• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

Thread Ripper 2, and the effect on Intel pricing

@Hotwired https://www.theregister.co.uk/2018/08/06/amd_epyc_intel_xeon_x86_server_revenue_share/
Mercury Research estimated Epyc is at only 1.3% in Q2 2018, with Intel making up the remaining 98.7%.
They gained 0.3% from Q1 2018 and are up from 0.5% in Q2 2017, total Epyc revenue for Q2 2018 sits at $57.66m which isn't really amazing because Epyc is a high margin product they need to ship in volume to improve their margins.

@Hades It's closer to Phenom II, very competitive and they're gaining a lot of market share. Back then Intel was more competitive price wise with their Core 2 Duo offerings, but the 6 core Phenoms were pretty great value and performance. Athlon was clearly superior to anything Intel had to offer, while not the same can be said about Ryzen imo, it's competitive, but maybe with Zen 2 they'll knock it out of the park.
 
Hades said:
I'm coming at this from a very uninformed position so please excuse me if I'm talking out of my rear. But from my standpoint Ryzen feels very much like their Athlon 64 felt like. Suddenly AMD could not only compete but were out in front.
Click to expand...
Ryzen and TR are amazing. No doubt.
But it's not as epic as athlon 64. That was leaps ahead of Intel is just about every use case. Best for games, best for work, best for everything. Faster, cheaper, you name it.
Ryzen is not a repeat of that. To have another athlon 64 moment you would have to have amd release ryzen the way it is now, at the price it is now, but running at 5.5Ghz.
 
AndreiD said:
Epyc's adoption thus far has been pretty poor though, think the MCM/NUMA latency issues are keeping some away, possibly the lower per core performance (mainly AVX) too since a lot of software is licensed based on # of cores.
Click to expand...
I doubt that numa latency issues are going to factor into the decision considering most servers that require high core counts are already multisocketed and have the usual numa related latencies to deal with already, be it Xeon or Epyc.

The lacklustre avx performance is possibly the more likely of the two though I imagine it really depends on whether the workloads even utilise avx instructions.

I imagine that past history is hard to ignore where amd all but abandoned the server market once the opteron series got left for dust post Athlon64. (Around nahelem time)
 
Last edited:
AndreiD said:
Epyc's adoption thus far has been pretty poor though, think the MCM/NUMA latency issues are keeping some away, possibly the lower per core performance (mainly AVX) too since a lot of software is licensed based on # of cores.
Click to expand...
First. AMD started from 0% on the server market.

The third party server certification process was completed last month. That matters a lot on the server market and is a process that takes up to 12 months when a new architecture comes out. (same applies to Xeon).
Is not like desktop where people can jump and buy new CPUs.
However since this process ended, we started to see already mainstream servers using EPYC since mid June.
AMD predicts to have a modest 5% market share by end of the year. That is billions in new income and the orders are there according to AMD, the certification had to be completed though.
 
Yup, the issue is that Intel had literal complete dominance in servers, even taking 1.3% is actually a monumental step. It means that with no in place working plan to validate/certify servers companies still took not of EPYC and started that work. EPYC 2 also adds to the belief that server guys should bother with AMD. EPYC 1 alone was a single chip and if AMD couldn't execute another step fairly quickly it would be seen as a flash in the pan. EPYC 1 is showing off Zen to the server market and they were absolutely interested, EPYC 2 is the thing that makes the server guys commit, which is why as EPYC 2 taped out a whole bunch of server makers and companies who buy servers got a whole lot more interested.

With EPYC 1 server guys mostly validated and worked on certification and building their own computers after it launched. With EPYC 2 they are doing much of this work during the sampling phase and concurrently with AMD also doing this stage of work. So with EPYC 1 there is a ~1 year lag between launch and the servers coming out, with EPYC 2 it should be possible that a few servers launch alongside the EPYC 2 launch and a lot more within the first few months. With the same socket and similarity of the platform making the second lot of servers will also be much much less work.

There is also issues in terms of volume, with Glofo it's hard to say how much production AMD were using but Glofo had other customers also and AMD was making it's whole line up at Glofo. With TSMC able to be involved due to the wafer agreement change it opens up vastly more volume for AMD.

1.3% is a much much bigger step, in the same way the saying goes that the first million is the hardest, once you get a foothold and once you get people actually building and validating servers and it's actually selling it's easier to convince the same company to make 5 different types of servers and then see them all selling the next time. THink of it like this, new product comes along, committing to build each specific server build might cost a few million, a bunch of work and a commitment to support that server for multiple years but at that time you don't even know if people will buy it because it's a new player. So you wait for the chips to launch, you build one box to limit exposure and dip your toe in the water of the market and see if anyone bites. When those servers sell easily (as I'm under the impression they are) you decide to build 10 different server configurations. You already know demand is there, you know you can sell them and you've been convinced. A year from now when those 10 different configs are available and knowing it will actually be the far better product than the competition this time around, you have no worry they'll fly off the shelves.
 
Lokken86 said:
Are you sure it's only 5k at stock? My 1950x gets dam near 3k at stock.
Click to expand...

Its at 3.0 to 3.4Ghz depending what its boosting to, the 1950X has a base clock of 3.4Ghz with an all core boost of 3.6Ghz, which doesn't mean all cores are running at 3.6Ghz, as Pants said there are TDP limitations at stock, its a 32 core CPU, the highest core count ever on a single package.

Also take into account core count scaling, its not 1:1
Its 70% faster than a 1950X, take scaling and lower clock speed in to account its probably about right.

rs4QzXt.png
 
4K8KW10 said:
It scales like crazy.
After the 32-core 2990WX, we don't need more cores, just die shrinks to increase the frequency and decrease the power consumption.
Wow!!
Click to expand...
Here's hoping even cpus like this or original threadripper will be good for years at least matching the 3570k for longevity. Now if only games would scale up with more then 4 cores we'll be laughing.
 
Lokken86 said:
Here's hoping even cpus like this or original threadripper will be good for years at least matching the 3570k for longevity. Now if only games would scale up with more then 4 cores we'll be laughing.
Click to expand...

Screw games. As long as productivity apps keeps scaling I'm good.
 
amigafan2003 said:
Screw games. As long as productivity apps keeps scaling I'm good.
Click to expand...
Wouldn't mind a bit more optimisation on adobe stuff. Still got plenty of cores to spare, not really a problem because I usually render projects in after effects and occupy myself with Photoshop and dreamweaver or YouTube whilst it goes but If I was just to do one render it would be nice to focus all the cores onto it.
 
Lokken86 said:
Wouldn't mind a bit more optimisation on adobe stuff. Still got plenty of cores to spare, not really a problem because I usually render projects in after effects and occupy myself with Photoshop and dreamweaver or YouTube whilst it goes but If I was just to do one render it would be nice to focus all the cores onto it.
Click to expand...

Yeah, the Adobe suite is a bit **** on Ryzen/Threadripper. Then again, it doesn't exactly make the most of Intel HEDT either!
 
HEDT is a drop in the ocean but much more significant will be how EPYC 2nd generation is priced and how Intel respond to that.
If Intel do release that 28C HEDT+ platform then I can imagine the motherboards being ridiculously expensive due to the complexity and the very low volume for the platform.
Maybe they will start at £400-£500 and the CPUs might as well be POA! :)
 
The 2990WX is £1600, Intel are not going to be dropping no 28 core for anything less than £3000 and at that price no one is going buy it.

AMD have disrupted HEDT to such an extent Intel just can't compete.
 
You must log in or register to reply here.
Back
Top Bottom