Speculation: AMD's response to Intel's 8-core i9-9900K

[AtenRa]

Odd reply. Yes, obviously AMD hasn't reached core for core parity. I don't see anyone claiming as such. Good point?

But AMD has reached chip for chip lead in many workloads, let alone chip for chip parity. Which is why the pricing is as it is at the moment.

And as I gave an easy example of above, there are plenty of people blinded by brand loyalty which don't even accept that. But this is the reason this whole thread exists. AMD currently has the performance lead in certain segments and they will loose this lead when intel 8-core mainstream chips come along. So the discussion is what will happen in the situation when AMD loses the performance lead in all mainstream metrics.

You believe that Intel will price the 8-core mainstream SKU at the same level as AMDs R7 2700X ??? Because you dont loose your performance lead if the competition still has 6-Cores at the same segment(price) like 2700X vs 8700K.

 

[coercitiv]

You believe that Intel will price the 8-core mainstream SKU at the same level as AMDs R7 2700X ??? Because you dont loose your performance lead if the competition still has 6-Cores at the same segment(price) like 2700X vs 8700K.

It's called performance lead for a reason, not price/performance. Meanwhile, the 9000 series i7 may still end up being an 8 core. See this.

 

[piesquared]

There was quite a commotion at one time to try and determine if the latency between CCXs had any significant impact on anything. If i remember right the naysayers won the day claiming that the latency had a big impact on performance. Given how well it already performs and winning in many multithreaded workloads that bodes really well for Zen 2 since increasing that efficiency is probably a major improvement coming and that alone should have a sizeable impact on performance. Multithreaded performance is now probably more important in the industry than a single cores performance.

 

[eek2121]

AMD had to know in advance that Intel would eventually respond by upping the core count. They are probably binning chips for both Threadripper and the 2800X and waiting to see if an 8 core drops.

I'm still betting on a 2800X release if Intel releases early, or they will just clear stock and get ready for 7nm if Intel releases too late. Keep in mind that AMD has historically been releasing around the April time frame. I'm willing to bet that 7nm Ryzen will drop around the same time next year. It's now July, and I doubt we will see anything from Intel for the next couple months at least...maybe when Threadripper 2xxx drops. Quite a few of us 1950X folks could get our chips to 4.2, even 4.3 GHz occasionally. I've had my chip as high as 4,450 MHz at obscene voltages. That's Ryzen gen 1. If I disable 8 cores I can run 4.2 easily and pass prime95. 4.3 requires 1.475V, but a Zen+ 2800X would push that to 4.5 or 4.6...right in line with anything Intel would be able to offer. Threadripper 1 also had improvedL1/L2, etc. latency over Ryzen gen 1. Who knows what Threadripper 2xxx will have. Any of those dies reserved for Threadripper 2xxx could be used in a Ryzen 2800X offering.

Note that I'm not a fanboy for either party, but if I were the CEO at AMD, I know how I'd play my cards...I'd stack the deck.

 

[AtenRa]

It's called performance lead for a reason, not price/performance.

Ok fair enough, I was thinking in the line of same Price segments.

 

[epsilon84]

That's what I suspect as well. 6 Cores of CFL at 5Ghz, OR, 8 Cores of CFL, at... ??? Ghz? 4.5? Higher? I really doubt 5Ghz though. Maybe Asus will deploy an "OC socket", if such a thing is possible these days. (Happened before.)

Does a 2700X overclock any worse than a 2600X? Not really. It runs hotter, sure, 33% more cores, but with adequate cooling they hit similar clocks. Think you'll see the same with the '9900K', with good cooling 5.0GHz should be possible like every CFL chip to date.

To the OP, I think a 2800X is possible, even likely, but like Stilt said, a small clockspeed bump isn't going to change much when the competition is introducing a 33% core count increase.

I'm not convinced AMD will have to adjust prices much though, unless Intel somehow prices these chips around $400. If these are $500+ as people speculate then there is nothing wrong with AMD charging $300 for a chip that would be, what, 10 - 15% slower perhaps?

 

[richaron]

You believe that Intel will price the 8-core mainstream SKU at the same level as AMDs R7 2700X ??? Because you dont loose your performance lead if the competition still has 6-Cores at the same segment(price) like 2700X vs 8700K.

Not a snowflake's chance.

But I do think any advantage AMD has with their 8 (vs 6 cores) from intel is less if intel also has an 8 core chip. Maybe you've missed the bunch of times I've said it, so I'll just copy & paste for you.

...why I think AMD will drop prices of current line, perhaps with the help of a 2800X release. Even though I don't think intel will even attempt to price the 8 core consumer part competitively (it will be relatively high) I think if AMD looses the mainstream multi-core advantage it will drive them to push for price advantage until next gen.

*bold added by me

Please read my posts in thread properly before questioning.

 

[epsilon84]

I have a question. Where do you think this 8 Core CF part will land on these hardware.fr charts?
https://www.hardware.fr/articles/974-19/indices-performance.html

Note that 8700K has super high all core turbo clock and it scores 281.4pts on desktop workloads average chart. On average, going from Ryzen 2600X to 2700X (normalized for slight difference in clocks) we gain ~22.8% more performance. Similar should go for CF 6->8C part. Unless intel somehow magically can force the same aggressive all-core Turbo clocks on new 8C part(s) I don't see a reason why AMD should do anything at all, even price could remain the same. 8C CF should be be scoring around 2700X/7820X or slightly above. On gaming chart nothing will change, 8700K will still be the top dog and 8C CF part could be on that level or maybe slightly lower on the chart- for AMD this changes nothing.

AMD is now price competitive with 8700K while clearly being superior value for the money(except for having slightly lower low-res. fps when paired with highest end cards, but even that can be mitigated with high clocked DDR4 memory to some extent). I'd say AMD is good to go as Pinnacle Ridge is well positioned against any CF part.

So AMD can achieve similar all core turbo clocks between a 2600X and 2700X yet Intel will need to 'somehow magically' do the same when going from 6C to 8C? Something doesn't compute with this logic. Also, not sure what you mean exactly by a 'super high' 8700K all core turbo, it is 4.3GHz, unless MCE is enabled, which I don't think is the case with these tests.

All things being equal, an 8C CFL chip should show a similar lead over a 2700X to what a 8700K does over a 2600X, which going by those charts, would be ~11% on average.

So I would actually agree about the pricing, AMD is already pricing the 2700X aggressively, they do lose the desktop performance crown and Intel gains a bit of positive PR, but value conscious buyers who want 8 cores will still likely opt for AMD unless Intel prices the 8C CFL chips close to the 2700X, which I highly doubt.

 

[TheELF]

There was quite a commotion at one time to try and determine if the latency between CCXs had any significant impact on anything. If i remember right the naysayers won the day claiming that the latency had a big impact on performance. Given how well it already performs and winning in many multithreaded workloads that bodes really well for Zen 2 since increasing that efficiency is probably a major improvement coming and that alone should have a sizeable impact on performance. Multithreaded performance is now probably more important in the industry than a single cores performance.

They only win if they have more cores though and if the intel one is limited to the low clocks of the zen...
so for the industry choosing is still a very complex matter although most of the times the only thing the industry cares about is how much processing power they can fit into their server room.
(And intel has like 8x socket mobos that takes 16c/32t cpus)

 

[maddie]

They only win if they have more cores though and if the intel one is limited to the low clocks of the zen...
so for the industry choosing is still a very complex matter although most of the times the only thing the industry cares about is how much processing power they can fit into their server room.
(And intel has like 8x socket mobos that takes 16c/32t cpus)

This merited a chuckle.

Most probably next year we might see 2x socket motherboards with 64c/128t or at least 48c/96t Rome CPUs.

 

[moinmoin]

(And intel has like 8x socket mobos that takes 16c/32t cpus)

Related read: Serve the Home: How the Intel Xeon Platinum is a Marketing Headache
Xeon Platinum are the Intel chips allowing for bigger than 2x sockets configurations. The market so far overwhelmingly prefers Xeon Gold which is limited to 2x sockets but offers the same capability as Platinum otherwise at a much lower price.

 

[jpiniero]

Gold 6xxx allows for quad socket.

 

[moinmoin]

You're correct, of course. Damn Gold 5xxx muddling the clear separation...

 

[Abwx]

I have a question. Where do you think this 8 Core CF part will land on these hardware.fr charts?
https://www.hardware.fr/articles/974-19/indices-performance.html

Note that 8700K has super high all core turbo clock and it scores 281.4pts on desktop workloads average chart..

325-330 pts at most, assuming the same TDP as the 8700K, FTR it s 110W with Prime 95...

 

[spikespiegal]

What's the anti-intel mindshare like?

Just askin'...

Goes like this...basically you think more cores that do less work makes your 'puter multitask better when there's no benchmark or application set I've ever seen that proves it. Other than that, I have no issue with AMD...just wish they'd quite with the nonsense about more cores.

Been running heavy multitasking environments like RDS and Citrix farms for almost two decades that make your gamer desktops seem like they are running DOS in comparison, and I'll take faster cores over slower ones anyday. LOL HP moonshot LOL. When I retire a dual or quad socket server that's 6 years old running 8/12/16 cores or more, and replace it with a single socket / 6 core box, and all the apps on the new box run drastically faster I'm not exactly buying into AMD's line that more cores are better because of some mythical claim about CPU scheduling.

 

[Markfw]

Goes like this...basically you think more cores that do less work makes your 'puter multitask better when there's no benchmark or application set I've ever seen that proves it. Other than that, I have no issue with AMD...just wish they'd quite with the nonsense about more cores.

Been running heavy multitasking environments like RDS and Citrix farms for almost two decades that make your gamer desktops seem like they are running DOS in comparison, and I'll take faster cores over slower ones anyday. LOL HP moonshot LOL. When I retire a dual or quad socket server that's 6 years old running 8/12/16 cores or more, and replace it with a single socket / 6 core box, and all the apps on the new box run drastically faster I'm not exactly buying into AMD's line that more cores are better because of some mythical claim about CPU scheduling.

Mythical is right. I have no idea where you got that. And when you replace a 6 year old box, the technology has increased to much that its hard to even consider how much faster the new ones are.

 

[PeterScott]

AMD had to know in advance that Intel would eventually respond by upping the core count. They are probably binning chips for both Threadripper and the 2800X and waiting to see if an 8 core drops.

I'm still betting on a 2800X release if Intel releases early....

Unlike forum tit for tat rhetoric, companies don't respond in any kind of timely fashion to their competitor products.

If the current GF 12nm process was really capably of running that much faster there should be some chips like that in the wild. 100 or 200 MHz is likely the best they could squeeze out, and that will be pretty inconsequential. Sitting on a huge pile of binned chips for months is just losing money in tied up, depreciating inventory.

AMD will do what it always does when Intel is ahead. Offer lower prices.

Intel will probably offer 8 core CL as a VERY expensive part, priced well above AMDs top 8 core Ryzen. By the time this happens AMD will likely have already saturated the market and will have a slight price drop to further sweeten the deal, and maybe offer some nice package deals on TR CPU/MB combos that are closer in price to 8 core Coffee Lake.

But a faster binned Ryzen 8 core: very unlikely.

 

[whm1974]

Well at Newegg, last years Ryzen 7 1700 is $220 while this years 2700 is $295:
https://www.newegg.com/Product/ProductList.aspx?Submit=ENE&IsNodeId=1&N=100007671 601294612
Not bad at all for 8c/16t decent performing CPU. Once Coffee Lake 8c/?t CPUs are out, then we could expect a price drop on the 8c/16t Ryzens, or maybe not. If Intel prices those CPU really high then perhaps AMD wouldn't need to do anything at all.

 

[eek2121]

Unlike forum tit for tat rhetoric, companies don't respond in any kind of timely fashion to their competitor products.

If the current GF 12nm process was really capably of running that much faster there should be some chips like that in the wild. 100 or 200 MHz is likely the best they could squeeze out, and that will be pretty inconsequential. Sitting on a huge pile of binned chips for months is just losing money in tied up, depreciating inventory.

AMD will do what it always does when Intel is ahead. Offer lower prices.

Intel will probably offer 8 core CL as a VERY expensive part, priced well above AMDs top 8 core Ryzen. By the time this happens AMD will likely have already saturated the market and will have a slight price drop to further sweeten the deal, and maybe offer some nice package deals on TR CPU/MB combos that are closer in price to 8 core Coffee Lake.

But a faster binned Ryzen 8 core: very unlikely.

Given that the PB + XFR can hit 4.3 GHz, and some users have reportedly actually gotten 4.4 GHz or even 4.5 GHz out of their 2700X chips, I'd say it's safe to bet that AMD could drop a 4.5-4.6 GHz 8 core pretty easily. Intel at 8 cores is either going to need extreme TDP to be able to stay at 5 GHz or it will likely have a slower boost than the 6700k...a Mythical 2800X vs this supposed 8 core chip would be a much closer race than you'd believe. Intel leads in clockspeed, but it also has higher TDPs than AMD chips. That becomes exponentially problematic as you increase cores.

By comparison, most users needed 1.375+V to hit 4.0 GHz depending on the chip. Meanwhile, the 1950X binned cores could do the same clock speed at 1.2V in many instances...and that's with 16 cores! My 1950X is capable of pushing 4.1 at 1.325-1.35V (can't remember, have an odd issue with my AIO that causes the CPU to heat up during prime95, but it worked fine on my other board). I believe 4.2 GHz was 1.45V - 1.475V, however, the motherboard could not handle pumping that much power and throttled. Just based off current data, you can expect that AMD could compete pretty easily if they want to. My thought? It depends on Intel's release of the 8 core chip. If it's close to time to clear inventory, I fully expect them to just clear inventory and wait for zen 2. If Intel releases it in 1-2 months, we'll probably see a 2800X.

 

[SlowBox]

Threadripper2 ? with 32c/64t

Yeah with 64 threads you can play like 10 games at once while you render in the background and have 30 windows open. For 64 threads the least you can do is accommodate it with lots of RAM. Man you wouldn't have to upgrade for 15 years. Ok maybe 10 years...I mean games will use at the most 6 cores and 2 threads at 4k resolution everything maxed out FSAA all that jazz on the heaviest engine game. However just FYI, I play Unreal Tournament Alpha and the graphics are better then any game sorry to say and it uses like avg 7 percent of my CPU power, So ya a gamer wants less cores and faster cores,,, sorta

 

[PeterScott]

Given that the PB + XFR can hit 4.3 GHz, and some users have reportedly actually gotten 4.4 GHz or even 4.5 GHz out of their 2700X chips, I'd say it's safe to bet that AMD could drop a 4.5-4.6 GHz 8 core pretty easily. .

Where?

 

[Markfw]

Yeah with 64 threads you can play like 10 games at once while you render in the background and have 30 windows open. For 64 threads the least you can do is accommodate it with lots of RAM. Man you wouldn't have to upgrade for 15 years. Ok maybe 10 years...I mean games will use at the most 6 cores and 2 threads at 4k resolution everything maxed out FSAA all that jazz on the heaviest engine game. However just FYI, I play Unreal Tournament Alpha and the graphics are better then any game sorry to say and it uses like avg 7 percent of my CPU power, So ya a gamer wants less cores and faster cores,,, sorta

OK, this is silly. You don't buy a 32 core cpu for gaming, its for productivity. You can only play one game at a time. This is trolling.

 

[ZGR]

Where?

Pretty sure that is with BCLK overclock with XFR.

 

[whm1974]

OK, this is silly. You don't buy a 32 core cpu for gaming, its for productivity. You can only play one game at a time. This is trolling.

Agree, I highly doubt anyone with a shred of sense would use more then a 8c/16t CPU for a gaming building anyway.

 

[epsilon84]

Given that the PB + XFR can hit 4.3 GHz, and some users have reportedly actually gotten 4.4 GHz or even 4.5 GHz out of their 2700X chips, I'd say it's safe to bet that AMD could drop a 4.5-4.6 GHz 8 core pretty easily. Intel at 8 cores is either going to need extreme TDP to be able to stay at 5 GHz or it will likely have a slower boost than the 6700k...a Mythical 2800X vs this supposed 8 core chip would be a much closer race than you'd believe. Intel leads in clockspeed, but it also has higher TDPs than AMD chips. That becomes exponentially problematic as you increase cores.

By comparison, most users needed 1.375+V to hit 4.0 GHz depending on the chip. Meanwhile, the 1950X binned cores could do the same clock speed at 1.2V in many instances...and that's with 16 cores! My 1950X is capable of pushing 4.1 at 1.325-1.35V (can't remember, have an odd issue with my AIO that causes the CPU to heat up during prime95, but it worked fine on my other board). I believe 4.2 GHz was 1.45V - 1.475V, however, the motherboard could not handle pumping that much power and throttled. Just based off current data, you can expect that AMD could compete pretty easily if they want to. My thought? It depends on Intel's release of the 8 core chip. If it's close to time to clear inventory, I fully expect them to just clear inventory and wait for zen 2. If Intel releases it in 1-2 months, we'll probably see a 2800X.

Wow, where did you find these golden 2700Xs that can do 4.5GHz? Some sources would be great.

I think it's virtually impossible for AMD to release a 2800X that turbos to 4.5GHz+, unless you mean for single core turbos, which is possible, considering a 2700X can already do 4.3GHz.

But all core turbo will be where the battle will be won and lost. AMD has the TDP headroom you say? Perhaps you should check some 2700X reviews again and see the power usage numbers. I've seen a 2700X exceed 140W at full load. 4.0GHz is the tipping point in terms of efficiency for the Ryzen 2000 chips. Any higher and power consumption goes up exponentially. A hypothetical 2800X would needed to be very aggressively binned using the absolute top silicon, to be feasible at say, 4.2GHz all core turbo. And even then, we'll probably see it exceed 150W under full load.