionality with the new Radeon HD 7000 series, and I bet Kepler will have similar functionality too, so not having QuickSync is not going to be a problem for someone building a new system now.

QuickSync uses an IGP that uses ~15W of power. Also, it blows away AMD APP for hardware encoding, beats software encoding (faster than 980X, probably faster than 3930K), I mean sure, perhaps HD 7000 and Kepler will improve on their encoding speeds but it's nowhere close to Intel's IGP. This is almost certainly due to Intel using only 12 EUs (compared to 512 CUDA cores on a 580 or 2048 SPs on a 7970). Video encoding is predominantly serial (doesn't scale well with 100s of GPU cores), so I further doubt Kepler/HD 7000 bringing anything comparable in terms of encoding speed.

edit: To clarify, AMD and Nvidia already have graphics cards that can do hardware encoding. If they improved by 50%, well... Ivy Bridge's HD 2500 and HD 4000 are going to improve by 50% over HD 2000 and HD 3000 respectively. It's not a necessary feature, but it's nice using my laptop, ripping a DVD and using only ~30% CPU on an i3 dual-core, and still encoding faster than pretty much any other method (desktop or laptop).
Edited by jrbroad77 - 1/29/12 at 6:50pm
 
Lanbox Lite
(16 items)
 
 
CPUGraphicsRAMHard Drive
i3-2310M Intel HD 3000 8GB DDR3 Samsung 830 64GB SSD 
Hard DriveOSMonitorCase
640GB Hitachi HDD Windows 7 13.3" LCD Magnesium Alloy, 3.2lbs 
CPUMotherboardGraphicsRAM
Phenom II X4 955 BE Gigabyte GA-MA78GPM-UD2H MSI Hawk R5770 3x2GB G.Skill DDR2 800 4-4-4-12 
Hard DriveHard DriveOptical DriveCooling
1TB Samsung Spinpoint F3 250 GB WD Caviar Black Samsung 20X DVD-R/RW Thermaltake MaxOrb 
CoolingOSMonitorPower
Noctua NF-B9-1600 Windows 7 Pro 64-bit BenQ E2420HD, 24" 1920x1080 TT Purepower 500W 
Case
TT Lanbox Lite 
  hide details  
Reply
 
Lanbox Lite
(16 items)
 
 
CPUGraphicsRAMHard Drive
i3-2310M Intel HD 3000 8GB DDR3 Samsung 830 64GB SSD 
Hard DriveOSMonitorCase
640GB Hitachi HDD Windows 7 13.3" LCD Magnesium Alloy, 3.2lbs 
CPUMotherboardGraphicsRAM
Phenom II X4 955 BE Gigabyte GA-MA78GPM-UD2H MSI Hawk R5770 3x2GB G.Skill DDR2 800 4-4-4-12 
Hard DriveHard DriveOptical DriveCooling
1TB Samsung Spinpoint F3 250 GB WD Caviar Black Samsung 20X DVD-R/RW Thermaltake MaxOrb 
CoolingOSMonitorPower
Noctua NF-B9-1600 Windows 7 Pro 64-bit BenQ E2420HD, 24" 1920x1080 TT Purepower 500W 
Case
TT Lanbox Lite 
  hide details  
Reply
post #70 of 86
Quote:
Originally Posted by jrbroad77 View Post

QuickSync uses an IGP that uses ~15W of power. Also, it blows away AMD APP for hardware encoding, beats software encoding (faster than 980X, probably faster than 3930K), I mean sure, perhaps HD 7000 and Kepler will improve on their encoding speeds but it's nowhere close to Intel's IGP. This is almost certainly due to Intel using only 12 EUs (compared to 512 CUDA cores on a 580 or 2048 SPs on a 7970). Video encoding is predominantly serial (doesn't scale well with 100s of GPU cores), so I further doubt Kepler/HD 7000 bringing anything comparable in terms of encoding speed.

Even if performance may be increased(doubt it's as fast though), as shown in the Anandtech review of SB, there are serious quality issues with AMD's APP/Nvidia's CUDA. If you're just transcoding video for your phone though, you can probably get away with it as quality doesn't really matter there. Quicksync has near identical quality* to x86 encoding.
Edited by PoopaScoopa - 1/29/12 at 6:55pm
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Hardware News
Overclock.net › Forums › Industry News › Hardware News › [Xbit]: Intel Core i7-3820 vs. Core i7-2700K and Core i7-3930K