The AMD Execution Thread [2018]

Status
Not open for further replies.
Speculation based on precedence set by other non-top tier GPU makers combined with what ASRock themselves are saying, sure. But better than speculation based on a phone conversation with a middle management sales manager from an Asian country which has since been redacted by the original source (Tomshardware) as being erroneous and something they never should have posted.

Regards,
SB

It was redacted they said to protect the senior sales manager from problems with AMD, not as you mention for being erroneous.
No-one has yet proved he is a middle-manager; as someone who works with such VPs and sales channels,product development,etc with teams in multiple regions over the years I can say it is highly unlikely the employee is as low as you deem relative to the PR (who made multiple miscommunications in their contact with Forbes).
How many in-country local sales guys do you know where you work keep an eye out for a sales channel breach in Taiwan/Australia/etc on another continent and then take time to try and resolve it themselves rather than pushing it up the chain?

You mention speculation, but isnt that exactly what you did as well throughout your post?
Anyway the recent points from others keep ignoring one important aspect; this is against the backdrop of Scott's quote earlier on about freedom of choice for consumers/retailers and importantly said 'supporting partners' in context of greater flexibility/freedom for partners.
Currently that statement-marketing strategy looks awkward but only time will tell because all the information is either conflicting or does not line up, so now we need to see what happens 3-6 months down the line.

Edit:
Also I do not think some appreciate just how big some of those benefits are with regards to which tier a partner sits in; which includes logistics-parts available and IHV manufacturer scheduling,financial and market incentives.
Considering how restrictive the sales agreement seems to be for ASRock, it is fair to think they are on a pretty low one.
This for an established core large international partner albeit currently on the motherboard side with its own sales channel agreement, but this is not entirely the same situation as a new partner.
 
Last edited:
One outlet I checked does have Biostar AMD cards in stock. Could be wrong, but don't think Biostar offers their one nvidia card (GeForce GT1030 2GB) .
If I'm not mistaken this outlet is one of the hoards who pick up hundreds of boards for mining and sells them at insanely high premiums to other miners
 
If I'm not mistaken this outlet is one of the hoards who pick up hundreds of boards for mining and sells them at insanely high premiums to other miners
It's an outlet in the US that sells Biostar AMD cards. At the moment the most desirable mining cards are at insanely high premiums no matter where you look in the US.
 
It's an outlet in the US that sells Biostar AMD cards. At the moment the most desirable mining cards are at insanely high premiums no matter where you look in the US.
Those aren't even the most desirable ones, Micron memories are preferred apparently these days
 
AMD Will Release a Raven Ridge APU Graphics driver once each Three Months
However, as the adopters of AMD Ryzen 3 2200G and AMD Ryzen 5 2500X, driver support has been lacking. And for an APU with low-level gaming capabilities, that means they miss out on the better game optimization that dedicated graphics card owners do take advantage of. Early May AMD released an Adrenalin driver that included support for Raven Ridge, making a lot of people happy. From there onwards it was expected that all Adrenalin driver would get APU support. As it turns out now, things are not as unified on a driver level at it seems. Recent Radeon driver updates all lacked Raven Ridge support.

As it turns out now, there will be a driver for Raven Ridge every three months. Social media manager Matt, from AMD, responded with the announcement that only the WHQL versions of the drivers will include RR support and these are released once every three months, these thus offer support for the APUs. That means that any hotfix or beta driver will not support Raven Ridge APUs. And yes, that is going to disappoint a lot of consumers, as they really would like that 0-day game support on their APUs as well.

It's a little weird to see, Intel is moving to 0-day game support with their drivers for IGPs, and AMD is falling back to just four drivers per year for their APUs. Then again, at least you get four sets of WHQL drivers for your APU each year.

http://www.guru3d.com/news-story/am...u-graphics-driver-once-each-three-months.html
 
Not sure if this is the right thread, but AMD's joint venture with THATIC has spawned Dhyana CPUs from Hygon.
Dhyana is practically AMD Epyc, same Linux patches work for both if you correct the device and manufacturer id's.
Apparently they get pass the x86-licensing woes like this:
AMD owns majority of HMC, which then has access to AMDs x86-license. HMC then licenses that IP to Hygon so Hygon can develop a chip design to sell back to HMC, which then manufactures it wherever and then sells back to Hygon, which then brings it to market. AMD owns minority share of Hygon, too.


https://www.tomshardware.com/news/china-zen-x86-processor-dryhana,37417.html
 
If "IPC" means "performance/MHz" and 256-bit AVX units are involved, it may not mean much for games and other tasks. Still good news, though.
 
It's a field the current generation is already strong in, so good to see increased IPC at these workloads if the rumours are true.
 
If "IPC" means "performance/MHz" and 256-bit AVX units are involved, it may not mean much for games and other tasks. Still good news, though.

AVX in Zen is half speed, 13% for full speed AVX is bad, it should be more like ~80% if it's 100% AVX instructions.
If the benchmark is scaling capped because of data dependencies it'd be a crappy benchmark for IPC. An IPC benchmark should ideally guarantee arbitrary positive number of "Instructions per Clock". E.g. you can create artificially constructed loops for Zen with an achieved IPC of 4.
 
AVX in Zen is half speed, 13% for full speed AVX is bad, it should be more like ~80% if it's 100% AVX instructions.
If the benchmark is scaling capped because of data dependencies it'd be a crappy benchmark for IPC. An IPC benchmark should ideally guarantee arbitrary positive number of "Instructions per Clock". E.g. you can create artificially constructed loops for Zen with an achieved IPC of 4.

As I understood it, +13% is an average value over a suite of benchmarks, some portion of which may include a portion of AVX instructions.
 
New The entire tech sector is getting slaughtered this month but today was raw for AMD.
Their GPU sales is significantly down, the death of the mining craze hit them hard. They also cite high channel GPU inventory problems. So I guess they suffered the same problem as NVIDIA did with Pascal.

quarterlyearnings.amd.com/static-files/8ffabcf1-4674-43d7-9299-ab2c942a77b2
 
Last edited:
Status
Not open for further replies.
Back
Top