Intel's 8th Generation Core Family - Coffee Lake (LGA 1151, 6C/12T)

Where do you expect Core i7-8700K's Turbo to land?

  • 3.8/3.9 GHz

    Votes: 0 0.0%
  • 4.0/4.1 GHz

    Votes: 3 23.1%
  • 4.2/4.3 GHz

    Votes: 6 46.2%
  • 4.4/4.5 GHz

    Votes: 3 23.1%
  • 4.6/4.7 GHz

    Votes: 1 7.7%

  • Total voters
    13
  • Poll closed .
You're right. It's not a game. It's a cherry picked example. And the cherry picking can go both ways.

Ryzen-Benchmarks-Stock.004-150x150.png


Oh no!...the precious 7700k is beaten by 40% in Cinebench R15.

Using my zoom-eye abilities... No, Zen IPC is not 40% ahead KBL. In fact, Zen has lower IPC than Broadwell in that bench.
 
Using my zoom-eye abilities... No, Zen IPC is not 40% ahead KBL. In fact, Zen has lower IPC than Broadwell in that bench.

I know you're obtuse, but you understand the point of cherry picking benchmarks to suit your point don't you? Here...I'll even post from the same pcper article you did.
clock-cb15-2.png
 
200+ dollar samsung ram to get the best performance vs buy whatever DDR4-2666 (around 140-150 usd atm)without worry it will work and won't hurt you performance wise in games and probably not in much of anything else. added cost on ram easily makes of for current motherboard price differences. the low end boards come out in January but who seriously buys a K sku CPU without a chipset that can make use of overclocking..

G. skill 3200 mhz kits (16 Gb) can be found for as low as $150 right now. There is a very little price savings over 2666 Mhz.
Also, Ryzen performance doesn't increase much after 3200 Mhz as stated in this rescent Anantech article:

https://www.anandtech.com/show/11857/memory-scaling-on-ryzen-7-with-team-groups-night-hawk-rgb/7

Finally, I doubt new 8700k buyers are purchasing them with low speed Ram. There is always that worry it may hinder performance... and no one wants to risk having to buy Ram twice.

If someone already has crappy ddr4 that can't go past 2666 mhz, then yes Ryzen may not be the best route right now.
 
G. skill 3200 mhz kits (16 Gb) can be found for as low as $150 right now. There is a very little price savings over 2666 Mhz.
Also, Ryzen performance doesn't increase much after 3200 Mhz as stated in this rescent Anantech article:

https://www.anandtech.com/show/11857/memory-scaling-on-ryzen-7-with-team-groups-night-hawk-rgb/7

Finally, I doubt new 8700k buyers are purchasing them with low speed Ram. There is always that worry it may hinder performance... and no one wants to risk having to buy Ram twice.

If someone already has crappy ddr4 that can't go past 2666 mhz, then yes Ryzen may not be the best route right now.

150 dollar 3200 kits are not samsung IC"s they are Hynix. if you mistakenly buy that for Ryzen and trust me i see Many people that do on weekly basis you wont have a chance at ever getting them to do 3200 on a Ryzen system. 2933 at best is what people end up with working fully stable..

edit: idk how you or anandtech comes to the conclusion that performance doesn't increase much past 3200 on ryzen... both cpu and gaming charts in their very own article show the 3333 outperforming every other setting in 75+ % of the tests what they should have done is not been dumb and tested all the way up to 3600 to get a proper big picture not a narrow range of 2400-3333
 
Last edited:
https://community.amd.com/community...emory-oc-showdown-frequency-vs-memory-timings

pastedImage_11.png


  • Conclusion #1: Dual rank DIMMs (yellow) offered the best performance amongst “set and forget” (light blue, orange, yellow) memory configured automatically by XMP profiles.
  • Conclusion #1a: But the increased overclocking headroom of single rank modules was more than enough to overpower the benefits of rank interleaving, so manually-tuned single rank DDR4-3200 and 3466 won the day (dark blue and green).

btw all these above are on samsung b-die except the dual rank which i think should be D-die samsung, nothing else can do these speeds/timings on either amd or intel.
 
I know you're obtuse, but you understand the point of cherry picking benchmarks to suit your point don't you? Here...I'll even post from the same pcper article you did.
clock-cb15-2.png


Amazing that you accuse others of being "obtuse" and "cherry picking benchmarks". Let me recall you the history of this discussion:

1) Shintai claims that the IPC gap is 30% when there is no bottleneck.
2) Nightfire replies that RyZen is not 30% behind KBL, except on some "poorly optimized game".
3) I give a non-game benchmark where RyZen IPC is 30% behind KBL.
4) You claim that KBL is beaten by 40%, but you give a benchmark comparing 8-core, 10-core, and 4-core at different clocks.
5) You further reply now with another benchmark comparing 8-core and 4-cores at same clocks.

When someones pretends that RyZen IPC is never 30% behind KBL outside of gaming, all what is needed is to find a single benchmark outside gaming where the IPC is 30% behind. That is not cherry picking.
 
This guy does:


He compares 3200 cas 14 to 3600 cas 16 which didnt see much difference except maybe me:a.

I wish he would have compared 3200 cas 16 to either one of the above combos.

In any case those 2 ram combos ARE expensive AND greatly reduce the i7 advantage so lets call it a wash.

The fastest ram in a legitimate review is 3200 cas 16.
 
This guy does:


He compares 3200 cas 14 to 3600 cas 16 which didnt see much difference except maybe me:a.

I wish he would have compared 3200 cas 16 to either one of the above combos.

In any case those 2 ram combos ARE expensive AND greatly reduce the i7 advantage so lets call it a wash.

The fastest ram in a legitimate review is 3200 cas 16.



lol once again i look through the benchmark results and see 3600 C16 on top for ryzen in the vast majority of games.. the scaling is pretty much perfect each step up when that is the case as well.. i'd question how good that youtuber is though at benchmarking things considering some how he got 3200 c14 to score higher in a few. not to mention i'd question this guys 7700k results off hand they look to be false to me based on tons of other much better reviews comparing overclocked 7700k to R7 overclocked like [H]'s very own one done https://www.hardocp.com/article/2017/05/26/definitive_amd_ryzen_7_realworld_gaming_guide/7
 
This guy does:


He compares 3200 cas 14 to 3600 cas 16 which didnt see much difference except maybe me:a.

I wish he would have compared 3200 cas 16 to either one of the above combos.

In any case those 2 ram combos ARE expensive AND greatly reduce the i7 advantage so lets call it a wash.

The fastest ram in a legitimate review is 3200 cas 16.



another big issue with his testing... he used a 1070...

I just duplicated his rise of the tomb raider test on my rig COMPLETELY STOCK 1080 Ti (2000 core, stock ram) i5-7600k @ 4.2 Ghz DDR4 - 2133!!

completely blew all his scores away, in any event i'm cpu bottlenecking hard so for "science" ill overclock my cpu to remedy that see what i get and then overclock ram and see if it helps

https://imgur.com/a/HC9fL


5.1 ghz... 1080 Ti still bottlenecking hard at these settings..

https://imgur.com/a/FoOXN https://imgur.com/a/Uk0Uy

point is you can't test cpu gaming performance if your gonna be gpu bound.

and finally 5.1 ghz 4133 ddr4

https://imgur.com/a/ZTQmh

still cpu bound but staying above 90% 2133 default ddr4 vs 4133 c19 is a pretty massive difference in memory performance though.

to summarize

4.2 ghz 2133 - 187.92 fps
5.1 ghz 2133 - 208.66 fps
5.1 ghz 4133 - 224.98 fps

using a fan cooled overclocked 1070 is a poor choice also, pascal dynamically adjusts clocks based on GPU temp even as low as 40C! the source of some of his variance i'd be willing to bet is GPU throttling.

I guess i really shouldn't call my 1080 Ti completely stock for the sake of transparency because its watercooled and has a special bios flashed to it removing power limits (also why it has a 2000 mhz "stock" boost clock now) but i can tell you due to watercooling it won't throttle and will always put out consistent figures an air cooled heavily overclocked 1070 is going to exhibit all kinds of throttling that is not readily apparent to casual observers and completely jack with any sort of consistency in producing comparable benchmark scores. if he is a decent reviewer he should have better graphics cards to use than that for these sorts of tests.
 
Last edited:
AMD continues giving us 4-core APUs for mainstream after a decade!

why would it matter what amd gives on APU? besides im talking about intel here you have gotta get that through lol. business A doing bad practice to milk consumer, regardless of what business B does, A still milks consumer. in this case, intel milks hard. i can feel it, 8 cores for 20 yrs from intel.
 
I've had a think about my budget and how much I use my desktop now that I'm very, very old. As much as it pains me to admit, I think I'm going to buy the 8400. It'll probably be 65 -> 70% as fast as the 8700k for almost exactly half the price.
It will also be a significantly better deal than the AMD 1600 (even overclocked) since I don't need a video card and I hate having to pony up $60 US for a good quality one with the AMD 1600.
So unless AMD drop the price of the 1600 to $140 US or cheaper in the next 4 weeks. I'm getting the 8400. Should last me until DDR5 and 5Ghz 10nm CPUs in 3 years time.

Just hard to justify as a non gamer.
 
I've had a think about my budget and how much I use my desktop now that I'm very, very old. As much as it pains me to admit, I think I'm going to buy the 8400. It'll probably be 65 -> 70% as fast as the 8700k for almost exactly half the price.
It will also be a significantly better deal than the AMD 1600 (even overclocked) since I don't need a video card and I hate having to pony up $60 US for a good quality one with the AMD 1600.
So unless AMD drop the price of the 1600 to $140 US or cheaper in the next 4 weeks. I'm getting the 8400. Should last me until DDR5 and 5Ghz 10nm CPUs in 3 years time.

Just hard to justify as a non gamer.
For me it's the opposite, I upgrade CPU/MOBO/RAM ideally every 4+ years so spending an extra $100 or even $200 out of the gate amounts to pennies in the long run. If I'm going to be stuck with it for a full generation, then I need to make sure I do it right.

The price of RAM today is a damn shame, though. I'd love to splurge on a DDR4-4000 kit or similar.
 
For me it's the opposite, I upgrade CPU/MOBO/RAM ideally every 4+ years so spending an extra $100 or even $200 out of the gate amounts to pennies in the long run. If I'm going to be stuck with it for a full generation, then I need to make sure I do it right.

The price of RAM today is a damn shame, though. I'd love to splurge on a DDR4-4000 kit or similar.

the time to get in on DDR4 was earlier this year before Ryzen dropped. the demand for good DDR4 just keeping going up and up and up along with its price tag.
 
Is there a DDR4 calculator somewhere that let's me plug in frequency and timings to compare bandwidth?
https://www.dropbox.com/s/7un7fpwg5y30wuj/DDR RAM True Latency Calculator.xlsx?dl=0

i made this so have fun. Now this is basic and doesnt do sub timings which can make a difference i guess where 2 RAMs with the same speed can have difference results due to different sub timings.

This gets the actual latency of the RAM for 1st-2nd-8th word. Raw BW is a different measurement and MTs is a basic way to compare RAM.

Also All RAM uses the same calculation. DDR3/4 are really just categories "in a sense"
 
The price of RAM today is a damn shame, though. I'd love to splurge on a DDR4-4000 kit or similar.

Yup. I've decided I'm just going to keep my slowish DDR4 2400 from my current build for now because RAM is so insane. Haswell-E didn't care about RAM speeds, and anything past 2400 changes the way you have to OC, so I just saved the money and got the 2400. Granted, back then I paid $150 for 32gb....

I figure I'll lose a tiny bit of performance, but I don't want to drop back down to 16gb right now just to get faster RAM.
 
For me it's the opposite, I upgrade CPU/MOBO/RAM ideally every 4+ years so spending an extra $100 or even $200 out of the gate amounts to pennies in the long run. If I'm going to be stuck with it for a full generation, then I need to make sure I do it right.

The price of RAM today is a damn shame, though. I'd love to splurge on a DDR4-4000 kit or similar.

Yeah the RAM is impacting the purchase to be honest. It's $410 AUD for 32GB of only 2666, ugh! What the hell.

That being said I mean I'd love to get an 8700k with DDR4, 4000 or something but the problem is, we're going PCI-e 5, we're going DDR5 kinda in the next 12 to 18 months, we're finally going 8 core and 10nm in the next 12 to 18 months.

Seems to me, I can buy this stuff, save $300 now and put that towards a beast in 2.5 years.

I don't blame you though I guess, be nice to have the stuff at the top of the pack.
 
Yeah the RAM is impacting the purchase to be honest. It's $410 AUD for 32GB of only 2666, ugh! What the hell.

That being said I mean I'd love to get an 8700k with DDR4, 4000 or something but the problem is, we're going PCI-e 5, we're going DDR5 kinda in the next 12 to 18 months, we're finally going 8 core and 10nm in the next 12 to 18 months.

Seems to me, I can buy this stuff, save $300 now and put that towards a beast in 2.5 years.

I don't blame you though I guess, be nice to have the stuff at the top of the pack.
There is probably zero chance of DDR5 and pci-e 5 being available in actual products in just 12 to 18 months. We probably wont even see pci-e 4 for anytime soon and DDR5 is at least 2.5 years away.
 
There is probably zero chance of DDR5 and pci-e 5 being available in actual products in just 12 to 18 months. We probably wont even see pci-e 4 for anytime soon and DDR5 is at least 2.5 years away.

To be fair an 8400 and 32GB might last me 2.5 to 3.5 years :D
 
Yikes, i can't imagine shelling out $400+ for 32 gb of Ram. Curious to know what programs take advantage of the extra RAM. I have yet to use much more than 12 GB, even with vmware and a video editor running.
I am making a budget build now and only ordered it with 8 gB of Ram. :p
 
Yikes, i can't imagine shelling out $400+ for 32 gb of Ram. Curious to know what programs take advantage of the extra RAM. I have yet to use much more than 12 GB, even with vmware and a video editor running.
I am making a budget build now and only ordered it with 8 gB of Ram. :p
Windows will utilize more ram if you have it so it certainly does not go to waste. 16 GB stopped cutting it for me over a year ago as I usually have a ton of tabs open and I was literally running out of memory when playing some games. IMO 16 GB is the minimum for a good gaming pc and 32 GB is nice if you like to leave lots of stuff open while playing games.
 
Yikes, i can't imagine shelling out $400+ for 32 gb of Ram. Curious to know what programs take advantage of the extra RAM. I have yet to use much more than 12 GB, even with vmware and a video editor running.
I am making a budget build now and only ordered it with 8 gB of Ram. :p

Some games already hit 16GB for me. 16GB in my next build is not enough.

With Scorpio its not going to get better for the PC either on a more overall basis.
 
How does Scorpio factor into the mix? It already has more avaiable ram than ps4 pro and more than enough for what the apu can chew though.

There is no reason for pc developers to ride on the coat tails of consoles.

Lastly, just like vram, ram used may not be the same as ram NEEDED. I would really like to see where 'only' 16 gb of ram is bottlenecking performance.
 
You're not wrong on the above points, but for:
I would really like to see where 'only' 16 gb of ram is bottlenecking performance.

This is a delta between game benchmarking and actual usage. While 16GB of RAM shouldn't represent a bottleneck for gaming today, when you factor in everything else people use their computers for that may be loaded simultaneously, 16GB could start looking a little cramped.

I'll admit that such a scenario would likely be a tad excessive, but we're at the point where it can happen for power users who are also gamers.
 
It seems like games started recommending 16 gb only a couple of years ago. 8 gb was recommended eons ago so it will be a long time before we see a 32 gb recommdation.

http://www.pcgamer.com/how-much-ram-do-you-really-need-for-gaming/
PCgamer asks this same question, and although they dont recommend against 32 gb, they say the money can go much further otherwise.

https://www.techspot.com/article/1043-8gb-vs-16gb-ram/page3.html
Techspot goes further and and actually tests some games out.
Although games like GTAV 'ask' for 9 gb or more, it was humorus to see that these games only lost a few frames when lowered all the way to 4 gb of available ram.
 
4D0rL8U.png


And this is why you shouldn't take passmark data seriously.

(1) The graph has changed today.

(2) It seems to show a reduction in AMD marketshare to sub 20 percent levels. Let us revisit this topic in one week when this quarter data self-averages itself.

(3) No media reported the graph yesterday. The media only make noisy news when the graph seemed to show an huge increase in AMD share in July. And no one of the media that published fake information in July retracted or updated the articles, when accurate data was available latter.
 
Last edited:
Is there a DDR4 calculator somewhere that let's me plug in frequency and timings to compare bandwidth?

DDR4 has the same technology than DDR3, the formula is

Bandwidth = Clock * 8 * channels

So dual-channel 2400MHz RAM has maximum bandwidth of 38.4 GB/s.

Total maximum BD is the product of the number of channels by the throughput of each channel.

Each channel has a wide of 64bits (= 8 bytes). So each channel transmits up to 8 bytes in a cycle.

Clocks expressed in MHz gives the Mega cycles per second.

M byte / s = (M cycle / s) * (byte / channel) * channel
 
If Intel Coffee Lake matches or is close to AMD multi-threading per price range, it will also be faster on less threaded type applications making Intel the better buy if costs are similar. Also the over clocking ability maybe looked at as well. Coffee Lake so far looks to be very good and I hope Intel keeps the price reasonable. I think AMD will need to lower prices, like getting the 1700 around $250 which would compete well with the 8600K price point. 1700x $299, 1800x $349.
 
Prices are already official by Intel. 359 for a 8700K. A 1800X needs to be much lower than that.

More slower cores that loses big time when scaling isn't close to linear can in no way expect a price premium. Its FX all over.
 
The 1800x needs to just go away.

Fx as in bulldozer? We are comparing Ryzen to bulldozer now.
This Forum has gone full retard.
 
The 1800x needs to just go away.

Fx as in bulldozer? We are comparing Ryzen to bulldozer now.
This Forum has gone full retard.

Prices if its not a too complicated subject. A 1800X will have to be priced around the area of top FX chips was. And that excludes something that starts with a 3.
 
Eh idk Ryzen 7 will sell well if priced just under 8700k still I think. the 6 and 4 core ryzen cpu's will likely need some price reduction sense they are gonna get slapped around by the 8600k and 8350k esp in gaming and that is where a lot of budget gamers are buying cpus at that price point level.

on the subject of amount of ram, 16Gb is still plenty but certainly there are some games out there that are starting to exceed 10-12 gb use so it is starting to get danger zone for power users/lazy gamers running too much in the background.
 
That's a sweet looking little itx. I think I've settled on the Maximus Hero (non wifi) for my own build. Although, the Asrock Taichi intrigues me. Anyone have any experience with Asrock boards?

Asrock makes great mobos the Tachi is a really popular middle of the road model in their lineup, Only folks i know running Tachi's are on Ryzen but they all love their mobos.
 
Back
Top