Review on Processor Intel Core i7-3820 LGA2011, 4 x 3600 MHz, BOX by Michal Adach ᠌

Name: Processor Intel Core i7-3820 LGA2011, 4 x 3600 MHz, BOX
Item: Processor Intel Core i7-3820 LGA2011, 4 x 3600 MHz, BOX
Rating: 4

May 29, 2023

Everything is fine, I liked the product, I bought it for a reason.

The problems of heat dissipation were successfully solved by spending about the same amount on cooling as on the chip itself.
I don’t consider the crazy heat dissipation a disadvantage, although it certainly annoys.
When overclocked to 5GHz, the declared 130 watts turn into 200.
Even with a good custom-loop CBO, you have to keep an eye on the temperature of the cores (especially if you get a badly smeared chip). In the air, it is hardly worth driving him at all.
Frankly, I personally dealt with 2 copies of this chip, + I am familiar with some of its other owners: I have not yet met a chip that would differ so much in overclocking properties depending on a particular copy.

In 2 words about the LGA2022 and X79 chipset with which this chip is inextricably linked:
The chipset is problematic in places, especially the raid controller. From time to time, disks drop out of the raid, it's better not to think about TRIM at all. Also, popular applications for backups and recovery (like Acronis) still do not want to be friends with the raid on X79. Due to 4-channel memory, high TDP processors and the need to support a large number of PCI-E slots at the same time, motherboards are equipped with large VRM blocks that are not always able to cool on their own.

After I squeezed everything I could out of it in 3DMark, I sold it and bought a 3820. I'm waiting for Ivy-E.

Pros

6 cores 12MB cache and 40 PCI-E lanes. Unfortunately, there is nothing more to add to the merits except for its main characteristics.

Cons

4 channel memory controller. It would seem that this should be one of the main advantages, but alas: this controller is actually two 2-channel controllers from the usual Sandy Bridge. Due to the incredibly high ping for such a link, its real throughput is lower than that of SB. (Which is quite sad - even using the 2-channel mode does not save the situation) It also affects the maximum memory frequency. The probability of getting a chip that will pull 2666MHz is almost the same as winning the lottery. Even 2400 is far from being pulled by everyone, although mine copes. lack of official support for PCI-E 3.0. In essence, the problem is not big, however, especially unsuccessful copies of the 3.0 chip are not really supported, which is very unfortunate, because otherwise what is the use of 40 lanes. In general, a weak Uncore block, the percentage is very sensitive to the voltage on the controllers (VTT and VCCSA), if something goes wrong, it can degrade very quickly.