Motherboard Incoming AM4 Mini-ITX boards

[EdZ]

New Well intel is to blame on lga 115x bad tim/heat transfert. Even on broadwell-e intel is doing far better.

Again, the TIM is just fine, the problem is with clearance between the top of the die and bottom of the IHS. The large-core LGA2011 chips are soldered because the dies are still large enough to do so without damage.

Intel did clearly even worse than skylake on Kabylake

Smaller dies will run hotter for the same power as larger dies, simply because you have less surface area to extract heat from. this has been the case for years, ever since TDPs stopped increasing (hitting ~100W for normal consumer chips) but processes kept shrinking.
Skylake is pushing 90W out of 112mm^2 (die size of Kaby lake is not yet confirmed). RyZen has been slated at 95W for an unknown die size, and do not have an iGPU to 'pad out' die area. If the 4c8t die is cut down rather than just a binned large die, it will run into the exact same thermal issues as Intel have been dealing with.

 

[MarcParis]

I understand the issue of the distance between ihs and die, no issue on that. I'm just saying that Intel could handle in a better way this problem, because on kabylake issue is becoming really noticeable.
On Skylake delliding was bringing between around 3-10 degrees improvement....now on kabylake it's reaching between 10-25 degrees...

 

[MarcParis]

Well, here is a trusted and objective review about core i7 7700K delid. Results are not that bad for Intel :
http://www.tomshardware.fr/articles/overclocking-kabylake-z270-wizerty-delid,2-2617-2.html

Just switching thermal paste from Intel stock, Kryonaut (best class non conductive paste), Conductonaut (metallic) on core i7 7700K @1,3v, prime 95

Between Intel and Kyronaut, there is no real interest to delid CPU. However switching to metallic pasta is just amazing.

Clearly there is way of optimisation. I'm sure Intel could propose a core i7 7700K premium with metallic pasta/soldering...people will buy it anyway..

To be honest, cpu delid is not worth with a 5GHz/1,3V OC...silicone lotery could be crual..

 

[EdZ]

I'd be wary of putting any 'liquid metal' TIM on a bare die. Those TIMs are gallium-based, which will happily infiltrate Silicon and form a eutectic alloy at ~30°C. The metal coatings on top of the die will provide some protection, but any imperfections in the coating, and the edges of the die as the TIM migrates around the chip, will provide avenues for infiltration. As that infiltration progresses, it will start to change the properties of the silicon in bulk, and it is the fine doping of the silicon that allows the chip to function.

tl;dr like with soldering small dies, an 'official' chip using eutectic Gallium TIM is living on borrowed time.
You can delid/repaste yourself and gain performance, but like adding NOS to an engine that performance comes at the price of long-term reliability (and risk of catastrophic failure as with a total delid and physical die damage), so you're living on borrowed time

 

[MarcParis]

I agree with you that metallic compound is quite intrusive/agressive. Based on my personal experience, I won't use it again..
That's why i like this review, it's clarifying delid efficiency between non conductive compound and metallic compound.

 

[Zerofool]

Yet another unconfirmed Ryzen leak. According to this one, there will be a 8C/16T model with just 65W TDP. If true, that's a great accomplishment. Now imagine what's possible with just 4C/8T or 4C/4T...
I'm so anxious to see the reviews!

 

[zovc]

Heck yeah, these are looking quite promising for my "home server" machine.

 

[jeshikat]

According to this one, there will be a 8C/16T model with just 65W TDP.

If Zen is real and not just overly optimistic hype again (looking at you Bulldozer), then I think AMD is going to sell a lot of chips.

 

[Zerofool]

If Zen is real and not just overly optimistic hype again (looking at you Bulldozer), then I think AMD is going to sell a lot of chips.

Let's hope so. Potential Zen failure will be a disaster for everyone.
Historically, Jim Keller has delivered every time (arguably), so that's encouraging. Judging by all the info so far, I think Ryzen will be a commercial success, even if it doesn't beat intel's products in pure performance ("close enough" will work for many, at the right price).

 

[zovc]

Yeah... is AMD even able to survive Zen not delivering?

 

[alexep7]

It's going to be a success for the average consumer regardless, which is what matters in the end. Even if it doesn't (and I don't believe it will) match Intel's performances.

 

[MarcParis]

It's going to be a success for the average consumer regardless, which is what matters in the end. Even if it doesn't (and I don't believe it will) match Intel's performances.

We already know that zen won't be at intel chips. At best, in specific apps, it's a little bit better or equal than broadwell-e.
Main interest for Ryzen is ratio performance/price where Intel is more or less weak due to absence of true competition.

Both AMD and INTEL consumers will win from this competition

 

[Sean Crees]

It looks like Ryzen is going to be a combo of 95w and 65w TDP parts.

http://wccftech.com/amd-ryzen-lineup-pricing-confirmed-8-cores-low-320/

The leak is from a retailer that apparently has a long history of getting AMD CPU product information up before official launches with accurate information.

Most expensive part is $500, with a 65w TDP R7 8c/16t for only $320.

 

[MarcParis]

If those prices are confirmed, AMD will ensure ryzen success.. fingers crossed..

 

[Phuncz]

If AMD fails, it hope it goes something like this:


(white = Intel, black = AMD)

 

[alexep7]

CanardPC: "There is a problem with the high-end Ryzen CPUs: 8c @ 3.6/4.0GHz for 95W TDP seems impossible with final stepping."

Ouch. Either the clock speeds will be lower or the TDP will be higher than expected. For most people this doesn't really matter, but for the SFF crowd it might be a big blow. I was excited about those promised 95W. I just hope the non-X version will still be 65W...

 

[MarcParis]

Still one month to wait...

 

[Phuncz]

CanardPC: "There is a problem with the high-end Ryzen CPUs: 8c @ 3.6/4.0GHz for 95W TDP seems impossible with final stepping."

Ouch. Either the clock speeds will be lower or the TDP will be higher than expected. For most people this doesn't really matter, but for the SFF crowd it might be a big blow. I was excited about those promised 95W. I just hope the non-X version will still be 65W...

I can hardly believe CanardPC isn't somehow fabricating this info. They'll most likely be enjoying the pageviews, but a CPU that's rumoured to be available within the month isn't going to be in a stage where it's not sure what TDP or clocks it's going to have. These are already finished, tested and being manufactured as we speak. Or else the March availability is not going to true.

 

[EdZ]

I can hardly believe CanardPC isn't somehow fabricating this info. They'll most likely be enjoying the pageviews, but a CPU that's rumoured to be available within the month isn't going to be in a stage where it's not sure what TDP or clocks it's going to have. These are already finished, tested and being manufactured as we speak. Or else the March availability is not going to true.

I'd suspect delays. For a massive CPU generation release that is little over a month and a half away at most, the only practical performance tests we have from AMD are:
- It's very competitive in one Blender benchmark (benchmark details publicly released)
- Competitive in a Handbrake on-stage demo (no benchmark details released)
- ZBrush and Keyshot demos (no performance numbers released)
- It can play Doom, Battlefield, Battlefront, and DOTA2 (on unknown settings, no framerate/frame timing numbers published)

We've got a grand total of one reproducible benchmark with hard numbers (Blender), and one unreproducible benchmark with a single point of comparison (Handbrake).

AMD are playing things very close to the wire if they're holding back on publishing real-world performance numbers this close to release.

 

[MarcParis]

My main concern remains this bug that already appears from f3 stepping. It's somehow related to smt.

On benchmark you forgot performance on cinebench where c8/t16, 3.4ghz @1188 multicore.
http://wccftech.com/amd-ryzen-zen-cpu-benchmarks-leak/