Ryzen 2700X OCed to 4.3Ghz (1.4v) across all cores, performance numbers included.

[Razor01]

32 minutes ago, leadeater said:

Think you missed the point, we don't look at CPUs we use them. Ohh shiny 8 cores, ok but.... you know this will actually get used right so how much?

 

 

That why I made the point if you want to look at price, anything above 4 core is a waste of money for most people.  Even for future proofing lol if there was such a thing.

[Razor01]

33 minutes ago, Taf the Ghost said:

The 2+2 Cannonlake part has vanished, leaving on the 2+0 U part that it appears only Intel itself will use in extremely small numbers.

 

So, yes, they've announced they're going to ship something from the baseline 10nm node, even if it probably doesn't ever really make it to market, existing as something of a novelty.

 

Icelake will be on 10nm+ process node, and full scale production isn't until July, I believe, which points towards a Q2 2019 release, though, as I said, they could paper launch some SKUs in late 2018.

 

 

Isn't 2+2 cannolake slated to be replaced by Skylake U 4 core?

[Trixanity]

53 minutes ago, Taf the Ghost said:

@Razor01

 

8c part won't be Icelake or 10nm. That's not until 2019.

Gotta move that goalpost when it doesn't fit that gosh darn agenda :-)

 

>pesky Zen consumes power like a truck 

>but that new 8c Intel chip will increase an already precarious amount of power drawn by CFL-S

>10nm will fix that

>8c won't be on 10nm

>mfw

 

We're probably at least a year away from desktop 10nm so I fully expect to see 10nm Intel vs 7nm AMD (desktop) within 3 months of each other.

[Razor01]

40 minutes ago, Trixanity said:

Gotta move that goalpost when it doesn't fit that gosh darn agenda :-)

 

>pesky Zen consumes power like a truck 

>but that new 8c Intel chip will increase an already precarious amount of power drawn by CFL-S

>10nm will fix that

>8c won't be on 10nm

>mfw

 

We're probably at least a year away from desktop 10nm so I fully expect to see 10nm Intel vs 7nm AMD (desktop) within 3 months of each other.

 

 

Actaully coffee lake doesn't use crazy amounts of power if you want to look at clock for clock thread for thread, core for core perf/watt, coffeelake is way ahead of Zen.  Looking at around 50% perf/watt depending on application.  All with an integrated GPU!

 

Now the GPU might not be used but its still powered up, so its burning some power even though its very little.

 

Moving goal posts you say?  You can't move goal posts with what I have stated.  Unless we talk about pricing and shit like that.  That is taking away from what I was stating. 

 

Zen is WAY behind when looking at Coffee Lake Core for Core performance which translates over to multithreaded performance too.  The SMT advantage Zen had over Kaby Lake is less with Coffee Lake.

 

Lets look at power figures for 8700 and compare that to 1700x.  Outside of gaming this Intel chip with 6 core 12 threads is still going to have a sizable advantage in pretty much everything, there are some apps that will be close and take the lead for the 1700x but few and far between.  With the 8700 at a TDP of lower than 30% of 1700x.  So if you want to look at perf/watt gotta combine performance and power draw right?  Its going to end up around 40 to 50% difference over all.  Was it really a silly statement when nV stated perf/watt is going to be what matters in the future?  To many non technical people it did, we all read the crazy posts about this.  But for people that understand how that affects product line ups and potential of intangibles, it makes a butt load of sense.  The ramifications of perf/watt over the course of developing these products don't change much from gen to gen unless design is significantly change.  You can't just tweak a few things and expect power consumption to drop.  Doesn't work that way.   We can also compare this to a 1600x and get similar conclusions too!

 

In a scenario like this Intel really doesn't even need the 10nm process to stay level with AMD products.  Of course its better if they do because it will give them breathing room.  We can even go so far to say if AMD gets to 7nm before Intel gets to 10nm, then they were actually be on level with Intel architectures on 14 nm!

 

Where did we hear things like this before, Pentium 4 vs Athlon 64, the node advantage will solve problems right?  We heard this excuse on the graphics side of things too.  Nah its not the node that will solve this type of problem, it needs to be done with design.

[Taf the Ghost]

8 hours ago, Razor01 said:

 

 

Isn't 2+2 cannolake slated to be replaced by Skylake U 4 core?

Correct. 2+0 and 2+2 U parts showed up, and there was a production run in the last weeks of December (so Intel's CEO could say 10nm is in production). Then the 2+2 disappeared and we have something called "Whiskey Lake" that'll fill in for it on the 2nd half of the year. Though the Taiwan production leaks also called it on 14nm+++, which no one has any clue what that mean. (Though a few other connected types have mentioned there's another revision out there, but it might just be some low-power tweaks.)

 

Right now it looks like the 2+0 will be in some Intel-based products in minuscule runs and 2+2 isn't happening. Part of me wants to track down one of the 2+0 parts just for the history of the disastrous 10nm node. Also, while vague, there was something really wrong with the base 10nm process. Yields apparently got good enough a bit ago, but something else killed it ever being put into full scale production. It's just guesses for what the secondary issue is.

[Jito463]

On 4/11/2018 at 8:55 AM, Razor01 said:

That's why I stated compare it to another 95 watt Intel chip!  One that actually has the same core count and threads!

The i5-8400 is a 65W chip, not a 95W one.  The Ryzen 5 1600x is a 95W chip.  The 1600 is the 65W part that it should be compared against for power draw, though.

[Razor01]

2 minutes ago, Jito463 said:

The i5-8400 is a 65W chip, not a 95W one.  The Ryzen 5 1600x is a 95W chip.  The 1600 is the 65W part that it should be compared against for power draw, though.

http://thetechaltar.com/ryzen-3-vs-coffee-lake-i3-per-clock-performance-comparison/2/

 

Clock for clock core for core thread per thread comparison of Coffee lake and Ryzen 1, this is without the CCX latency issues too.

 

What do we see no SMT advantage for AMD at all.  Everything comes down to a pure advantage of IPC that Intel chips have over AMD's.

[Taf the Ghost]

56 minutes ago, Razor01 said:

http://thetechaltar.com/ryzen-3-vs-coffee-lake-i3-per-clock-performance-comparison/2/

 

Clock for clock core for core thread per thread comparison of Coffee lake and Ryzen 1, this is without the CCX latency issues too.

 

What do we see no SMT advantage for AMD at all.  Everything comes down to a pure advantage of IPC that Intel chips have over AMD's.

Good link, though I'd be really curious what would happen on Ryzen if we got 4-way SMT. That'd be funny.

 

https://www.anandtech.com/show/10435/assessing-ibms-power8-part-1/4  IBM went all the way to 8-way, haha.

 

And we're talking about "Effective" IPC. Most of Intel's IPC advantage is in the AVX2 (so 256-bit geometry) instructions. That's what produces most of the difference in workloads that don't thrash core-to-core communications. i.e. Encoding tasks. What matters for gaming is actually not the IPC itself: it's the Memory & I/O systems. Intel has had a memory advantage on AMD for as long as I can remember. I don't know exactly why, but I'd guess it is one of those areas where just raw amounts of Engineers on the project matter.

 

The Ryzen 3 1200 is still a 2+2 CCX design part, so it gets the memory penalty. Also, the testing was done at 2400 with loose timings, which should hurt the Ryzen part more than the Intel part, but both would respond positively to going to at least the max JEDEC spec of 2666 that they support without an OC.

 

I find the topic interesting simply because it's actually Intel that has a bigger problem. It's very clear the areas AMD will be focusing on for the next few generations, though some of the memory latency issues will never quite go away, though as AMD finds ways to lower them, they'll be far less of an issue. (Along with the scheduler improving over time.)

[Tellos]

I fail to see why some are upset a refresh didint get to 5hz or something. Seems like nobody is ever satisfied.

[Jito463]

3 hours ago, Razor01 said:

*snip*

I wasn't talking about the performance differences.  Earlier in the thread, you were commenting about the power draw between the 1600x and the 8400, while stating both were 95W chips.  The 8400 is actually a 65W chip, so the equivalent chip power-wise would be the 1600 (non-X).

[Razor01]

Just now, Jito463 said:

I wasn't talking about the performance differences.  Earlier in the thread, you were commenting about the power draw between the 1600x and the 8400, while stating both were 95W chips.  The 8400 is actually a 65W chip, so the equivalent chip power-wise would be the 1600 (non-X).

 

 

Sorry my original point got all muddled up, I was trying to say if we look at equivalent thread for thread core for core, Intel still has an upper hand with CFL.  Too many side thoughts going on lol.