Coffeelake thread, benchmarks, reviews, input, everything.

[eddman]

My point is that many workloads aren't like this. An example where the opposite is true would be encoding. ... a high core count CPU, should scaling work, excel in the latter. However this detail is apparently lost by many because they only tend to focus on Cinebench/POV-ray type benchmarks.

Well, 7800X does quite well in 4K and/or HQ handbrake tests which seem to benefit from more cores, and 8700K should do better.

https://www.anandtech.com/show/1155...core-i9-7900x-i7-7820x-and-i7-7800x-tested/13
https://www.techspot.com/review/1433-intel-core-i9-core-i7-skylake-x/page2.html
http://www.bit-tech.net/reviews/tech/cpus/intel-core-i7-7800x-review/3/

EDIT: It seems two programs that do not benefit from higher clocks but do with more cores are Corona, and Blender in certain workloads. Tom's hardware's multiple blender results for 8700K should be interesting.

 

[Zucker2k]

What? You did exactly the opposite when you claimed this:

So what is it? Clock vs clock - is it right or wrong? Make up your mind. I remember it was you who claimed that it's irrelevant for something like Bulldozer because that has a high-clocking architecture

How so? Nothing said there contradicts my stand on the clock for clock question. If IPC = Instruction Per Clock/Cycle, I'm saying don't stop there; the more important question is how many cycles each chip is capable of per a given time (frequency). Frequency is an inherent, and a direct influence on ipc design. Basically, what this means is that chip architects must marry an architecture to a process whose characteristsics (high frequency or low power, but a balance is inherently more desirable) suits the ipc potency of the architecture. This is why process tuning is so crucial to realizing the best results. Eg. you can't just slap the skylake architecture on Intel's 90nm process, for what should be obvious reasons; die size, power consumption, heat, etc.
So, from this point of view, it is simplistic, at best, for an enthusiast to pit Nehalem against Bulldozer in an ipc battle In the greater scheme of things it means nothing, because it only tells part of the story. The whole story, is if the enthusiast allows the chip to stretch its legs at its thermal design power and cooling, incorporating whatever performance enhancements the manufacturer deems fit. If this is so hard for you to understand, then instead of a clock for clock comparison at 3.4GHz, let's try raising the clock a bit; to a skylake-comfortable 4.5GHz, on air.

 

[mikk]

Apparently you cannot comprehend how turbo modes work, even when somebody posted them on this same page!


Now compare with Ryzen 3 specs.

The gaming IPC of Ryzen is comparable to Ivy Bridge in many games, but sometimes even worse. Another recent example from Project Cars 2: http://gamegpu.com/racing-simulators-/-гонки/project-cars-2-test-gpu-cpu

i5-2500K 3.4 Ghz (4C Turbo) 70 FPS
Ryzen 1300X 3.6 Ghz (4C Turbo) 70 FPS

Gaming IPC below Sandy Bridge! And that's with a huge memory disadvantage for the Sandy Bridge system (DDR3-1333 vs DDR4-2667)

 

[tamz_msc]

The gaming IPC of Ryzen is comparable to Ivy Bridge in many games, but sometimes even worse. Another recent example from Project Cars 2: http://gamegpu.com/racing-simulators-/-гонки/project-cars-2-test-gpu-cpu

i5-2500K 3.4 Ghz (4C Turbo) 70 FPS
Ryzen 1300X 3.6 Ghz (4C Turbo) 70 FPS

Gaming IPC below Sandy Bridge! And that's with a huge memory disadvantage for the Sandy Bridge system (DDR3-1333 vs DDR4-2667)

You sound like *THAT* guy. What did the Rise of The Tomb Raider patch tell us about software optimizations?

 

[tamz_msc]

How so? Nothing said there contradicts my stand on the clock for clock question. If IPC = Instruction Per Clock/Cycle, I'm saying don't stop there; the more important question is how many cycles each chip is capable of per a given time (frequency). Frequency is an inherent, and a direct influence on ipc design. Basically, what this means is that chip architects must marry an architecture to a process whose characteristsics (high frequency or low power, but a balance is inherently more desirable) suits the ipc potency of the architecture. This is why process tuning is so crucial to realizing the best results. Eg. you can't just slap the skylake architecture on Intel's 90nm process, for what should be obvious reasons; die size, power consumption, heat, etc.
So, from this point of view, it is simplistic, at best, for an enthusiast to pit Nehalem against Bulldozer in an ipc battle In the greater scheme of things it means nothing, because it only tells part of the story. The whole story, is if the enthusiast allows the chip to stretch its legs at its thermal design power and cooling, incorporating whatever performance enhancements the manufacturer deems fit. If this is so hard for you to understand, then instead of a clock for clock comparison at 3.4GHz, let's try raising the clock a bit; to a skylake-comfortable 4.5GHz, on air.

You're confused.

You're beating around a bush that's full of incoherent ideas regarding Instructions Per clock, Instructions per cycle(the two are not the same), latency per (specific) instruction, process limitation, high FO4 design(Bulldozer) vs low FO4 design, architectural differences, a rigid view that IPC is supposedly dependent on the clock speed and so forth.

Why should an IPC comparison at 3.5GHz be any different at 4.5GHz? There are Lake chips that operate at both the 3.5 GHz mark and the 4.5GHz mark.

 

[RichUK]

https://www.eteknix.com/asus-introduces-intel-z370-motherboard-lineup/

Asus Z370 mobos.

Apex goodness:

 

[psolord]

The gaming IPC of Ryzen is comparable to Ivy Bridge in many games, but sometimes even worse. Another recent example from Project Cars 2: http://gamegpu.com/racing-simulators-/-гонки/project-cars-2-test-gpu-cpu

i5-2500K 3.4 Ghz (4C Turbo) 70 FPS
Ryzen 1300X 3.6 Ghz (4C Turbo) 70 FPS

Gaming IPC below Sandy Bridge! And that's with a huge memory disadvantage for the Sandy Bridge system (DDR3-1333 vs DDR4-2667)

Well at least AMD can now do with 4/4 what it could do with 8/8 before. That is no small feat.

Also the bigger Ryzen brothers are amongst the top cpus and are able to almost max out a 1080ti.

Most Ryzens can also max out a GTX 1080, which is also great.

It seems to me that the glass is half full, not half empty. No scrap that, it's nearly full.

 

[psolord]

More on topic....is it me, or are most motherboard manufacturers hell bent to offer less sata ports for the z370 mobos compared to the past chipsets?

 

[tamz_msc]

More on topic....is it me, or are most motherboard manufacturers hell bent to offer less sata ports for the z370 mobos compared to the past chipsets?

If you're referencing the Apex motherboard from Asus, then they've always aimed this at extreme OCers and skimped on I/O. I expect most to offer 8, just like before.

 

[phillyman36]

More on topic....is it me, or are most motherboard manufacturers hell bent to offer less sata ports for the z370 mobos compared to the past chipsets?

Yes seems like everyone is offering less USB and Sata.

 

[RichUK]

I only use one SATA drive. The rest of my storage is on an external USB 3 SSD.

My main interest in the Apex board is maxing memory performance as I've got 2x8GB 4266 c19 sticks, and my Z270 Code struggles to get them over 4000.

Plus it doesn't have all that aesthetic plastic guard rubbish over the board.

 

[psolord]

If you're referencing the Apex motherboard from Asus, then they've always aimed this at extreme OCers and skimped on I/O. I expect most to offer 8, just like before.

No, not just the APEX. I mean it's the general feeling I'm getting. Maybe I'm affected by the Taichi Z370 losing 2 sata ports, but generally I don't see manufacturers going large on the Sata ports. If I am not mistaken I have not seen any of them having 10 Sata. 8 is the max and 6 the most common option.

 

[AtenRa]

The gaming IPC of Ryzen is comparable to Ivy Bridge in many games, but sometimes even worse. Another recent example from Project Cars 2: http://gamegpu.com/racing-simulators-/-гонки/project-cars-2-test-gpu-cpu

i5-2500K 3.4 Ghz (4C Turbo) 70 FPS
Ryzen 1300X 3.6 Ghz (4C Turbo) 70 FPS

Gaming IPC below Sandy Bridge! And that's with a huge memory disadvantage for the Sandy Bridge system (DDR3-1333 vs DDR4-2667)

Or its the game that is not optimized as it should,

http://gamegpu.com/rts-/-стратегии/total-war-warhammer-ii-test-gpu-cpu

Core i5 4670K 3.6GHz 4C Turbo = 65/86fps
Ryzen R3 1200 3.1GHz 4C Turbo = 64/83fps

 

[hnizdo]

You sound like *THAT* guy. What did the Rise of The Tomb Raider patch tell us about software optimizations?

It did tell us specifically: 1] without optimalization ryzen performance can be very poor not only in games, 2] only about two games out of hudred got optimalizations

 

[mikk]

Or its the game that is not optimized as it should,

Typical AMD excuse.

 

[Glo.]

Pulling back this thread into its rails:

So, only K models have seen price hike, not the regular SKUs? So 8400 CPU should have the same price as 7400? 8100 should have the same price as 7100?

 

[Zucker2k]

You're confused.

You're beating around a bush that's full of incoherent ideas regarding Instructions Per clock, Instructions per cycle(the two are not the same), latency per (specific) instruction, process limitation, high FO4 design(Bulldozer) vs low FO4 design, architectural differences, a rigid view that IPC is supposedly dependent on the clock speed and so forth.

Why should an IPC comparison at 3.5GHz be any different at 4.5GHz? There are Lake chips that operate at both the 3.5 GHz mark and the 4.5GHz mark.

(1) Aha! If the two are not the same, how do you then pretend they are by testing @ clock for clock? Core chips scale very well by design. 4.5GHz because it is impossible for the AMD chip to do 4.5Ghz on air. Now, ask yourself how that may have influenced the ipc design of the zen architecture - a design that bottoms out after 4Ghz. Now, IPC aside, you're also crying over ILP and core scaling, ie. CB. So what exactly do you want from those tests since you're pro 'clock for clock'? A badly optimized code that hinders the true IPC potential of Intel chips?

 

[DrMrLordX]

Core chips scale very well by design. ?

All chips scale well by design, so long as the cache and memory subsystems can keep up. There is nothing about <insertnameofuarchweshouldn'tbediscussinginthisthread> that makes it unable to scale well beyond a certain clockspeed. You are confusing uarch and process limitations.

IPC comparisons should be valid at any clockspeed, again assuming the cache and memory subsystems can keep up. Though you can do IPC comparisons at different fixed clockspeeds if it worries you that much. 3 GHz with fast RAM is usually a good place to start. Though all turbo modes and similar should be deactivated for such testing.

 

[tamz_msc]

(1) Aha! If the two are not the same, how do you then pretend they are by testing @ clock for clock? Core chips scale very well by design. 4.5GHz because it is impossible for the AMD chip to do 4.5Ghz on air. Now, ask yourself how that may have influenced the ipc design of the zen architecture - a design that bottoms out after 4Ghz. Now, IPC aside, you're also crying over ILP and core scaling, ie. CB. So what exactly do you want from those tests since you're pro 'clock for clock'? A badly optimized code that hinders the true IPC potential of Intel chips?

Have you ever looked at a CPU instruction fetch/decode timing diagram? Because if you did, and understood what happens during that process, you would immediately take back what you've said.

 

[Zucker2k]

All chips scale well by design, so long as the cache and memory subsystems can keep up. There is nothing about <insertnameofuarchweshouldn'tbediscussinginthisthread> that makes it unable to scale well beyond a certain clockspeed. You are confusing uarch and process limitations.

IPC comparisons should be valid at any clockspeed, again assuming the cache and memory subsystems can keep up. Though you can do IPC comparisons at different fixed clockspeeds if it worries you that much. 3 GHz with fast RAM is usually a good place to start. Though all turbo modes and similar should be deactivated for such testing.

I'm not. I purposely brought that up to show the importance of process to the overall performance of a uarch. If you're a chip architect who's designing a uarch for a process that could scale to 5GHz with 40% more power than another process at 4GHz, how do you try to bridge the performance gap for the 4GHz chip if you have 40% more power at your disposal. Do you increase workload per cycle, hence stronger ipc? So if you were to compare the two chips at 4Ghz, who comes up tops? I know a test like that is useful to some, but for me it's meaningless. I don't know why I'm being called out on it.

Edited for clarity.

 

[TahoeDust]

Why is it that every Intel thread turns into an Intel vs AMD crap fest, but AMD threads hardly ever do?

 

[TheF34RChannel]

Why is it that every Intel thread turns into an Intel vs AMD crap fest, but AMD threads hardly ever do?

Because we lack the obsession to downplay something or force what we like on others. Good threads get ruined. This is not baiting, merely my observations across the web.

 

[tamz_msc]

Why is it that every Intel thread turns into an Intel vs AMD crap fest, but AMD threads hardly ever do?

Because there's always that one person who comes along with a cherry-picked/skewed/single data point for exaggeration and doesn't listen when told to look at averages.

It's mostly the same people who complain about every Intel thread turning into a slugfest but have little to no interest in threads like the server CPU benchmark thread or the technically insightful Stilt's thread.

 

[Hayateazekura]

Because there's always that one person who comes along with a cherry-picked/skewed/single data point for exaggeration and doesn't listen when told to look at averages.

It's mostly the same people who complain about every Intel thread turning into a slugfest but have little to no interest in threads like the server CPU benchmark thread or the technically insightful Stilt's thread.

No its because people like you. You feed them. Instead say "Hey this is a coffeelake thread stay on topic." Instead you start debating about amd vs intel. I know you love Ryzen and what not. But you are part of the problem.

 

[tamz_msc]

No its because moderators like you. You feed them. If you were a proper moderator you would just say "Hey this is a coffeelake thread stay on topic." Instead you start debating about amd vs intel. I know you love Ryzen and what not. But you are part of the problem.

You've got me confused with someone else, and that moderator, who actually created this thread, was not doing anything wrong in trying to explain how a power consumption result was an outlier.

It's always because someone who claims things like "25 percent more ST performance, Ivy Bridge IPC, Intel 6 cores better than AMD 8 cores because reasons" etc. that the thread turns into a slugfest because God forbid they can't stand it when someone attempts to counter them with meaningful data.