Coffeelake thread, benchmarks, reviews, input, everything.

[Paratus]

I would be interested to see which applications are running 50-60% faster on CFL than Ryzen, with significantly higher power usage to boot. Not that I actively disbelieve you, I just haven't yet seen this.

Ask for CFL vs Ryzen benchmarks he gives you Haswell vs FX.

Why would i used numbers from the horri-bad "benchmark" that you are proporting is worth anything in the first place as any kind of baseline?

i5 2500k / FX-4100
69 / 53 = 1.30188679245

4670k / 2500k
86 / 69 = 1.24637681159

And obviously Skylake/Kabylake/Coffeelake will be higher than 4670k.

While you don’t like Cinebench maybe consider finding a benchmark you do like with the numbers relevant to his question?

Just a thought.

 

[Yuriman]

Yeah I'm not sure what's relevant about 4670K vs 2500K in the discussion of CFL vs Ryzen.

 

[PG]

I like AMD but Intel is way ahead in terms of IPC and single threaded performance. It does show up in certain things like professional CAD programs. These are multithreaded programs, but more cores don't really matter. Engineers make parts one feature at a time. If you need to make a change to an early feature, then all following features need calculated again, one feature at a time. It can't be done in parallel. Rotations of parts on the screen works the same way and is basically single threaded. It's calculating how the part should look one feature at a time. I say single threaded, but maybe 2 cores are really getting used at most.

Puget Systems has lots of articles and benchmarks if you look around their website. (No, I don't work there or have one of their systems.) They did cpu benchmarks for SolidWorks and Revit, among others. Solidworks is probably the most popular 3D CAD software out there for engineering, and Revit is the most commonly used program by people in architecture.
AMD is way behind in each of these for most tasks, except rendering. AMD of course does well in rendering, but is way behind in almost all other tasks.
Engineers who design parts aren't going to be doing much rendering to be honest, and that's last after the parts have been designed.

Some examples:

https://www.pugetsystems.com/labs/a...offee-Lake-vs-Skylake-X-vs-Threadripper-1105/

https://www.pugetsystems.com/labs/articles/Revit-2018-Coffee-Lake-CPU-Comparison-1052/

 

[Yuriman]

Why would i used numbers from the horri-bad "benchmark" that you are proporting is worth anything in the first place as any kind of baseline?

i5 2500k / FX-4100
69 / 53 = 1.30188679245

4670k / 2500k
86 / 69 = 1.24637681159

And obviously Skylake/Kabylake/Coffeelake will be higher than 4670k.

Ok, I'll bite.

2500k vs FX-4100 had a 30% advantage in a new FPS which showed good CPU scaling.

I grabbed the two most recent FPS titles listed on GameGPU (so we're using the same site). They don't have CFL, but they do have Kaby, Skylake and Broadwell-X.

I'm not seeing Intel 30% ahead here. The 1300X essentially ties with the i5 6600 (basically same CPU as i3 8100). The 1600X is up next to the 5960X and 6850K.

So, your specific example isn't supporting what you're trying to assert.

Which is not to say that I'm not aware that Intel has better IPC AND higher clockspeed. This just isn't the same as the Bulldozer fiasco.

 

[24601]

Yeah I'm not sure what's relevant about 4670K vs 2500K in the discussion of CFL vs Ryzen.

https://us.hardware.info/reviews/78...arks-gtx-1080-ti-rise-of-the-tomb-raider-dx12

Ryzen is ~= Sandy Bridge in most games

 

[24601]

Ok, I'll bite.

2500k vs FX-4100 had a 30% advantage in a new FPS which showed good CPU scaling.

I grabbed the two most recent FPS titles listed on GameGPU (so we're using the same site). They don't have CFL, but they do have Kaby, Skylake and Broadwell-X.

I'm not seeing Intel 30% ahead here. The 1300X essentially ties with the i5 6600 (basically same CPU as i3 8100). The 1600X is up next to the 5960X and 6850K.

So, your specific example isn't supporting what you're trying to assert.

Which is not to say that I'm not aware that Intel has better IPC AND higher clockspeed. This just isn't the same as the Bulldozer fiasco.

I am deliberately using a 4 thread game so that the game isn't just cache thrashing all over the place like happens when you are using a 8 thread or 16 thread game.

What people attribute to "new most betterar" in "multicore scaling" in terms of console ports generally is due to the game cache thrashing with it's 8-16 thread load that it is stuffing onto less threads/cores on the CPU and one of the main reasons why the console ports from the Xbox One and PS4 play horribly on PC and why you are seeing more "multicore scaling" than there actually exists in the game engine, where you are seeing simply less cache thrashing when you are using a higher core count CPU in aforementioned games.

 

[Yuriman]

I am deliberately using a 4 thread game so that the game isn't just cache thrashing all over the place like happens when you are using a 8 thread or 16 thread game.

What people attribute to "new most betterar" in "multicore scaling" in terms of console ports generally is due to the game cache thrashing with it's 8-16 thread load that it is stuffing onto less threads/cores on the CPU and one of the main reasons why the console ports from the Xbox One and PS4 play horribly on PC and why you are seeing more "multicore scaling" than there actually exists in the game engine, where you are seeing simply less cache thrashing when you are using a higher core count CPU in aforementioned games.

The reality of things is that many (maybe even most) of the games people play are console ports. So, it's practically cherry-picking to exclude these "real-world" examples.

 

[24601]

The reality of things is that many (maybe even most) of the games people play are console ports. So, it's practically cherry-picking to exclude these "real-world" examples.

Ok, if you just want a cache latency and throughput test = IPC then sure, you can tout those badly ported console games as your measure.

Btw, thats exactly what Cinebench is testing as well, so we're back to literally the same place in your argument.

 

[Yuriman]

Ok, if you just want a cache latency and throughput test = IPC then sure, you can tout those badly ported console games as your measure.

Btw, thats exactly what Cinebench is testing as well, so we're back to literally the same place in your argument.

PG shared some benchmarks which are relevant to professional work being much slower on Ryzen, but I'm still waiting on you to post an actual graph or some real numbers for consumer workloads, rather than the nebulous "Ryzen = Haswell" in 5 year old benchmarks of what is probably the single most notoriously IPC-dependent PC game.

 

[24601]

PG shared some benchmarks which are relevant to professional work being much slower on Ryzen, but I'm still waiting on you to post an actual graph or some real numbers for consumer workloads, rather than the nebulous "Ryzen = Haswell" in 5 year old benchmarks of what is probably the single most notoriously IPC-dependent PC game.

It seems more like you don't want an example, as I gave one, and even gave the entire reasoning for the entire choice, design, and result of the test.

Literally everything was explained to the last detail and you simply discarded it so I see no point in wasting time in spoonfeeding you further .

 

[Yuriman]

It seems more like you don't want an example, as I gave one, and even gave the entire reasoning for the entire choice, design, and result of the test.

Literally everything was explained to the last detail and you simply discarded it so I see no point in wasting time in spoonfeeding you further .

"Ryzen is the same as Haswell, so here's a 5 year old game where Haswell is 25% faster than Sandy Bridge, which is 30% faster than Bulldozer. And obviously Coffee Lake is faster than Haswell. Therefore Ryzen = Bulldozer."

Do I have it right?

i5 2500k / FX-4100
69 / 53 = 1.30188679245

4670k / 2500k
86 / 69 = 1.24637681159

And obviously Skylake/Kabylake/Coffeelake will be higher than 4670k.

 

[24601]

"Ryzen is the same as Haswell, so here's a 5 year old game where Haswell is 25% faster than Sandy Bridge, which is 30% faster than Bulldozer. And obviously Coffee Lake is faster than Haswell. Therefore Ryzen = Bulldozer."

Do I have it right?

Who said "Ryzen is the same as haswell"?

I said Ryzen is the same as Sandy Bridge.

Level up your reading comprehension bro.

I'm starting to get the feeling that you didn't even read most of my posts, and simply skimmed everything I said instead and applied your own bias filter to "what I think that person said"

 

[Paratus]

"Ryzen is the same as Haswell, so here's a 5 year old game where Haswell is 25% faster than Sandy Bridge, which is 30% faster than Bulldozer. And obviously Coffee Lake is faster than Haswell. Therefore Ryzen = Bulldozer."

Do I have it right?

He pulled this shtick in VC&G. He passes his opinion off as fact. When asked to support it he doesn’t or does so poorly. Then he insults the other party by suggesting that supporting his opinion is spoonfeeding. When further challenged he gets huffy and leaves.

For whatever reason he’s emotionally invested in being anti-AMD.

 

[24601]

He pulled this shtick in VC&G. He passes his opinion off as fact. When asked to support it he doesn’t or does so poorly. Then he insults the other party by suggesting that supporting his opinion is spoonfeeding. When further challenged he gets huffy and leaves.

For whatever reason he’s emotionally invested in being anti-AMD.

Literally 90% of the forum attacking me personally in every post and I'm the one that gets banned, hilarious.

Reality isn't a popularity contest.

Banning me simply means you feel threatened by my knowledge.

It's not your "knowledge" we get tired of.
It's the trolling and personal attacks (and borderline
mod call-outs).

AT Mod Usandthem

 

[Paratus]

Literally 90% of the forum attacking me personally in every post and I'm the one that gets banned, hilarious.

Reality isn't a popularity contest.

Banning me simply means you feel threatened by my knowledge.

I have no power to ban you so I’m not sure why you are on about being banned

Second being asked to justify your position in technical forum is hardly being attacked.

You made the statements that CFL was 50-60% faster than Ryzen

What we would like to see is a benchmark you approve of, since you don’t like Cinebench, that shows CFL 50-60% ahead of Ryzen.

Right now that statement CFL 50-60% > Ryzen is only backed up by your opinion.

Benchmarks on this site show Ryzen IPC roughly around Haswell/Broadwell with multi threaded performance on average being faster than Skylake or Kabylake due to the largee number of cores.

Coffelake has significantly higher IPC than Ryzen but not 50-60% more and certainly not in multithreaded applications, outside of few outliers, as most of us understand it.

So instead of defensively attacking those who ask for clarification why don’t you provide the information you used to support your opinion.

No moderator here is going to ban you for supporting your opinion.

 

[24601]

It seems more like you don't want an example, as I gave one, and even gave the entire reasoning for the entire choice, design, and result of the test.

Literally everything was explained to the last detail and you simply discarded it so I see no point in wasting time in spoonfeeding you further .

Lack of reading comprehension seems to be a problem here.

We've had enough. You keep doing the
same things, and expecting different results.
See you in a month.

AT Mod Usandthem

 

[Dayman1225]

6 Core 12 Thread CFL Xeon E 2716M has been spotted on SiSoft with what I assume is a 2.7GHz base clock.

Here

 

[ondma]

Ok, I'll bite.

2500k vs FX-4100 had a 30% advantage in a new FPS which showed good CPU scaling.

I grabbed the two most recent FPS titles listed on GameGPU (so we're using the same site). They don't have CFL, but they do have Kaby, Skylake and Broadwell-X.

I'm not seeing Intel 30% ahead here. The 1300X essentially ties with the i5 6600 (basically same CPU as i3 8100). The 1600X is up next to the 5960X and 6850K.

So, your specific example isn't supporting what you're trying to assert.

Which is not to say that I'm not aware that Intel has better IPC AND higher clockspeed. This just isn't the same as the Bulldozer fiasco.

Kind of sad. Game.gpu used to be a great site for gaming results. Dont even look at them anymore. They have the latest and fastest AMD cpus, but their selection of mainstream Intel cpus is pathetic. Yes they have K/L, but only a middle of the road locked i5. Yes they have S/L, but even that 3 gen old processor is the locked, lower clocked 6700. Pathetic.

 

[IntelUser2000]

6 Core 12 Thread CFL Xeon E 2716M has been spotted on SiSoft with what I assume is a 2.7GHz base clock.

Here

Correct. Xeon E 2176M is a Xeon version of Coffeelake H.

 

[dullard]

My need for a new computer is high and I've been on the verge of buying since last fall. I was really wanting an 8600. But it never was released. The 8400 is underwhelming and the 8600K costs 41.2% more (plus an annoyingly higher TDP) for a measly 7.9% turbo speed boost for the vast majority of those of us who don't overclock.

Well the 8600 that I wanted was finally released and I'm left a bit disappointed. Sure, getting 8600K turbo speeds for $44 less with the 8600 is a nice thing. And it is great to have something fill the massive price void between the 8400 and 8700 for those who don't overclock. But, there just isn't enough difference between the 8600 and the 8400 to justify the 8600. I think this time around, the 8500 is the best value i5 processor.

 

[Eug]

Well the 8600 that I wanted was finally released and I'm left a bit disappointed. Sure, getting 8600K turbo speeds for $44 less than the 8600 is a nice thing. And it is great to have something fill the massive price void between the 8400 and 8700 for those who don't overclock. But, there just isn't enough difference between the 8600 and the 8400 to justify the 8600. I think this time around, the 8500 is the best value i5 processor.

This is what happens when there is an embarrassment of riches. All of these chips you mentioned are excellent IMO, at least when compared to what was available last year.

 

[jpiniero]

AT didn't cover this, but there are "B" models of the 8700, 8500 and 8400. The only difference is that these are BGA. Ark says they are compatible with all the chipsets as well.

 

[Bouowmx]

I think this time around, the 8500 is the best value i5 processor.

Day 1 shenanigans of pricing of i5-8500 and 8600 vs below-MSRP 8400.

 

[epsilon84]

Well the 8600 that I wanted was finally released and I'm left a bit disappointed. Sure, getting 8600K turbo speeds for $44 less with the 8600 is a nice thing. And it is great to have something fill the massive price void between the 8400 and 8700 for those who don't overclock. But, there just isn't enough difference between the 8600 and the 8400 to justify the 8600. I think this time around, the 8500 is the best value i5 processor.

In the grand scheme of things, a $10 or $20 difference doesn't equate to much in terms of the total cost of ownership.

For example:
8400 ($180) + B360 mobo ($90) + 16GB DDR4 2666 ($170) = $440 platform cost for 3.8GHz MT / 4.0GHz ST
8500 ($190) + B360 mobo ($90) + 16GB DDR4 2666 ($170) = $450 platform cost for 3.9GHz MT / 4.1GHz ST
8600 ($210) + B360 mobo ($90) + 16GB DDR4 2666 ($170) = $470 platform cost for 4.1GHz MT / 4.3GHz ST

Doesn't look like such a bad value when viewed in that context, does it? And that's just the CPU + mobo + RAM, if you include all the other system parts like a GPU, SSD, PSU, case, monitor etc the extra $10 or $20 pales into insignificance, really.

 

[dullard]

In the grand scheme of things, a $10 or $20 difference doesn't equate to much in terms of the total cost of ownership....if you include all the other system parts like a GPU, SSD, PSU, case, monitor etc the extra $10 or $20 pales into insignificance, really.

Your point is valid. Although, it is closer to $31 difference in MSRP for the 8600.

The problem with your approach though is that it leads to computer inflation. Oh, what is $30 more on a CPU, $20 more on a GPU, $10 more on a SSD, $25 more on a PSU, $5 more on a case, etc. Looked at as a system, each change is meaningless. But then suddenly your $400 system costs $700.

I prefer to analyze each component in isolation. That way my analysis is true for upgrades or whole builds. It also eliminates build inflation. But, I do keep in mind that if one thing is just a measly $10 more, then I can just get it since $10 alone is nothing. Which is why I think the 8500 is a better value than the 8400. The extra 100 MHz to 200 MHz might not be noticed, but it is certainly better. Whereas the $10 will never be noticed.

I was just hoping that the 8600 would be like the 7600 or the 6600 (500 MHz to 600 MHz better than the 7400 and 6400 comparison). Even if you go all the way back to Nehalem, the top non-K i5 has always been at least 400 MHz faster (usually 600 MHz faster) than the bottom i5. Suddenly now it is just 300 MHz. Similarly with the 7700 and 6700, they were just 100 MHz faster than the 7600 and 6600. But now there is a 200 MHz to 300 MHz gap between the 8700 and 8600.