PCPer Mailbag #10 - 9/22/2017

Subject: Editorial | September 22, 2017 - 09:00 AM |
Tagged: video, Ryan Shrout, pcper mailbag, pcper

It's Friday, which means it's time for PC Perspective's weekly mailbag, our video show where Ryan and team answer your questions about the tech industry, the latest and greatest hardware, the process of running a tech review website, and more!

Here's what you'll find on today's show:

00:22 - PCPer Mailbag audio podcast?
01:14 - Games with DirectX, OpenGL, and Vulkan?
02:44 - Where are the AMD-based laptops?
06:51 - Does faster RAM = higher IPC?
08:28 - Using an iGPU with a discrete GPU?
10:55 - Why are Vega GPUs still so expensive?
14:41 - Do you need to reinstall Windows after upgrading CPU?
16:48 - How to minimize screen tearing without G-SYNC?
18:58 - Dummy dies and 32-core/64-thread Threadripper parts?
22:14 - The Cincinnati Bengals offense?

Be sure to subscribe to our YouTube Channel to make sure you never miss our weekly reviews and podcasts, and please consider supporting PC Perspective via Patreon to help us keep videos like our weekly mailbag coming!

Source: YouTube

September 22, 2017 | 03:10 PM - Posted by edwinjamesmiller36

"So you're telling me there's a chance." Jim Carey in "Dumb and Dumber"

BTW Nice try for free jerseys.

September 22, 2017 | 05:39 PM - Posted by DeepPocketsWantVega10 (not verified)

AMD's Raven Ridge(Zen/Vega APU) will have to arrive first before AMD makes any Real headway into the OEM laptop market. Though I'm looking at a gaming laptop with a current Bristol Ridge APU(Excavator/v2 28nm 3rd generation with "Volcanic Islands" integrated GPU graphics) and a discrete RX 560 GPU based gaming GPU(In the Laptop). And That Bristol Ridge even though it's a 28nm/Excavator/v2 design supports the AM4 motherboard and DDR4 memory, unlike the first generation Excavator/Carrizo SKUs.

With regards to why Vega is turing out to be more costly, it's because the professional market is wanting GPUs for GPU compute ability more than Graphics performance so it's compute(Miners/Others) demand that is causing more AMD Vega/Polaris GPU demand on the consumer side and more Demand causes the price to go higher if the supply is not there to meet Demand(Economics 101).

Also really even the Tech-Press in not really looking at this and it's another reason is that Vega 10, as designed by Raja under orders from Lisa and AMD's BOD, is a compute heavy GPU micro-arch that can also be used for gaming. So the professional markets(Compute/AI) gets first dibs on any Vega 10 "Cherry" die based SKUs for that market's needs. And the professional market will pay the proper mark-up without hardly a thought to get that compute/AI ability that's so in damand now that the GPU AI/compute market has taken off big time.

So gamers are not getting the top binned Vega 10 dies and neither are the miners. This is because of demand for the Vega 10(Top Binned/"Cherry") dies with the best thermal/clock performance metrics(Before Gamers/Miners get to fight over any lower binned dies) is coming from the professional markets and Vega 10's top binned dies are going into the Radeon Pro WX(Professional compute) and Radeon Instinct MI25(AI/infrencing) SKUs that get a much higher mark-up than and comsumer SKUs.

So that AMD/Server Partner Project 47 SKU uses 20 Epyc 32 core CPU SKUs and 80 Radeon Instinct MI25, Vega 10 die, based SKUs. So that's 80 Vega 10 dies used in those Radeon Instinct MI25 SKUs for one Project 47 system. And AMD/Server Partner(For Project 47) will be selling a lot of those systems at 80 Vega 10 dies needed at a time for that AI/inferencing workload market that has become all the rage.

So the miners are only competing with the gamers for the lower binned Vega 10 die based SKUs(Vega 64/56) that are available after the Professoinal compute/AI market demands are filled, and that AI market has taken off like a starship doing warp 10+. And those Professional Radeon Pro WX Compute/Radeon Instinct MI25 AI SKUs are gitting all the best Vega 10 silicon and earning AMD the proper mark-ups to really make the revenues. Don't forget the HBM2 dies also as the professional Radeon GPU Compute/AI SUKs are getting the best binned Samsung 8-Hi HBM2 die stacks, at 2/4 stacks, or more, per GPU SKU for the professional Radeon Pro WX 9100/Radeon Instinct MI25/AI SKUs.

Project 47 gets 80 Vega 10(Very Cherry) dies and that Project 47's Petaflops supercomputer SKU has a really great performance/watt metrics on those compute/AI workloads that will make the demand for Vega 10 dies continue to increase. And that's before the lower binned dies are available for the gamers and miners to get their hands on what Vega 10 dies are remaing that do not make the grade to be used in the professional SKUs. So the Vega 64/56 SKUs supplies will continue to be tight for a good long while.

SK Hynix needs to really get their HBM2 production issues fixed or Samsung will be the only game in town charging even more for its HBM2. DRAM prices and fab capicity is really at as near to 100% and it takes a few years to get any new plants/lines up and running, so that's not helping matters. And that all DRAM pricing is going up HBM2 and DDR4, GDDR5/5X/6 pricing, and NAND pricing also.

September 22, 2017 | 08:04 PM - Posted by Ipoopwhenifart (not verified)

Thumbnail caption:

"I just stealth farted on Allyn and Ken got blamed for it, damn I'm good"

September 28, 2017 | 01:58 PM - Posted by Sean (not verified)

Question for next mailbag: With the move away from 3 and 4-way SLI (and now crossfire as well), do you think we will see more of a focus on mATX motherboards (and consequently cases)?

Post new comment

The content of this field is kept private and will not be shown publicly.
  • Lines and paragraphs break automatically.
  • Allowed HTML tags: <a> <em> <strong> <cite> <code> <ul> <ol> <li> <dl> <dt> <dd> <blockquote><p><br>
  • Web page addresses and e-mail addresses turn into links automatically.

More information about formatting options

By submitting this form, you accept the Mollom privacy policy.