PC Perspective Podcast #531 - Radeon VII Review, New Logitech Headsets, and More!

Subject: General Tech | February 7, 2019 - 09:54 AM |
Tagged: Z390, W-3175X, tuf, scythe, radeon vii, podcast, owc, logitech, Katana 5, gaming headset, g935, g432, DLSS

PC Perspective Podcast #531 - 2/6/2019

Join us this week for a look at the AMD Radeon VII launch and benchmarks, a review of a new TUF Z390 board from ASUS, new gaming headsets from Logitech, and more!

Subscribe to the PC Perspective Podcast

Check out previous podcast episodes: http://pcper.com/podcast

Show Topics
00:01:09 - Review: AMD Radeon VII
00:32:20 - Review: ASUS TUF Z390 Motherboard
00:39:50 - Review: Logitech G935 & G432 Headsets
00:43:57 - Review: Scythe Katana 5 CPU Cooler
00:46:25 - Review: OWC USB-C 3.1 Drive Dock
00:50:18 - Sponsor: Molekule
00:53:01 - News: NVIDIA DLSS Driver Update
00:57:30 - News: ASRock DeskMini A300 Mini-STX PC
01:02:24 - News: Intel Xeon W-3175X Availability
01:07:26 - News: Liquid-Cooled DDR4 Memory
01:10:00 - News: Synology DS1819+ NAS
01:15:29 - News: Misbehaving Canadian Telcos
01:17:56 - Picks of the Week

This episode is brought to you by Molekule. Get $75 off your first order by using promo code PCPER at checkout.

Picks of the Week
Jim: Humble Paradox Bundle
Jeremy: Fanatical Ancestor Bundle
Josh: Samsung BAR Plus Flash Drive
Sebastian: Sony a6000 Mirrorless Camera

Today's Podcast Hosts
Sebastian Peak
Josh Walrath
Jeremy Hellstrom
Jim Tanous

Video News


February 7, 2019 | 01:29 PM - Posted by RushedOutBrokenOnRadeonVIIsomewhat (not verified)

I think that AMD may have lowballed the Driver Team's funding on Radeon VII hoping to offset the hardware BOM costs on that GPU SKU. So Folks are in for some needed driver fixes to be offered by AMD to fix all of what is broken. AMD also refactored their API on Radeon VII so a lot of the third party monitoring software is still in need of code refactoring as well to make proper use of the new API on Radeon VII for various GPU metrics monitoring tasks and such.

February 7, 2019 | 03:54 PM - Posted by Sean (not verified)

Hooray! The rundown has returned.

February 7, 2019 | 04:06 PM - Posted by Sean (not verified)

Nevermind, I guess it was added in post. I normally watch it live, so I don't see it.

February 8, 2019 | 01:06 AM - Posted by pdjblum

you fucker jim

i never listen to those fucking ads, but you had me hooked instantly, just as you did last week

fucking impressive

those sponsors can't be paying you guys enough for jim

and if you take the time to think about it, the vega 7 is brilliantly placed to maximize revenues and attention while readying navi

it might not be a 'gaming card' as some suggest, but it is the same as 2080 at the same price, which ain't too shabby

so the price for prosumers is insanely cheap and appreciated

and some of us gamers believe textures will take up more memory far sooner than ray tracing will be in more than a few titles

not to mention, the hardware on the 2080 can only do reflections without tanking fps when dlss is implemented to downscale and ai create the scene

so that is all you are going to get with the present hardware on the 2080

but the only thing i can say for sure is you fucker jim, you got me again

February 8, 2019 | 11:11 AM - Posted by FlashipGamingPoopDIEsForTheGamerScrubz (not verified)

Vega VII like GP102, and TU102 are a way to get some funds from the Professional market DIEs that do not make the grade to become the Quadros, Radeon Pro WXs/Radeon Instincts, and other Professional Parts. GP102 and TU102 have more quadro variants and only single gaming GPU variants that Nvidia brands GTX 1080Ti and RTX 1080Ti.

So all of Nvidia's and AMD's Flagship gaming cards are made from failed professional DIEs and it's those poop dies that the gamers will get at reduced cost. I'm sure that both AMD and Nvidia would rather those DIEs make the grade for Professional usage as the markups are well and proper to pay for the GPU Part's BOM/R&D Costs and still turn a tidy profit.

All the flagship Gaming GPU dies are failures as far as their makers are concerned and are harvested to bin out some usable SKUs that are they feed for the gamers at low margins. It's better than throwing those poop DIEs in the trash.

What you Red Gamng Scrubs have to wish for is a Dual Vega 20 poop DIEs on a single PCIe card variant as for sure there will even be lower binned Die turds that do not have even enough working shader cores to become Radeon VII poops.

Just think maybe AMD will take some 50 nCUs enabled Vega 20 dies that each still have 64 working ROPs but non sufficient working shader cores for even 60 working nCUs. So that would become some Radeon Pro Duo(Vega 20) Variant for Flagship gaming and Prosumer usage at $1200.

And Maybe since Vega 20 fully supports xGMI(Infinity Fabric) AMD could interface the 2 Vega DIEs on that single PCIe card's PCB with xGMI links instead of PCIe protocol links and fool the Software into thinking that it's just one big logical GPU with 100 nCUs and 128 ROPs. So the crappy gaming code will not have to be refactored to scale well across the 2 GPU DIEs.

If you look at that Dual Vega 10 Die based Radeon Pro V340 cloud game streaming/visualization/graphics SKU well that's 2 Vega 10 DIEs on a single PCIe card with the 2 Vega 10 DIEs each having a Vega 56 complement of Shaders:TMUs:ROPs. So there is always a possibility that some even lower binned Vega 20 DIE samples could be doubled up for a Radeon Pro Duo refresh.

AMD's all in with TSMC's 7nm and that's not only for the performance gains that's also because at 7nm that Vega 20 DIE's size/area is much smaller than Vega 10's Die Size/Area at 14nm. So that's more DIEs/Wafer on TSMC's 7nm and those 7.5T layout libraries are denser so that's one more reason that AMD loves that 7nm node. The Chip yields go up as a function of the DIE's area/size decreases and more DIEs can fit on a single Wafer.

Just remember gamers all of your Flagship Gaming SKUs are made from disappointing pro DIEs bins that the GPU makers would have really prefered to have made the professional binning grade. So That's Poops to you gamers and maybe a Double Vega 20 DIE Poop for AMD's gaming customers at some point in time with double the ROPs and Pixel Fill Rate, Double the VRAM/other metrics also.

February 8, 2019 | 06:07 AM - Posted by Anonymous123 (not verified)

Daaamn, now someone got me hyped because of an ad C-:

February 8, 2019 | 12:05 PM - Posted by BawHaHaChipzilla (not verified)

Now Adobe wants its own hardware(1) running on its cloud!
There is RISC-V that's an open source ISA with no licensing fees and now MIPS(Been aroud since 1981) ISA that's now open sourced(2) also. Or they could go ARM requires licensing($$).

There you go, AMD, maybe You can dust of your K12 custom ARM Verilog and your semi-custom folks could work up some semi-custom solution for Adobe. To make that even lower cost maybe AMD could switch K12 over to running the MIPS ISA, it all depends on what's lower cost in the long run.

"Intel must be starting to feel like Frito-Lay. It had an entire market to itself for a while, but now that other companies have started to sense weakness, it's hard to find someone that isn't planning to make its own chips. The latest would-be competitor? Well, according to Axios, it's Adobe.

That's right: the company that makes the software that creative professionals have to sell their organs to pay for each month is reportedly thinking of making its own processors. Or at least licensing Arm's designs so it can better integrate the software it's known for with the hardware on which it runs." (1)

I'm Actually more excited about the MIPS ISA going open source what with the MIPS software ecosystem being around as long as the ARM based software ecosystem has. So that's a pretty mature source for compilers/APIs/Other software ecosyetem tool chains and driver/firmware ecosystems as well that work with the MIPS ISA.

(1)

"Adobe Considers Manufacturing Custom Processors"

https://www.tomshardware.com/news/adobe-cpu-processor-arm-manufacture,38...

(2)

"MIPS Goes Open Source"

https://www.eetimes.com/document.asp?doc_id=1334087#

February 11, 2019 | 02:51 AM - Posted by Dark_wizzie

Radeon VII is a fail. I don't care about the memory bandwidth of a GPU or what process node it's made under. I care about price vs performance.

Also, did Pcper just review Asus mobo with zero mention of its lackluster vrms for z390?

February 11, 2019 | 12:19 PM - Posted by OthersMeasuureSucessAndFailureDifferentlyThanGamers (not verified)

If you are only a gamer and want the higest FPS no matter if it's overkill or not then maybe Radeon VII fails your expectations. But from a business standpoint AMD will probably still sell more Radeon VIIs to content creators and miners. Some initial mining benchmarks are startng to show up for R-VII and the underclocking/undervolting headroom on Radeon VII is not looking too shabby.

A Titan V's FP64 performance is 7,450 GFLOPS (1:2) for $2,999. And the Radeon VII's FP64 performance is 3,460 GFLOPS (1:4) at $699. So 2 Radeon VII's cost around $1400 and have a combined DP FP number of 6920 GFLOPS which is not bad for the price. Even 3 Radeon VIIs are still less than one Titan V with 3 Radeon VII's costing around $2100 and bring with that a total DP FP metric of 10380 GFLOPS and at $900 savings.

But realistically there will probably be some good numbers of dual Radeon VII based rigs out there and I really want to see the DX12 and Vulkan Explicit Milti-GPU Adaptor enabled game titles tested once the driver toothing pains are fixed on Radeon VII.

Hey maybe there will be a Vega 20 based Radeon Pro Duo released at some point in time as AMD did at one time release the Radeon Pro Duo(Fiji) and Radeon Pro Duo(Polaris) based Prosumer Dual GPU/Single PCIe card variants in the past. But this time around AMD has got its Radeon Instinct and probabily Vega 20 based Radeon Pro WX market sales to consider also that have xGMI fully enabled. So AMD has a larger Professional GPU marke to protect.

February 11, 2019 | 12:54 PM - Posted by JustMoreProsumerGoodnessFromRed (not verified)

Now some more good news(1) for Radeon VII's prosumer users!

"You uninstall your Radeon Adrenalin 2019 Edition drivers and replace them with the Radeon Pro 19.Q1 drivers to access pro features.

These include access to ProRender, certifications for various 3D, CAD, and CGI suites, SecureMI security, enterprise virtualization, and more. Over 320 professional applications are certified for the Radeon Pro 19.Q1 drivers, all of which will seamlessly run on the Radeon VII. AMD will also introduce a feature that lets you switch between the Radeon Pro and Radeon Adrenalin drivers on-the-fly (without needing reboots), so you don't lose your ability to play the latest games with day-one optimizations from AMD. These drivers will make the Radeon VII an incredible value in the enterprise space, as the GPU offers performance rivaling professional graphics cards priced well north of $3,000. It also blunts the feature-set edge the RTX 2080 holds over the Radeon VII." (1)

Now if AMD would just enable enough DP FP on Radeon VII so that 2 Radeon VIIs could match the Titan V's DP FP metrics of 7,450 GFlops. So Radeon VII's DP FP is currently at 3,460 GFLOPS (1:4) and maybe AMD could make that 1:3 in the GPU BIOS and get the DP FP closer to 4000 GFLOPS DP FP on single Radeon VII. Titan V cost $2999 and 2 Radeon VII's cost $1400.

Better yet where is the Radeon Pro Duo(Vega 20 Edition).

(1)

"AMD to Unlock Professional Features for Radeon VII to Blunt RTX 2080's Ray-tracing Edge"

https://www.techpowerup.com/252475/amd-to-unlock-professional-features-f...

February 11, 2019 | 11:21 PM - Posted by MoreSecurityResistantSuperBugsFromIntelInside (not verified)

More Bugs in the soup from Chipzilla(1)!

"Security researchers have found that Intel's Software Guard Extensions (SGX) don't live up to their name. In fact, we're told, they can be used to hide pieces of nasty malware that can silently masquerade as normal applications.

SGX is a set of processor instructions and features for creating a secure enclave in which code can be executed without scrutiny or interference from any other software – not even the operating system or hypervisor can look in. It's aimed at processing financial transactions, performing anti-piracy decryption of protected Hollywood movies, and similar cryptography in private away from prying eyes." (1)

(1)

"Intel SGX 'safe' room easily trashed by white-hat hacking marauders: Enclave malware demo'd"

https://www.theregister.co.uk/2019/02/12/intel_sgx_hacked/

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

CAPTCHA
This question is for testing whether you are a human visitor and to prevent automated spam submissions.