Coffee Lake scarce in Scandinavia, no K-models until 2018

[MageTank]

Just now, System Error Message said:

how do you think AMD beats intel at blender and cinebench?

Why do people think intel has higher IPCs than AMD? AMD beat intel despite being lower clocked and having less cores, that was actually in their demo. They beat intel in a math heavy workload. When it comes to simpler stuff and some of the instructions, intel is faster.

Blender was shown to be using AMD optimized settings. Did you conveniently forget that? Cinebench, AMD comes nowhere near Intel in single thread, while they only beat Intel at multi-thread by having higher core counts. This is plain as day when looking at Intel's X99 or X299 SKU's at identical clocks. What in the world are you even on about? lol

[Unhelpful]

3 hours ago, System Error Message said:

Thats really simple. That means clock per clock they are equal. lets take their total frequency value.

6 cores at 5Ghz is 30Ghz total

8 cores at 4Ghz is 32Ghz total.

That's not how that works. If all things are equal (IPC and cache) and you have a program perfectly optimized for multithreaded workloads... Then you can almost apply your brain-dead oversimplification. But those things are not true.

[System Error Message]

6 minutes ago, MageTank said:

Blender was shown to be using AMD optimized settings. Did you conveniently forget that? Cinebench, AMD comes nowhere near Intel in single thread, while they only beat Intel at multi-thread by having higher core counts. This is plain as day when looking at Intel's X99 or X299 SKU's at identical clocks. What in the world are you even on about? lol

that AMD has significantly changed their architecture when they came out with ryzen. Infact i went through some of the lower details.

The point is intel comes out with coffee lake to beat ryzen only to be beaten by the next AMD CPU.

 

This video will help. If a company was in intel's position and innovated rather than do what intel does AMD would already be dead. Infact intel and AMD actually took a few lessons from cyrix that had produced the x86 much faster than them because of doing things differently.

 

This is why despite intel being faster, its not like the coffee lake is that much faster to offset the price difference. And we still havent seen optimisations for the infinity fabric yet.

 

I dont want to see the repeats of what happens. In europe even UK things are sold for a lot more than they're worth that its actually cheaper to import from US and pay import duty and VAT than it is to buy locally.

 

The point of the video is to say no to overpricing. If intel can release their CPU competitively in price to AMD than it would be worth to consider but they still are doing things in their old way.

[Misanthrope]

19 minutes ago, LAwLz said:

I think it's cool how everyone goes around saying that Ryzen 7 is significantly cheaper, yet I am fairly sure there have been no prices announced.

What the hell are you talking about, officially announced prices from intel:

 

 

https://arstechnica.com/gadgets/2017/09/intel-coffee-lake-price-specs-release-date/

 

Regional pricing might not be available but for the most part we know.

[MageTank]

Just now, System Error Message said:

that AMD has significantly changed their architecture when they came out with ryzen. Infact i went through some of the lower details.

The point is intel comes out with coffee lake to beat ryzen only to be beaten by the next AMD CPU.

 

This video will help. If a company was in intel's position and innovated rather than do what intel does AMD would already be dead. Infact intel and AMD actually took a few lessons from cyrix that had produced the x86 much faster than them because of doing things differently.

 

This is why despite intel being faster, its not like the coffee lake is that much faster to offset the price difference. And we still havent seen optimisations for the infinity fabric yet.

Okay, let me stop you right here. If you honestly think Intel NEEDS a new architecture to answer Ryzen, you are mistaken. They can simply lower prices and put several nails in AMD's coffin. Remember, Ryzen's ONLY appeal is it's prices and PCIe lane counts. Weaker IMC, inferior IPC, high cross-CCX latency (and poor scaling of the infinity fabric on lower CCX count SKU's), etc.

 

Coffeelake has less to do with Intel answering Ryzen, and more to do with them being greedy enough to sell consumers the same architecture 3 times in a row (Skylake 3.0). As stated earlier, answering ryzen is as simple as lowering prices. 

 

I also like how you completely changed your own argument from AMD being as fast or faster than Intel, to one that is about price:performance. As for "AMD would already be dead", that's simply not true. They have been bailed out a few times now, and that trend would likely continue if Intel became over-competitive. For the sake of not letting you get away with your misinformation, allow me to go back and double down on my points.

 

https://hardforum.com/threads/amd-ryzen-blender-benchmark-scores-thread.1919896/

 

Quote

EDIT!

As many people noticed, the actual download file AMD provided via their website was giving different results than the test performed at the New Horizon event. AMD has aknowledged this and updated the download. If you have not re-downloaded the benchmark in the last 24 hours, please do so and run again. Your results will look wimpy and flacid compared to the new results. 

https://www.pcworld.com/article/3151464/hardware/you-can-find-out-how-your-cpu-compares-to-amds-ryzen-for-free.html

Quote

Step 4: Check the render sample size

You're almost ready to hit the launch button, but the last thing you should check before you proceed is the render sample size. If you downloaded the file within the last day, it should default to a sample size of 150. 

There’s been a lot of confusion over the sample size, because AMD has performed the demo at 3.4GHz twice with two different settings.

The initial test, which was captured by PCPer.com on video, was done for press and analysts. It used a sample size of 100 and took 24 seconds to complete. AMD officials told me this was done to make the test run a little shorter for the attention-span-limited press (OK, they didn’t call the press that, I did.)

The public demo AMD did at its unveiling of the Ryzen used a sample size of 150 and took 36 seconds. I verified the results and the sample size with AMD as of Friday morning. AMD also initially released the test file with a 100 sample size but has since updated it to 150. So yeah, there’s a lot of confusion. Just make sure you set yours to 150.

People did not know about this render sample change until after the initial tests, which made Ryzen look much faster than what it actually was. When the leaked Blender benchmarks came out, people were trying to test this themselves on stock blender settings (which is what they assumed AMD used) and saw themselves have extremely slower times than AMD, which caused them to assume AMD was X percentage faster, when in reality, they were on par with Stock Intel X99 CPU's, and slower when comparing OC vs OC. 

 

Instead of trying to justify your misinformation with a video on why Intel is a bad company, why not present some actual performance results instead? That would be far more beneficial to your claims. 

[MageTank]

2 minutes ago, Sampsy said:

 

Again, cannonlake is mobile only. 

No, it's not. Where are you getting this?

[Bouzoo]

59 minutes ago, System Error Message said:

Why do people think intel has higher IPCs than AMD? 

I really don't know. Let's see for instance what Anandtech has to say about that. 

Quote

As you would expect, AMD still lags in IPC to Intel, so a 4.0 GHz AMD chip can somewhat compete in single threaded tests when the Intel CPU is around 3.5-3.6 GHz, and the single thread web tests/Cinebench results show that.

Oh right, that's why. 

[Bananasplit_00]

 

Spoiler

NOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO!!!!

Intel can you not? i really want everyone here in Sweden to go and upgrade right away.... that way i could finaly get a 7700K and an OCFormula board ;-;

[MageTank]

https://www.intel.com/content/www/us/en/processors/core/8th-gen-processor-family-s-platform-datasheet-vol-1.html

Going to get ahead of it this time around, before people start saying DDR4 XMP kits will kill the IMC. As with literally every other product launch, Intel only recommends stock JEDEC voltages + 5%. So 1.2 + 5% = 1.26v for vDIMM. DDR4 now has XMP kits going up to 1.5v. For those of you that think vDIMM will kill your IMC, take solace in the fact that almost every person using any modern CPU, has been going way outside of Intel's recommended spec with no negative consequences. 

 

Also, the ARK pages for CFL is up, if anyone wants to compare them against Skylake and Kaby:

https://ark.intel.com/products/126684/Intel-Core-i7-8700K-Processor-12M-Cache-up-to-4_70-GHz

[TidaLWaveZ]

All my Scandinavians hit me up and I will source for a small fee. 

[rip]

4 hours ago, VegetableStu said:

no. You don't say 24133562 Usain Bolts run at the speed of light.

UB max speed is 28mph

24133562 x 28 = 675739736 mph

Speed of Light is 186000 mi per sec = 670200000 mph

 

Im impressed actually 

 

[rip]

44 minutes ago, TidaLWaveZ said:

All my Scandinavians hit me up and I will source for a small fee. 

"small" 

[Cookybiscuit]

I wish Ryzen fanboy shitposters would stop shitting up Coffee Lake threads.

[Dabombinable]

Perhaps the fabs are being pushed to their limits with all of the monolithic designs?

[snortingfrogs]

If this is indeed true I will probably let my relatives in Las Vegas buy me an 8700K (if I decide to go the Coffee Lake road instead of Skylake-X).

[Dabombinable]

5 hours ago, LAwLz said:

I think it's cool how everyone goes around saying that Ryzen 7 is significantly cheaper, yet I am fairly sure there have been no prices announced.

Hell, we don't even know the performance (one poorly written review from a fairly unknown source does not count) and people are already talking about which chip performs the best.

 

Really interesting to see people talk about price:performance with high confidence when neither price nor performance are known yet...

Look at the pricing this way. Coffee Lake also has the cost of an iGPU added in unlike Ryzen, which would be partly the reason for Intel's typically higher prices post Core 2/Core i gen 1. (Not having to pay for an iGPU is a positive for my friend-GTX 650ti 2GB wreck any of Intel's iGPU)

[Princess Luna]

5 hours ago, Ashiella said:

@Princess Cadence just move to the sweeds

Aw, I feel bad that people might not be capable of having this processor for so long, in fact by the time it gets available it will be "worth waiting canonlake" any ways.

 

Lets see how it goes here in Brazil, I should be selling the i7 6700+mobo+ram this next week already, won't have a personal desktop until I can get the 8700k and the cheapest asus z370 possible

[Sakkura]

2 hours ago, TidaLWaveZ said:

All my Scandinavians hit me up and I will source for a small fee. 

Winter is coming, we'll just go with Skylake-X to keep warm.

[Misanthrope]

4 hours ago, VegetableStu said:

Oh wow I somehow completely missed that O_O

 

Launch SRP comparison:

• 8700K: 359
• 7700K: 339 (346 adjusted for inflation)
• 8700: 303
• 7700: 303 (309)
• 8600K: 257
• 7600K: 242 (247)
• 8400: 182
• 7500/7600: 192 - 213 (196 - 218)
• 8350K: 168
• 7350K: 192 (196)
• 8100: 117
• 7100: 117 (120)

In fairness to you and @LAwLz someone has pointed out to me that those prices are for retailers buying batches of 1000 chips. There might be anywhere between a little and a lot of price hikes once it actually goes on sale. Anywhere from just 20 or so bucks to far over 400 if this scarcity rumors turn out to be true.

 

So we might be looking at 8700k at 400 at launch and maybe down the line it will come down to 360-370.

[MageTank]

3 minutes ago, Misanthrope said:

In fairness to you and @LAwLz someone has pointed out to me that those prices are for retailers buying batches of 1000 chips. There might be anywhere between a little and a lot of price hikes once it actually goes on sale. Anywhere from just 20 or so bucks to far over 400 if this scarcity rumors turn out to be true.

 

So we might be looking at 8700k at 400 at launch and maybe down the line it will come down to 360-370.

Yes, those are the tray prices. The tray prices can be above or below MSRP, depending entirely on market supply and demand. Intel has not added the "recommended customer price" yet on ARK, but if you visit this page after Oct 5th, you will likely see their recommended price: https://ark.intel.com/products/126684/Intel-Core-i7-8700K-Processor-12M-Cache-up-to-4_70-GHz

 

I expect retailers to take advantage of the hype and high demand to charge $400 for these CPU's just like they did with Skylake, and it may take a month or so for the demand to die down. Places like Microcenter actually sell these CPU's much closer to cost, or even at a loss to help get customers into the store to buy other various components, so those of  you that may have a Microcenter nearby, you may luck out and score one of these CPU's at a slight discount. 

[Notional]

6 hours ago, MageTank said:

Ryzen brought haswell-level IPC, not Skylake/Kaby/CFL (which all 3 have the same IPC).

Ryzen is Broadwell level IPC, not Haswell. The difference between Skylake and Broadwell is 5-10% IPC. Kaby is 0%. Intel mostly has the clock rate benefit.

[MageTank]

1 minute ago, Notional said:

Ryzen is Broadwell level IPC, not Haswell. The difference between Skylake and Broadwell is 5-10% IPC. Kaby is 0%. Intel mostly has the clock rate benefit.

Every test I saw showed Haswell-level IPC. Even so, broadwell only has a 2-3% advantage over Haswell at best. I don't quite understand where you are getting the 10% IPC boost over Broadwell on Skylake, as I have not seen that anywhere. I find it to be yet another 3-4% at most. I've actually tested this myself, as I have a 4790k sitting behind me, a 7700k in my system, and a 6700k sitting on my desk doing absolutely nothing, lol.

 

Do you have any tests you would like me to run with these chips so that we can try to get a definitive number on IPC? I know it varies per instruction set used, but I'd like to get a range of different tests that could cover the general basis for most real-world functions. 

 

Using the data Anandtech produced, their findings seem very similar to what I myself saw when comparing strictly using static clocks and various tests from the Aida64 test suite:

Quote

This graph shows that:

Sandy Bridge to Ivy Bridge: Average ~5.8% Up
Ivy Bridge to Haswell: Average ~11.2% Up
Haswell to Broadwell: Average ~3.3% Up
Broadwell to Skylake (DDR3): Average ~2.4% Up
Broadwell to Skylake (DDR4): Average ~2.7% Up

Oh dear. Typically with an architecture update we see a bigger increase in performance than 2.7% IPC.  Looking at matters purely from this perspective, Skylake does not come out well. These results suggest that Skylake is merely another minor upgrade in the performance metrics, and that a clock for clock result compared to Broadwell is not favorable. However, consider that very few people actually invested in Broadwell. If anything, Haswell was the last major mainstream processor generation that people actually purchased, which means that:

Haswell to Skylake (DDR3): Average ~5.7% Up.

PCper seems to believe Haswell to Skylake is a 10% difference in IPC using Cinebench, though my personal tests with Cinebench has it at roughly 7-8%. Perhaps the difference in IPC results stems from the fact that I heavily tweak my ram on every platform (2133 C8-10-10-18-1 on Haswell with extremely tight tertiary timings, 3600 C14-14-14-28-2 on Skylake and above with extremely tight tertiary timings). Their usage of 2133 JEDEC DDR4 surely puts it at an extreme disadvantage over JEDEC DDR3 1600, even when comparing completely untouched secondary/tertiary timings. We are talking a 3ns difference in absolute latency alone. 

https://www.pcper.com/reviews/Processors/Intel-Core-i7-6700K-Review-Skylake-First-Enthusiasts/Testing-Setup-and-SiSoft-San

 

Still, I'll gladly test any test you personally want to see, as when I bring this discussion up, most people seem to be in disbelief of my numbers, yet never really give me a methodology to test against. 

[Notional]

53 minutes ago, MageTank said:

Every test I saw showed Haswell-level IPC. Even so, broadwell only has a 2-3% advantage over Haswell at best. I don't quite understand where you are getting the 10% IPC boost over Broadwell on Skylake, as I have not seen that anywhere. I find it to be yet another 3-4% at most. I've actually tested this myself, as I have a 4790k sitting behind me, a 7700k in my system, and a 6700k sitting on my desk doing absolutely nothing, lol.

 

Do you have any tests you would like me to run with these chips so that we can try to get a definitive number on IPC? I know it varies per instruction set used, but I'd like to get a range of different tests that could cover the general basis for most real-world functions. 

 

Using the data Anandtech produced, their findings seem very similar to what I myself saw when comparing strictly using static clocks and various tests from the Aida64 test suite:

PCper seems to believe Haswell to Skylake is a 10% difference in IPC using Cinebench, though my personal tests with Cinebench has it at roughly 7-8%. Perhaps the difference in IPC results stems from the fact that I heavily tweak my ram on every platform (2133 C8-10-10-18-1 on Haswell with extremely tight tertiary timings, 3600 C14-14-14-28-2 on Skylake and above with extremely tight tertiary timings). Their usage of 2133 JEDEC DDR4 surely puts it at an extreme disadvantage over JEDEC DDR3 1600, even when comparing completely untouched secondary/tertiary timings. We are talking a 3ns difference in absolute latency alone. 

https://www.pcper.com/reviews/Processors/Intel-Core-i7-6700K-Review-Skylake-First-Enthusiasts/Testing-Setup-and-SiSoft-San

 

Still, I'll gladly test any test you personally want to see, as when I bring this discussion up, most people seem to be in disbelief of my numbers, yet never really give me a methodology to test against. 

https://www.pcper.com/reviews/Processors/AMD-Ryzen-7-1800X-Review-Now-and-Zen/Clock-Clock-Ryzen-Broadwell-E-Kaby-Lake

 

Ryzen wins some and loses some. Indeed the gaming bench favours Intel quite a bit, but we both know games needs to be optimized for Ryzen architecture, just like games are now optimized for Intel architectures.

The Lame mp3 bench makes little sense as the 7700K at same clock beats the 6900k at same clock.

 

Then again, it seems there is less than 5% difference between Haswell and Broadwell. Just makes Intel look worse imho. Intel is still the IPC king, but it's mostly their clock rate advantage and software optimizations giving them the first prize.

[MageTank]

Just now, Notional said:

https://www.pcper.com/reviews/Processors/AMD-Ryzen-7-1800X-Review-Now-and-Zen/Clock-Clock-Ryzen-Broadwell-E-Kaby-Lake

 

Ryzen wins some and loses some. Indeed the gaming bench favours Intel quite a bit, but we both know games needs to be optimized for Ryzen architecture, just like games are now optimized for Intel architectures.

The Lame mp3 bench makes little sense as the 7700K at same clock beats the 6900k at same clock.

 

Then again, it seems there is less than 5% difference between Haswell and Broadwell. Just makes Intel look worse imho. Intel is still the IPC king, but it's mostly their clock rate advantage and software optimizations giving them the first prize.

Yeah... from a pure IPC perspective, Intel has not released anything interesting since Haswell. It's also why I try to tell people that Ryzen matching Haswell was more than enough for them to compete, as there is not that big of a difference between Haswell and Skylake (or Coffeelake, which will have identical IPC, albeit different core counts and clock speeds). 

 

As for the MP3 bench being slightly in favor of the 7700k over the 6900k, it might prefer Skylake's slightly higher IPC over Broadwell, as there appears to be a 4-5% difference in seconds here,though it's hard to certain without knowing cache speeds and memory speeds as they seemed to have neglected to disclose those. By default Kaby seems to run at a 200mhz cache deficit in my testing relative to any core clock you manually set.

 

I need to get my hands on the bigger Ryzen chips, as I really want to do extensive tests on inner-CCX latency and the impact it has on the memory sub-system. We see what it does on Threadripper, but I have not seen the tests ran on Ryzen yet. It might also explain the poor scaling on the lower end Ryzen 3 SKU's compared to ryzen 7, where the Ryzen 3 SKU's seem to have even higher memory latency as a whole. 

[Shreyas1]

This compliments the GPU market very well.