Intel's Future Chips: News, Rumours & Reviews

[-Fran-]

Ugh... In my opinion, that answer is not really clear cut. ESPECIALLY when discussing "longevity", BUT!

Not taking a snarky jab or sarcastic comment here, because I do agree with doing it, but the whole point and argument in favor of testing in 1080p is for longevity purposes, so, since the trend is going wide with both AMD and Intel actually going wide for real this time into the mainstream, I'm pretty damn sure game engines will start going wide noticeably in the short term. Without going too far, there are plenty of titles that do escale pretty damn well at 6 and beyond threads. The only exception to that trend are, obviously, popular DX11 (or prior) based games or game engines (Valve being the weird exception).

There are also reasons for me to advocate in favor of the i7 and not the i5 as well: the performance delta in MP apps (not only games) is going to be even bigger this time, so the extra dosh the i7 will be asking for, might be really worth it this time around at purchase time and not only as a "longevity" play. Again, this is the reason I went for the i7 and not the i5 back in 2012 and it seems I wasn't wrong.

Budget oriented people will now have in the i3 all they need, specially since there's a K version now. So, the i5 is in a very weird position, at least in my eyes, from a value perspective. Specially since the 1600X is actually (according to the leaked data) still competitive in price to performance ratio against the i5.

Cheers!

 

[manleysteele]

how much difference will 12 threads make versus 6core/6t 8600k? i wonder if 6 cores is enough to provide the bulk of the performance with these chips..

I asked a similar question and was told that going from 6c/6t to 6c/12t would be negligible for gaming.
I'm sure there'll be plenty of tests coming to determine that but with Ryzen out now I think that's already been pretty much done.

It depends completely on the games. 32 bit Skyrim can not effectively use more than 6 threads. Assigning more threads usually slows it down. The difference between 4 threads and 6 threads is negligible without higher quality meshes and textures.

64 bit Fallout 4 can use more threads very well. So can 64 bit Skyrim SE. I don't know that anyone on the Nexus has found the optimal thread count for the 64 bit Gamebryo engine. I don't know that they haven't, either, because I haven't been looking.

 

[gamerk316]

Another leaked benchmark shows the $360 six-core CoffeeLake dangerously close to the $500 8-core RyZen even in benchmarks that scale above six-cores

Just for kicks and giggles, as it's been a while since I've done one of these [usual disclaimer, ignoring Turbo and SMT effects]:

Performance = IPC * Clock * Number of Cores

Ryzen 7 1800x:
1576 = IPC * 3.6 * 16
1576 = 57.6 * IPC
IPC = 27.36

i7-8700k:
1521 = IPC * 3.7 * 12
1521 = 44.4 * IPC
IPC = 34.26

% Difference: 22.4%

I obtain 25% difference for your numbers

34.26 / 27.36 = 1.252

But I would do it different. CB15 is a small workload. I would use the all-core turbo frequencies instead the base frequencies. Using the all-core frequencies I obtain 8% difference in IPC. Considering this is a MT workload and RyZen has higher SMT yields, the final result would be ~15% difference in IPC.

% Change != % Difference. The first only applies when you are comparing apples to apples, such as an old value to a new one. In this case, % Difference is the correct formula.

Percentage Difference = |V1-V2| / (V1+V2) / 2 * 100

Typically, I avoid using Turbo frequencies due to differences in how AMD/Intel implement Turbo. It's imperfect, but better then a very high but rarely used max turbo clock giving absurdly low relative IPC numbers.

I also used the MT scores to figure out scaling, since single-core scores you can pretty much eyeball as being relatively small differences. I'll do single-core later, but aren't expecting much.

Scores look reasonable for the i5 at a glance: 50% fewer CPU cores, 2.7% faster clock, 45% slower then the i7. Not sure what's surprising about the result. i5 actually as a slightly higher IPC, which makes sense as it doesn't have to deal with HTT effects.

 

[KirbysHammer]

Another leaked benchmark shows the $360 six-core CoffeeLake dangerously close to the $500 8-core RyZen even in benchmarks that scale above six-cores

Just for kicks and giggles, as it's been a while since I've done one of these [usual disclaimer, ignoring Turbo and SMT effects]:

Performance = IPC * Clock * Number of Cores

Ryzen 7 1800x:
1576 = IPC * 3.6 * 16
1576 = 57.6 * IPC
IPC = 27.36

i7-8700k:
1521 = IPC * 3.7 * 12
1521 = 44.4 * IPC
IPC = 34.26

% Difference: 22.4%

I obtain 25% difference for your numbers

34.26 / 27.36 = 1.252

But I would do it different. CB15 is a small workload. I would use the all-core turbo frequencies instead the base frequencies. Using the all-core frequencies I obtain 8% difference in IPC. Considering this is a MT workload and RyZen has higher SMT yields, the final result would be ~15% difference in IPC.

% Change != % Difference. The first only applies when you are comparing apples to apples, such as an old value to a new one. In this case, % Difference is the correct formula.

Percentage Difference = |V1-V2| / (V1+V2) / 2 * 100

Typically, I avoid using Turbo frequencies due to differences in how AMD/Intel implement Turbo. It's imperfect, but better then a very high but rarely used max turbo clock giving absurdly low relative IPC numbers.

I also used the MT scores to figure out scaling, since single-core scores you can pretty much eyeball as being relatively small differences. I'll do single-core later, but aren't expecting much.

Scores look reasonable for the i5 at a glance: 50% fewer CPU cores, 2.7% faster clock, 45% slower then the i7. Not sure what's surprising about the result. i5 actually as a slightly higher IPC, which makes sense as it doesn't have to deal with HTT effects.

The i5 has 6 cores, not 4.

 

[-Fran-]

The other option no one is considering for the difference between the i5 and i7 is Intel beefing up HT.

Cheers!

 

[aldaia]

Another leaked benchmark shows the $360 six-core CoffeeLake dangerously close to the $500 8-core RyZen even in benchmarks that scale above six-cores

That benchmark also shows the $260 (MSRP) 6-core 8600k getting beat pretty soundly (~50% performance improvement) by the Ryzen 5 1600X, which currently sells for under $240.

Of course, it is an image in a different language with no translation, so the image is missing a lot of context.


EDIT: In other words, AMD may actually still be in legitimate competition with Intel from a price/performance standpoint at some budget ranges. That's good news for everyone. It has been FAR too long since we actually had any chance of legitimate competition in the CPU market.

I guess there is something wrong with the i5. Its score cannot be 75% behind the i7.

It may also be that there is something wrong with the i7. My feeling is i7 was overclocked.

 

[aldaia]

Same review also shows i7-8700K is hot as hell and uses quite a lot of power. In my opinion too much compared with i5.
Temp:

Power:

 

[aldaia]

Another review compares i7-7700K vs i7-8700K and i7-7800X
At stock settings 8700K scores only 1262 at cinebench wich sugests the other review was overclocked.

 

[-Fran-]

Oh, it smells like the i7 8700K might be worth buying in non-K form then. Looks like it's base clocks and turbo are quite high already, unless initial reviews prove a decent OC on top of those figures is feasible.

Also, more heat generated, the more impact will have the toothpaste, I would imagine.

Cheers!

 

[aldaia]

It seems that Coffelake cores are exactly the same Kabylake cores (at least that is what single core performance at same frequency suggests). So basically 8th generation is 7th generation with +2 cores and small diferences in frequency.
In terms of IPC Coffelake = Kabylake = SkyLake. Anyone still doubting that coffe lake is intel's reaction to AMD Ryzen?

 

[goldstone77]

I think we all know this is a direct response to AMD. And I did mention a few months ago, watch how fast Intel will start turning out products, because of AMD. Intel has not failed my expectation! They added 14-18 core products to X299 line within a few months. Dropped prices on 10 cores almost in half. release 8 generation ahead of schedule, and added 2 cores to Kabylake, and are selling them for comparable price to previous generation. Now, let's see how fast they can bring 10nm to market.

 

[juanrga]

1262 isn't the score for the i7-7800k. The real score will be close to 1500. Also expereview is not strictly testing at stock settings, memory is underclocked on CoffeeLake.

CoffeeLake is not a direct response to AMD RyZen. CoffeLake was planned before Zen was tappeout. We have known since 2015 that six cores are coming to mainstream. In fact, the original plan from Intel was to give up to 8-core for mainstream platform, after Skylake

but then the 10nm node was delayed and both 4C Kabylake on 14nm+ and 6C CoffeeLake on 14nm++ were planned as a stopgaps before 10nm was ready.

It is not true that Intel dropped prices on 10 cores almost in half as consequence of RyZen. No. Intel has dropped prices since the ancient 45nm node, by adding moar cores in the same price bracket:

4C Nehalem (i7-975X): $999
6C Westmere (i7-990X): $999
6C Sandy Bridge (i7-3970X): $999
6C Ivy Bridge (i7-4960X): $999
8C Haswell (i7-5960X): $999
8C Broadwell: $1089
10C Skylake: $999

The trend is obvious. So pretending that 10C Skylake is now a direct response to Zen is like pretending that 6C Westmere was a direct response to Zen or that 8C Haswell was a direct response to Zen, because on all the cases Intel added moar cores for the same price. In fact if we count process nodes instead families we have

45nm --> 4-core
32nm --> 6-core
22nm --> 8 core
14nm --> 10-core

From the trend we can expect that 10nm HEDT was 12-core in the same price bracket than 10-core today.

10nm --> 12-core?

Intel always places the new chips in the same price bracket than older chips that are replaced. So Intel selling CoffeeLake at same prices that Kabylake is normal. We knew before leaks that i7-8700k was going to be priced similarly to the i7-7700k, because that is how Intel works. In fact googling a bit you can find me saying exactly this to people as OrangeKrush who was pretending, in another forum, that the new i7-8700k would be priced around $500.

The only direct response to AMD are the SXL-X models above 12C. Initially Intel only planned up to 12C. After ThreadRipper, Intel added the new 14/16/18-core models to the X299 platform. To be fair ThreadRipper wasn't initially planned by AMD either, but was a response to Skylake-X.

 

[goldstone77]

1262 isn't the score for the i7-7800k. The real score will be close to 1500. Also expereview is not strictly testing at stock settings, memory is underclocked on CoffeeLake.

CoffeeLake is not a direct response to AMD RyZen. CoffeLake was planned before Zen was tappeout. We have known since 2015 that six cores are coming to mainstream. In fact, the original plan from Intel was to give up to 8-core for mainstream platform, after Skylake

but then the 10nm node was delayed and both 4C Kabylake on 14nm+ and 6C CoffeeLake on 14nm++ were planned as a stopgaps before 10nm was ready.

It is not true that Intel dropped prices on 10 cores almost in half as consequence of RyZen. No. Intel has dropped prices since the ancient 45nm node, by adding moar cores in the same price bracket:

4C Nehalem (i7-975X): $999
6C Westmere (i7-990X): $999
6C Sandy Bridge (i7-3970X): $999
6C Ivy Bridge (i7-4960X): $999
8C Haswell (i7-5960X): $999
8C Broadwell: $1089
10C Skylake: $999

The trend is obvious. So pretending that 10C Skylake is now a direct response to Zen is like pretending that 6C Westmere was a direct response to Zen or that 8C Haswell was a direct response to Zen, because on all the cases Intel added moar cores for the same price. In fact if we count process nodes instead families we have

45nm --> 4-core
32nm --> 6-core
22nm --> 8 core
14nm --> 10-core

From the trend we can expect that 10nm HEDT was 12-core in the same price bracket than 10-core today.

10nm --> 12-core?

Intel always places the new chips in the same price bracket than older chips that are replaced. So Intel selling CoffeeLake at same prices that Kabylake is normal. We knew before leaks that i7-8700k was going to be priced similarly to the i7-7700k, because that is how Intel works. In fact googling a bit you can find me saying exactly this to people as OrangeKrush who was pretending, in another forum, that the new i7-8700k would be priced around $500.

The only direct response to AMD are the SXL-X models above 12C. Initially Intel only planned up to 12C. After ThreadRipper, Intel added the new 14/16/18-core models to the X299 platform. To be fair ThreadRipper wasn't initially planned by AMD either, but was a response to Skylake-X.

https://ark.intel.com/products/94456/Intel-Core-i7-6950X-Processor-Extreme-Edition-25M-Cache-up-to-3_50-GHz
Processor Number i7-6950X Recommended Customer Price $1723.00 - $1743.00
Launch Date Q2'16
Performance
# of Cores
10
# of Threads
20
Intel Core i9-7900X Processor Price: $962.89 Free Shipping for Prime Members which is selling below retail now.
I think there is an argument that can be made this is because of AMD. Agree to disagree.

Intel already had an HEDT platform, so ThreadRipper, if we go by the story AMD tells us it was to establish a presence in the HEDT market, and give enthusiast uncompromised choice and value. But I would believe it attempt to have products in every segment for consumer markets. You can say everything Ryzen/ThreadRipper is a response to Intel Processors as a whole since they haven't been able to compete with Intel in years.

 

[8350rocks]

nobody plays games on HEDT @ 720p though...

The 720p testing has a different reason. The reason has been explained a dozen of times. Ignoring the reason will not stop reviews from performing 720p and 1080p tests.

It does have a different reason, to create an artificial benchmark that proves nothing. performance there does not accurately indicate performance in anything.

Believe it or not some people play 1080p at high refresh rates. Mind blown.

Testing in 720p allows you to see the framerate at which the CPU can push to at maximum. Let's say in 1 year 200hz@1080p monitors become standard for 1080p. People upgrade GPUs more often than CPUs. While Ryzen mght only be able to push 110fps, An Intel chip might push 160fps or more.

This is a fallacy.

720p does not create a situation where you can learn any useful real world performance. Show me one example where 720p performance is indicative of a real world performance result in anything.

1080p is also ridiculously lauded as the best resolution for gaming, the only people who say that either want a big number in the corner of their monitor (because they cannot truthfully tell the difference between 90 FPS and 190 FPS, despite what they claim), or people who have never played on 4K or ultrawide.

1080p looks great if you came from a laptop. If you have 4K and see the clarity and fidelity routinely, 1080p looks like an etch-a-sketch version of what you could be seeing.

There is nothing useful coming out of 720p benchmarks, and as 1440p and greater become more ubiquitous, then 1080p will become the new 720p benchmark creating an artificial scenario that does not justifiably show performance that has any meaning.

Testing at a lower resolution allows you to remove the GPU from the equation (you'd hope) so that you are getting the max FPS the CPU can push which should be well below the max FPS the GPU could push at that resolution. So for CPU testing a low resolution is used. Most use 1080 as their test resolution though.

E-sports gamers play at 1080 because yes, they want max FPS and they don't care about all the distracting eye candy.

It's more then that, it's about knowing if your CPU becomes a problem when paired with future GPU upgrades. Its about performance headroom.

Yes, at 1440p and GPU X, two CPUs might be equal. But what about a more powerful GPU? Can each CPU drive it? How much performance headroom is on the table? That's the point of low resolution tests. It's not about "real world, today", it's about "real world, tomorrow".

Ryzen can push 4x Vega FE cards in mGPU config for gaming:

https://www.reddit.com/r/Amd/comments/72lrp3/my_1950x_quad_vega_fe_workstation_build_is/

Something tells me that if the CPU can handle 100 TFLOPs of GPU, what it does at 720p means nothing.

 

[juanrga]

1262 isn't the score for the i7-7800k. The real score will be close to 1500. Also expereview is not strictly testing at stock settings, memory is underclocked on CoffeeLake.

CoffeeLake is not a direct response to AMD RyZen. CoffeLake was planned before Zen was tappeout. We have known since 2015 that six cores are coming to mainstream. In fact, the original plan from Intel was to give up to 8-core for mainstream platform, after Skylake

but then the 10nm node was delayed and both 4C Kabylake on 14nm+ and 6C CoffeeLake on 14nm++ were planned as a stopgaps before 10nm was ready.

It is not true that Intel dropped prices on 10 cores almost in half as consequence of RyZen. No. Intel has dropped prices since the ancient 45nm node, by adding moar cores in the same price bracket:

4C Nehalem (i7-975X): $999
6C Westmere (i7-990X): $999
6C Sandy Bridge (i7-3970X): $999
6C Ivy Bridge (i7-4960X): $999
8C Haswell (i7-5960X): $999
8C Broadwell: $1089
10C Skylake: $999

The trend is obvious. So pretending that 10C Skylake is now a direct response to Zen is like pretending that 6C Westmere was a direct response to Zen or that 8C Haswell was a direct response to Zen, because on all the cases Intel added moar cores for the same price. In fact if we count process nodes instead families we have

45nm --> 4-core
32nm --> 6-core
22nm --> 8 core
14nm --> 10-core

From the trend we can expect that 10nm HEDT was 12-core in the same price bracket than 10-core today.

10nm --> 12-core?

Intel always places the new chips in the same price bracket than older chips that are replaced. So Intel selling CoffeeLake at same prices that Kabylake is normal. We knew before leaks that i7-8700k was going to be priced similarly to the i7-7700k, because that is how Intel works. In fact googling a bit you can find me saying exactly this to people as OrangeKrush who was pretending, in another forum, that the new i7-8700k would be priced around $500.

The only direct response to AMD are the SXL-X models above 12C. Initially Intel only planned up to 12C. After ThreadRipper, Intel added the new 14/16/18-core models to the X299 platform. To be fair ThreadRipper wasn't initially planned by AMD either, but was a response to Skylake-X.

https://ark.intel.com/products/94456/Intel-Core-i7-6950X-Processor-Extreme-Edition-25M-Cache-up-to-3_50-GHz
Processor Number i7-6950X Recommended Customer Price $1723.00 - $1743.00
Launch Date Q2'16
Performance
# of Cores
10
# of Threads
20
Intel Core i9-7900X Processor Price: $962.89 Free Shipping for Prime Members which is selling below retail now.
I think there is an argument that can be made this is because of AMD. Agree to disagree.

Intel already had an HEDT platform, so ThreadRipper, if we go by the story AMD tells us it was to establish a presence in the HEDT market, and give enthusiast compromised choice and value. But I would believe it attempt to have products in every segment for consumer markets. You can say everything Ryzen/ThreadRipper is a response to Intel Processors as a whole since they haven't been able to compete with Intel in years.

As shown above, Intel have been adding moar cores in the same price bracket since the ancient 45nm. Thus getting now 10-core Skylake for the same price than former 8-core Broadwell is not due to Zen. Just as getting 8-core Wetmere for the same price than former 4-core Nehalen is not due to Zen. Evidently giving moar cores for the same price is the same than reducing the price of each core. Intel has been reducing the price of each core since the 45nm node: $249 (Nehalem) --> $166 (Westmere, Sandy Bridge, Ivy Bridge) --> $124 (Haswell, Broadwell) --> $99.9 (Skylake).

Pretending that Intel has reduced the price per core on Skylake due to Zen is so incorrect as pretending that all the former reductions on price are due to Zen.

 

[ibjeepr]

nobody plays games on HEDT @ 720p though...

The 720p testing has a different reason. The reason has been explained a dozen of times. Ignoring the reason will not stop reviews from performing 720p and 1080p tests.

It does have a different reason, to create an artificial benchmark that proves nothing. performance there does not accurately indicate performance in anything.

Believe it or not some people play 1080p at high refresh rates. Mind blown.

Testing in 720p allows you to see the framerate at which the CPU can push to at maximum. Let's say in 1 year 200hz@1080p monitors become standard for 1080p. People upgrade GPUs more often than CPUs. While Ryzen mght only be able to push 110fps, An Intel chip might push 160fps or more.

This is a fallacy.

720p does not create a situation where you can learn any useful real world performance. Show me one example where 720p performance is indicative of a real world performance result in anything.

1080p is also ridiculously lauded as the best resolution for gaming, the only people who say that either want a big number in the corner of their monitor (because they cannot truthfully tell the difference between 90 FPS and 190 FPS, despite what they claim), or people who have never played on 4K or ultrawide.

1080p looks great if you came from a laptop. If you have 4K and see the clarity and fidelity routinely, 1080p looks like an etch-a-sketch version of what you could be seeing.

There is nothing useful coming out of 720p benchmarks, and as 1440p and greater become more ubiquitous, then 1080p will become the new 720p benchmark creating an artificial scenario that does not justifiably show performance that has any meaning.

Testing at a lower resolution allows you to remove the GPU from the equation (you'd hope) so that you are getting the max FPS the CPU can push which should be well below the max FPS the GPU could push at that resolution. So for CPU testing a low resolution is used. Most use 1080 as their test resolution though.

E-sports gamers play at 1080 because yes, they want max FPS and they don't care about all the distracting eye candy.

It's more then that, it's about knowing if your CPU becomes a problem when paired with future GPU upgrades. Its about performance headroom.

Yes, at 1440p and GPU X, two CPUs might be equal. But what about a more powerful GPU? Can each CPU drive it? How much performance headroom is on the table? That's the point of low resolution tests. It's not about "real world, today", it's about "real world, tomorrow".

Ryzen can push 4x Vega FE cards in mGPU config for gaming:

https://www.reddit.com/r/Amd/comments/72lrp3/my_1950x_quad_vega_fe_workstation_build_is/

Something tells me that if the CPU can handle 100 TFLOPs of GPU, what it does at 720p means nothing.

I tell you it's to remove the GPU from the equation and you come back with 4 of them...?

If you want to compare which CPU is handling a program better, you need to isolate the test variable.
CPU X vs CPU Y with out the impact of the GPU. At high resolution you start seeing the video card's limits instead of the CPU's.

 

[goldstone77]

1262 isn't the score for the i7-7800k. The real score will be close to 1500. Also expereview is not strictly testing at stock settings, memory is underclocked on CoffeeLake.

CoffeeLake is not a direct response to AMD RyZen. CoffeLake was planned before Zen was tappeout. We have known since 2015 that six cores are coming to mainstream. In fact, the original plan from Intel was to give up to 8-core for mainstream platform, after Skylake

but then the 10nm node was delayed and both 4C Kabylake on 14nm+ and 6C CoffeeLake on 14nm++ were planned as a stopgaps before 10nm was ready.

It is not true that Intel dropped prices on 10 cores almost in half as consequence of RyZen. No. Intel has dropped prices since the ancient 45nm node, by adding moar cores in the same price bracket:

4C Nehalem (i7-975X): $999
6C Westmere (i7-990X): $999
6C Sandy Bridge (i7-3970X): $999
6C Ivy Bridge (i7-4960X): $999
8C Haswell (i7-5960X): $999
8C Broadwell: $1089
10C Skylake: $999

The trend is obvious. So pretending that 10C Skylake is now a direct response to Zen is like pretending that 6C Westmere was a direct response to Zen or that 8C Haswell was a direct response to Zen, because on all the cases Intel added moar cores for the same price. In fact if we count process nodes instead families we have

45nm --> 4-core
32nm --> 6-core
22nm --> 8 core
14nm --> 10-core

From the trend we can expect that 10nm HEDT was 12-core in the same price bracket than 10-core today.

10nm --> 12-core?

Intel always places the new chips in the same price bracket than older chips that are replaced. So Intel selling CoffeeLake at same prices that Kabylake is normal. We knew before leaks that i7-8700k was going to be priced similarly to the i7-7700k, because that is how Intel works. In fact googling a bit you can find me saying exactly this to people as OrangeKrush who was pretending, in another forum, that the new i7-8700k would be priced around $500.

The only direct response to AMD are the SXL-X models above 12C. Initially Intel only planned up to 12C. After ThreadRipper, Intel added the new 14/16/18-core models to the X299 platform. To be fair ThreadRipper wasn't initially planned by AMD either, but was a response to Skylake-X.

https://ark.intel.com/products/94456/Intel-Core-i7-6950X-Processor-Extreme-Edition-25M-Cache-up-to-3_50-GHz
Processor Number i7-6950X Recommended Customer Price $1723.00 - $1743.00
Launch Date Q2'16
Performance
# of Cores
10
# of Threads
20
Intel Core i9-7900X Processor Price: $962.89 Free Shipping for Prime Members which is selling below retail now.
I think there is an argument that can be made this is because of AMD. Agree to disagree.

Intel already had an HEDT platform, so ThreadRipper, if we go by the story AMD tells us it was to establish a presence in the HEDT market, and give enthusiast compromised choice and value. But I would believe it attempt to have products in every segment for consumer markets. You can say everything Ryzen/ThreadRipper is a response to Intel Processors as a whole since they haven't been able to compete with Intel in years.

As shown above, Intel have been adding moar cores in the same price bracket since the ancient 45nm. Thus getting now 10-core Skylake for the same price than former 8-core Broadwell is not due to Zen. Just as getting 8-core Wetmere for the same price than former 4-core Nehalen is not due to Zen. Evidently giving moar cores for the same price is the same than reducing the price of each core. Intel has been reducing the price of each core since the 45nm node: $249 (Nehalem) --> $166 (Westmere, Sandy Bridge, Ivy Bridge) --> $124 (Haswell, Broadwell) --> $99.9 (Skylake).

Pretending that Intel has reduced the price per core on Skylake due to Zen is so incorrect as pretending that all the former reductions on price are due to Zen.

I do see what you are saying, so now we get 16/32 for the same price as last years 10 core!

Processor Number i7-6950X Recommended Customer Price $1723.00 - $1743.00
So, 16/32 for a cheaper price than last years 10 core $1699 is a direct response to ThreadRipper breaking that trend by more than 50% cores for a cheaper price!

 

[-Fran-]

Why are you guys talking about the HEDT platform when the "news" is around the *mainstream* platform getting more than 4 cores from the Intel camp. And, around 2015, AMD announced Zen to be an 8 core part? Am I remembering this incorrectly?

Cheers!

 

[goldstone77]

Why are you guys talking about the HEDT platform when the "news" is around the *mainstream* platform getting more than 4 cores from the Intel camp. And, around 2015, AMD announced Zen to be an 8 core part? Am I remembering this incorrectly?

Cheers!

New AMD Zen X86 APU boasts up to 16 cores
by FUAD ABAZOVIC on31 MARCH 2015

 

[juanrga]

Why are you guys talking about the HEDT platform when the "news" is around the *mainstream* platform getting more than 4 cores from the Intel camp. And, around 2015, AMD announced Zen to be an 8 core part? Am I remembering this incorrectly?

8-core Zen CPUs aren't mainstream parts. The mainstream parts are Raven Ridge APUs.

Regarding Intel, the ancient Westemere mainstream 'APU' was 2-core and SandyBridge introduced the first 4-core. Originally, Skylake (14nm) was going to be the last 4-core 'APU', and CannonLake (10nm) would double the number of cores up to 8. But then Intel did start having problems with 14nm and 10nm nodes and the whole plan was changed. The Tick-tock strategy was replaced by tick-tock-optimization-2nm-optimization and we got first 4-core Kabylake and now 6-core CoffeLake.

Once the 10nm problems are solved, we will get 8-core on mainstream platform. Last rumors is that this is coming next year with IceLake.

On the other hand, it is AMD which doesn't advance with its core-counts on the mainstream platforms.

"Llano" (2011) --> 4-core
"Trinity" (2012) --> 4-core
"Richland" (2013) --> 4-core
"Kabini" (2014, SoC) --> 4-core
"Kaveri" (2014) --> 4-core
"Carrizo" (2016) --> 4-core
"Bristol Ridge" (2017) --> 4-core
"Raven Ridge" (2018) --> 4-core
"Picasso" (2019) --> 4-core

New AMD Zen X86 APU boasts up to 16 cores
by FUAD ABAZOVIC on31 MARCH 2015

So, AMD did hear that Intel planned 8-core 'APU' for mainstrean and planned a 16-core!!!

Well no, I was joking just to demonstrate how easy is to play this game about who is influencing. The true is this 16-core APU is for HPC, would use the SP3 socket (EPYC) or some variation, but doesn't appear in any official roadmap.

 

[adamsleath]

...both amd and intel have been competing in the field for decades....they dont operate in a market vacuum.
ive even heard someone use the expression "oh yeah amd is like a clone of an intel chip" (many years ago) but that's what someone said to me about it. this is when i hadnt even heard of amd :lol: such is/was the intel presence in the market (in australia anyway)...but i guess i was just the ignoramus that hadn't heard of amd.

ryzen 3/5/7 arent mainstream? could have fooled me. they sure have mainstream pricing.

 

[-Fran-]

Why are you guys talking about the HEDT platform when the "news" is around the *mainstream* platform getting more than 4 cores from the Intel camp. And, around 2015, AMD announced Zen to be an 8 core part? Am I remembering this incorrectly?

8-core Zen CPUs aren't mainstream parts. The mainstream parts are Raven Ridge APUs.

Regarding Intel, the ancient Westemere mainstream 'APU' was 2-core and SandyBridge introduced the first 4-core. Originally, Skylake (14nm) was going to be the last 4-core 'APU', and CannonLake (10nm) would double the number of cores up to 8. But then Intel did start having problems with 14nm and 10nm nodes and the whole plan was changed. The Tick-tock strategy was replaced by tick-tock-optimization-2nm-optimization and we got first 4-core Kabylake and now 6-core CoffeLake.

Once the 10nm problems are solved, we will get 8-core on mainstream platform. Last rumors is that this is coming next year with IceLake.

On the other hand, it is AMD which doesn't advance with its core-counts on the mainstream platforms.

"Llano" (2011) --> 4-core
"Trinity" (2012) --> 4-core
"Richland" (2013) --> 4-core
"Kabini" (2014, SoC) --> 4-core
"Kaveri" (2014) --> 4-core
"Carrizo" (2016) --> 4-core
"Bristol Ridge" (2017) --> 4-core
"Raven Ridge" (2018) --> 4-core
"Picasso" (2019) --> 4-core

New AMD Zen X86 APU boasts up to 16 cores
by FUAD ABAZOVIC on31 MARCH 2015

So, AMD did hear that Intel planned 8-core 'APU' for mainstrean and planned a 16-core!!!

Well no, I was joking just to demonstrate how easy is to play this game about who is influencing. The true is this 16-core APU is for HPC, would use the SP3 socket (EPYC) or some variation, but doesn't appear in any official roadmap.

I'd buy that argument, if they weren't used in the same platform. And, more importantly, the same chipset and socket combo.

I know you'll try arguing about the "CPU" itself not being shown as "mainstream" in the slides, but it's the same thing you said not long ago about the Ice Lake rumor about the i7 with 8C for the Z platform not being a "mainstream" part. That is semantics and I disagree with that take.

Cheers!

EDIT: Changed word.

 

[juanrga]

...both amd and intel have been competing in the field for decades....they dont operate in a market vacuum.
ive even heard someone use the expression "oh yeah amd is like a clone of an intel chip" (many years ago) but that's what someone said to me about it. this is when i hadnt even heard of amd :lol: such is/was the intel presence in the market (in australia anyway)...but i guess i was just the ignoramus that hadn't heard of amd.

ryzen 3/5/7 arent mainstream? could have fooled me. they sure have mainstream pricing.

I just provided AMD roadmap. AMD claims that the mainstream platform is Raven Rdige APUs. The RyZen CPUs aren't mainstream products. It is also the reason why the impact of RyZen sales in AMD marketshare is is in the 4%. Those chips aren't masintream chips and don't produce real volume sales.


Pricing is not conclusive. The $499 R7 1800X is not a mainstream chip. Or didn't just forget all the slides and demos, where AMD compared the 1800X with the HEDT Broadwell-E 8-core chip?

After CoffeLake launch AMD will be forced to drop the price of the 1800X to sub $300 levels, but it will continue being not a mainstream chip.

 

[logainofhades]

Threadripper is HEDT. R7 are high performance mainstream, the same as your i7 7700k, and soon to be released 8700k. The platform is what makes it HEDT, not the CPU.

 

[Isaiah4110]

Threadripper is HEDT. R7 are high performance mainstream, the same as your i7 7700k, and soon to be released 8700k. The platform is what makes it HEDT, not the CPU.

+1