The AMD Execution Thread [2019]

Normally that's the case. However, when that happens PC shipments also fall. In this case, PC shipments rose, while discrete graphics shipments took a rather large tumble. PC shipments 9.25% while discrete shipments (AIBs) dropped by 16.62%. Or you might see a small uptick in PC shipments accompanied by a fall in discrete graphics shipments like in Q4 2018 when PC shipments rose 1.61% while discrete shipments fell. That was also still very close to when the crypto bubble popped so that exacerbated things.

I rarely look at Quarter on Quarter numbers due to seasonality, however, you can look at the relative performance of items within a market from quarter to quarter to see how things are progressing within the market itself.

It's kind of been the trend ever since Turing came out. But at first the picture was muddled due to the crypto bubble popping. But it's been almost a year now since Turing came out. I do wonder if AMD's release of RDNA will reverse the trend by making performant discrete GPUs a bit more affordable, but I'm not sure it'll be enough.

Much like Apple pushed the price bubble too high with the latest iPhones and suffered for it in shipments if not in revenue, same is happening with NV's GPU shipments.

Regards,
SB

 

Actually the trend didn't change much since Turing, NVIDIA is still in a commanding lead, that increases or decreases according to seasonality. For Q2 I think we are looking at a stock clearance situation, old 2080, 2070 and 2060 models are cleared in preparation for their Super replacements.

 

New card shipments generally bumb up the numbers in the quarter before the actual launch, so Supers are already in there.

 

That means Navi numbers are in there too.
I somehow doubt that.

 

Yes, Navi 10 numbers should be there too

 

I find this really odd.
The only time AMD got a marketshare jump this big was when they had clearly superior GPU offerings, and think that last time that happened was when the DX11 TeraScale 2 competed with the 1st-gen Fermi.
I don't see the 5700/XT as having that kind of advantage at all.

 

How AMD gained market share

https://www.fudzilla.com/news/graphics/49323-how-amd-gained-market-share

 

That article seems to be just a pile of crap. He's denying impact of Navi argumenting "Navi started shipping for revenue in July" (Q3). AMD had to ship all the reference boards in June (Q2) to be able to launch the card at the beginning of July. So AMD sold them in Q2.
As for the impact of RX 570 / RX 580 with the exception of a few models, prices were falling slowly and continuously in most markets. There was no big sudden discount which would make Polaris significantly cheaper in Q2 than in Q1. According to ComputerBase even GTX 1060 got more significant price discount during Q2 (20 %) than RX 580 (9 %) or RX 570 (13 %). Historically AMD brought more significant discounts of last-gen parts in past, but none of them resulted in sales increased by 1.5×. In this case, market was oversupplied by Polaris after the end of mining craze it was extremely cheap on second hand markets and nobody bought it despite even bigger price reduction. I highly doubt that ~10 % average discount in 2019 would change mind of (+50 %) AMD customers.

 

On the other hand, one Navi chip covering narrow market segment can hardly be the explanation.

 

Two competitive models with very good price / performance ratio can't reach 9 % of the discrete market share? What would be the reason? Polaris is more than 3 years old now and quite a lot of users are looking for upgrade. GTX 1660 / GTX 1660 Ti offers 15-30 % higher performance for $250-300, so these are hardly a worthy upgrade. Radeon RX 5700 (XT) and RTX 2060 (S) are exactly those products, which are worth upgrading (55-90 % higher performance), but aren't excessively expensive.

 

I think Tahiti and Hawaii both had models that were clearly better than their nvidia competition. They consumed more power but thats of little consequence IMO. Id also say polaris was clearly a better buy than the 1060.

 

AMD's earnings last quarter show that whatever GPU's AMD sold to gain those 9% must have been low revenue and low margin.

Fire sale prices on 570's and 580's fit that bill.

You really have to have been blind to not have seen the constant DEALS on those over the last quarter.

 

Argumentum ad hominem. Nice.

Anyway, it's nonsense. Those deals were not Q2-related, they ran since the end of the last year:

November 2018 - RX 580 for $165: https://www.pcworld.com/article/332...-best-graphics-card-deals-in-a-long-time.html
February 2019 - RX 580 for $169: https://www.tomshardware.com/news/msi-amd-rx-580-newegg-deal,38580.html
April 2019 - RX 580 for $160: https://www.pcgamer.com/this-overclocked-radeon-rx-580-8gb-is-the-cheapest-around-at-dollar160/

This can hardly explain marketshare jump between Q1 2019 ad Q2 2019.

 

https://venturebeat.com/2019/10/29/...enue-as-it-ships-more-high-end-cpus-and-gpus/

AMD sees $1.8 billion in Q3 2019 revenue thanks to high-end processors and GPUs

In the third quarter ended September 30, Advanced Micro Devices reported earnings that roughly matched Wall Street’s expectations for the maker of processors and graphics chips.

Santa Clara, California-based AMD is benefiting from a rare advantage in chip design over archrival Intel, which has been slow on both the manufacturing and design fronts to come up with a processor that challenges AMD’s Zen and Zen 2-based designs.

AMD CEO Lisa Su said in a statement, “Our first full quarter of 7-nanometer Ryzen, Radeon and Epyc processor sales drove our highest quarterly revenue since 2005, our highest quarterly gross margin since 2012, and a significant increase in net income year-over-year. I am extremely pleased with our progress as we have the strongest product portfolio in our history, significant customer momentum, and a leadership product roadmap for 2020 and beyond.”

 

AMD today released the long waited 16-core Ryzen 9 3950X for AM4 and previewed Ryzen Threadripper 3960X (24C/48T) and 3970X (32/64T) for new TRX40-platform and sTRX4 socket. Threadrippers have 88 PCIe lanes total which of 8 + 8 have been reserved for CPU-Chipset communication so 72 usable lanes and 4 channel DDR4-3200 support.
Also new Athlon 3000G APU (unlocked for OC) to beat the crap out of Intel Pentium & Celeron

In addition, according to Mercury Research AMD gained market share in all x86 segments, servers excl. IoT up 0.9 to 4.30%, desktop excl. IoT up 0.9 to 18.00%, mobile excl. IoT up 0.7 to 14.70%, client excl. IoT up 0.8 to 15.80% and overall x86 share less semi & IoT up 0.7 to 14.60%
Highest shares AMD has had since 2013/2014

 

ِApparently that was the true explanation of the Q2 AMD jump, because as of Q3, AMD shares took a hit to 27%, NVIDIA rose to 73%. NVIDIA increased sales tremendously and raised the AIB market by 42%.



https://www.jonpeddie.com/press-rel...t-soars-led-by-nvidia-reports-jon-peddie-rese

 

How on earth can you pull that kind of conclusions from that?
- RX500 prices have been the same since November last year, there was no "firesale" in Q2
- If a company which has ~68% marketshare increases their sales by 42%, their share won't jump by mere 5%. Both AMD and NVIDIA increased their sales significantly, NVIDIA just more than AMD but not that much more since the share shifted by only 5%

 

How AMD gained market share
https://www.fudzilla.com/news/graphics/49323-how-amd-gained-market-share

Yes it really was fire sale prices of older products in Q2 2019 that moved the needle for AMD and not NAVI for if it was NAVI then why the retraction of 5% in market share for AMD and the paltry 473 thousand unit increase Q-Q.

You are misreading market share increases with overall add-in board market expansion. The add-in board market expanded by 42% not Nvidia's unit sales. Note Nvidia's unit sales actually increased by 52.8% Q-Q: See math below showing why.

Read the full report here:

Global Q3’19 add-in board market soars led by Nvidia, reports Jon Peddie Research
https://www.jonpeddie.com/press-rel...t-soars-led-by-nvidia-reports-jon-peddie-rese

Here is the math for AMD and Nvidia unit sales for Q2 and Q3:

In Q2 7.384 million AIBs shipped = 10.5 million / 1.422

Q2:

AMD: 7.384 million x 32.1% = 2.37 million units
Nvidia 7.384 million x 67.9% = 5.01 million units

Q3:

AMD: 10.5 million x 27.08% = 2.843 million units
Nvidia 10.5 million x 72.92% = 7.657 million units

Q-Q: Nvidia increased unit by 2.647 million units vs AMD's 473 thousand unit increase

Q-Q Unit increases:

Nvidia 52.8% (7.657 / 5.01)
AMD 20% AMD ( 2.843 / 2.37)

From this Q3 quarter onward older products from both Nvidia and AMD won't be of any significance it will be all Turing and Navi products.

 

Looks like exactly the opposite actually happened in Q3. Maybe the companies selling these things actually know what the market will bear.

 

If you are referring to the Jon Peddie link from the NV thread.

So, everyone still isn't back to where they were a year ago (total GPU shipments down 5.6% Year to Year). It is, however, encouraging that everyone had seasonal growth from Q2 to Q3.

However, if you look back at Q3 2018 (https://www.jonpeddie.com/press-rel...ch-reports-gpu-shipments-up-in-q318-from-las/ ), that also had a 2.2% decrease in Year to Year shipments.

GPU shipments are still trending downwards. Q3 almost always sees a seasonal increase in GPU shipments compared to Q2. At least this year the seasonal increase was larger than last year, otherwise this years shipments could have seen a rather precipitous decline. However, the yearly trend for GPU shipments continues to go downwards.

Which means everyone still isn't back to where they were 2 years ago.

Regards,
SB