Ryzen vs i7 (Mainstream); Threadripper vs i9 (HEDT); X299 vs X399/TRX40; Xeon vs Epyc

ajc9988,

Okay, lets move on then.

Here is where I'm coming from (more or less):

See:
https://www.pugetsystems.com/labs/a...aby-Lake-X-Broadwell-E-Kaby-Lake-Ryzen-7-976/

See:
https://www.pugetsystems.com/labs/a...aby-Lake-X-Broadwell-E-Kaby-Lake-Ryzen-7-973/


The above real world results are very closely tied to my bread and butter. Ryzen 7 1800X isn't even in the running vs. anything Intel has offered in the last few years. (Meaning; there is no point to upgrade an older (but not too old) Intel system to a Ryzen 7 1800X platform when running PS or LR).

For those that don't want to click the links above:

To compare and contrast to the Ryzen 7 1800X platforms; they are ~71% slower in LR when importing/exporting (just like ole!!! hints at w/regards to storage performance) vs. an i7-7820X and ~18% slower when converting to DNG/smart previews/image scroll/panorama/HDR workloads are involved in LR vs. the same processor (i.e. both 8C).

With a mere $100 difference between the CPU's (and ignoring the rest of the platform $$$) - AMD is not competitive even many months later in these type of workloads.

In PS, the same processors compared leave AMD with an ~20% deficit in HDR creation, an ~15% deficit in Photomerge and an ~8% deficit when running general actions.

Something to keep in mind with the above: these are not just 'scores' from meaningless BM's. This is a real world test run on actual software with real data.

The two comments below are also interesting:

Question:

Reply from author of article:

See:
https://www.pugetsystems.com/labs/a...orkstation-for-Adobe-Lightroom-CC-2015-8-901/

The above link shows the performance/productivity progress Intel has offered for workflows similar to mine for the last half dozen years. Not a small improvement (as many seem to state, time and time again against Intel... - btw, gaming has long been replaced as a taxing 'workload' on any modern system and using it to compare platforms (except for gaming, of course) isn't in the best interest of anyone really interested in the potential performance/productivity of any particular platform, in any other workload/workflow).


AMD's track record and my experience though would be to have a wait and see attitude until their products actually land at a store near you.

To actually plan to buy/implement an AMD system from all the info we all have today? Not for me. Not by a long shot.

Yet, I do see where AMD's products may be compelling to others - based on the limited info we have from AMD of those products today...

As I keep repeating: AMD - deliver! The benefits (even for me and my Intel based workflows) are already showing up.

 

@hmscott

Sent from my SM-G900P using Tapatalk

 

The issue is Adobe optimization at this point, which is slow and always has been. LR and PS still both perform best on single thread, high speed CPUs. For videos, since graphics cards are used more, you don't see the need to optimize as much for Ryzen or HCC, which is why you get closer performance there. Don't put on AMD the responsibilities of the software developer.

Sent from my SM-G900P using Tapatalk

 

that is correct, adobe has crappy optimization atm for that specific workload with more than 4 cores. generally takes 4 core cpu and that has highest IPC/clock performs better.

when buying into a product, i'd have to make plans and sometimes estimates or guesses. basically, the legacy software that i use needed for single threaded will likely never EVER get an update, hence i choose intel that would benefit for 80% of my usage environment.

my environment may differ from yours, hmscott or tillers so thats why i dont purchase what other purchase here. now i also have to put multi threading into action and make some sacrifices. TR is excellent in terms of value and energy efficiency fosho, but i'd have to sacrifice a lot of my software workload for it, which intel at this point even its power hungry it'll still able to provide me most of what i need, also it'll be for future proof as 8-10c would be good for a few yrs assuming i decide to not get 18c.

theres also other things like storage performance etc. gaming is great but not a must, i can simply afford a more powerful GPU if i want higher fps so quad core high ghz dont matter to me as much as 8-10cores almost as high ghz.

@tilleroftheearth pointed this out many times for making a decision not on speculation but rather benchmark, more realistic usage based, (of course included whats best now and might or might not be good later on) but almost always its neglected that people would pick 1 or 2 things they disagree with in his entire comment and simply say everything is wrong. that imo is faulty thinking and wrong approach to talk about things.

 

Now, read my recommendations a page or two back detailing my picks on HEDT. I base it on what is needed and where I think the compromise will be. For performance, you have to weigh TCO verse time in service compared against potential optimizations (or lack thereof), etc. I still feel the 12-core is best for those that need both faster clocks and core count, but AMD has a competitive product. Evidently, though, the 12 core will have a 2.9 base clock, don't know the boost or overclock yet, so may have to revise moving forward (I think someone mentioned base on the 18-core at 2.7).

But, if what is wrong with the statement undermines other parts of the statement, it can invalidate the whole. Also, when done with enough frequency, it can undermine trust in statements moving forward. Etc.

Sent from my SM-G900P using Tapatalk

 

That's why I always include the video title in text, so I can search for it if it goes private or is taken down for some reason - more frequent than you might think - so I can fix broken video links down the road.

The video has been set to private, so we can't view it, if you remember the title, or the channel, please look it up and see if there a replacement has been uploaded.

What was it?

 

All this discussion on non optimized multi-thread apps. I look at it as with my i7-3940xm I get single core r15 of 138-143. With a Threadripper I should get 173, an over 20% boost. Since with these non multi threaded apps I am happy now, I use LR and Photoshop, I am sure the boost will be appreciated.

 

It was the new pcworld full nerd webisode that focused on AMD exclusively. It was the live stream, so they break it into 3 parts later to show it on YouTube. Threadripper, Ryzen 3, and Vega. It should be back up later today.

Sent from my SM-G900P using Tapatalk

 

no way bro, skylake-x is only around 15-16% faster than ivy in terms of IPC lets not glorify ivy too much lol, of course with 138-143, what kind of clocks you run matters too mate.

 

I run on single thread up too 4.1 GHz. This is mobile so it is not all that fast but fast enough for my use.

 

what the heck, i run mine at 4.7 ghz lol 1.37v got a pretty decent 3940xm here.
if running skylake or kaby at 4.1ghz it'll have a much lower score too like 170.

 

alrighty, things to buy, a skylake-x rig with x299 E-ATX w onboard VGA mobo at 8x8x8x4 if 8c, 18c if i can't get a 5ghz 10c chip.

clevo coffeelake 6c laptop? http://wccftech.com/intel-6-core-coffee-lake-i7-8700k-i5-8600k/ looks like its pulled ahead again ROFL intel can't make up their minds bastard.

 

Until Clevo figures out what it is doing on graphics cards, I'll wait (although that is what I'm on now, a ZM). In fact, I'll likely wait for ice or Zen 2 before upgrading my laptop.

Sent from my SM-G900P using Tapatalk

 

Locked bios, I got too 4.1 as I had a 3820qm and the +400 was activated in UEFI.

 

@ole!!! @TANWare some info on Intel mesh OC

Sent from my SM-G900P using Tapatalk

 

video doesnt make sense tbh. i think its more due to cache rework rather than mesh causing performance issue. with that in mind i'd say its cache causing performance issue which, optimization takes time on software's end. i think intel is smart because their new cache rework is similar to how AMD does theirs, and intel knows software will be optimizing for Ryzen in near future so it'll also improve on intel's architecture, if related, any at all, should benefit intel too.

now once again for gaming, not caring so much. though i do have a tiny bit concern is that if say one were to clock 6c 7800x to 4.5ghz same as 7700k 4.5ghz both cpu all cores overclocked, would they get similar performance? if they do, there is no issue at all. if not just like the video showed, which 4.9ghz 7700k at some games quite a bit better than 4.7ghz 6c then it will be the optimization im talking about.

 

Oh, this wasn't a dig and the take away is, regardless of what others say, here is 3-10%, depending on if stock or overclocked on the core, ripe for the taking. Every optimization counts and this is enough that I'd do it if I was buying one. Because you were, I wanted you to see an easy way to help, even if a little.

Sent from my SM-G900P using Tapatalk

 

AMD Threadripper 1950x 16core vs E5-2679v4 20core Cinebench R15 benchmark

 

3-10% in game. so the real take away is, on ipc and as per tomhardware, pcper, anandtech, we realize theres no ipc difference.

 

Not exactly true. Take for example overclocking the ring. You never got much from it. Always prefer the core to the ring or mesh. But it depends on the program as to whether you see anything. Now, I'm going off of you saying you were delidding and overclocking. This is at least something showing that you can get marginal increases from it. May not be like what was stated on reddit (I don't really read reddit, going off of statements in the video), but it shows something.

 

12-core Intel Core i9 7920X will get a 400 MHz slower base-clock-Guru3d.com

"And that makes things interesting as I immediately want to compare towards Ryzen Threadripper here."

 

The full nerd episode 26 ( @hmscott )

Edit: backup link: http://www.idg.tv/video/78996/the-full-nerd-episode-25

 

https://www.pcgamesn.com/intel/intel-core-i9-7920x-vs-i9-7900x
Just a source on the 2.9 base clock on the 12-core 7920X. So they are selling Xeons as HEDT this round. I don't know that my 100-200MHz by 2 cores will remain true.

 

What a mess. Intel is taken on the bed by Amd.

 

links;
http://www.tomshardware.com/news/intel-core-i9-7920x-amd-threadripper,35038.html

 

let see if clevo will make a new lappy with bigger heatsink and allow us to do 6c 5ghz.

 

Why would Clevo want to go through all that effort to put in a pre-failing 6 core Intel, when they can put in the effort and get a range of Ryzen 3,5,7 + follow on CPU's??

 

why would it pre fail lol? i mean kabylake 7700k is one of the most popular cpu around.. i'd think coffeelake will be popular as well with pricing per core ratio dropped.

 

Man, did they ever dump on Intel... and they didn't even have Intel subjects in the agenda... it's all through the broadcast - ThreadRipper, Ryzen 3, etc

 

I meant it won't compete against Ryzen 7 8 core to start, so it's known to "fail" before it even releases. I wouldn't be surprised if the CL 6 core is higher priced than the Ryzen 8 core even at that point in time.

Ryzen might have new parts by then, 10/12 core on AM4 motherboards instead of requiring TR4 x399, which makes Ryzen AM4 based 4 / 6 / 8 (/ 10 / 12) core possible in laptops.

CL 6 core might as well not even release.

 

cl 6c is most definitely priced higher than ryzen for sure, as it is a faster cpu than R5 1600x or possibly on par with 1700/1800 just because it has a higher overclock. as for mobo chipset, intel will likely continue milk its way through. though cl 6c might not be the deadend* because its just one theres still no 100% way knowing if cNL 10nm will not fit in same socket.

intel might not give us the world until they really realize they are hurt by amd, deeply in their yearly bonus lol.

their strategy right now is to have 6c/6t and 6c/12t replaces older i7/i5 as the new i7/i5 and probably wait to see how their sales go. if sales go bad they'd really have to take drastic measure, right now they are small changes and wait to minimize loss of their "used to" huge profit margin.

 

Yeah, that's what I am saying Intel's CL 6 core 6t/12t is going to be in trouble of being irrelevant for most people.

Ryzen 7/5 will fill that need if not better then at least as well and cheaper, with more cores.

I think CL is pushed out to 1H 2018, there's a lot ground between here and there for AMD to come out with higher count cores for AM4, and maybe a stepped up revision with higher OC.

I think Intel's going to be "back on their heels" all of 2017, and worse in 2018.

 

i twas pulled ahead again to next month lol. first it was 2H 2017 like 2-3yrs ago, then Q1 2018, then Q3 2017, then back to Q1 2018 now back to 2H 2017 rofl.

 

Even so, CL 6 core 6t / 12t is not actually out yet, and the clock of viability for it started ticking 3+ months ago!

Ryzen's here, and it's not waiting for Intel to release CL 6

Intel is panicing true to form, they are rushing the Coffee Lake release, and may have another x299-like disaster waiting for them.

 

Doesn't matter. Will be killed by the low entry oc'd R7 1700/1700X and 1800X. See also post below. Mainstream Intel lying with the back broken!! Higher clocks can't help it from be eaten by Ryzen.

BTW... See Intel Core i7 [email protected] fry at 110C


"A note on VRM heat - during overclocking I pointed my thermal camera at it and the chokes certainly light up, but they always will of course. Power cables however did not show heat levels that worried me. However that 7900X proc is another thing, we halted tweaking as the 1000 USD processor hits 110 Degrees C at relatively moderate settings . We applied an x46 multiplier at all cores and tried the AUTO setting for voltage as well as manual 1.275 and 1.30 Volts. All of them booted fine, the processor simply overheats on the Corsair LCS cooler we used. It even might be a BIOS issue as this was a new yet very dangerous record to see. Hey in good news, we did finish CB at 2500 points though Slowly but steadily I am also feeling that the 4600+ MHz clocks simply are too much for Skylake-X, especially long term tweaks ."

AMD Ryzen 7 1700 @ HWBOT - AMD Ryzen 7 1700X @ HWBOT - AMD Ryzen 7 1800X @ HWBOT - Intel Core i7 7800X @ HWBOT

FYI. The newest 6c/12t Intel i7 at 5.0GHz

 

This is exactly why those "high scores" are useless for the every day user of these CPU's, as noone could run them on a daily basis at the OC, temps, power usage that was needed for those "once in a lifetime" scores.

So they made a high score @ 110c, right before it started to cook itself off, and they had to shut it down.

" Um yeah, we aborted overclocking as at 4600 Mhz / 1.30 Volts the 10 cores reached 110 Degrees C. We did finish one CB15 run though, close to a 2500 CB score."

Big Whoop. ®

The real day to day out of the box scores is what people need to see, as most of them aren't ever going to OC. If the Motherboard and CPU auto boost, that's the best they are going to use, in daily use.

99% of the CPU's used their owners won't waste their valuable time OC'ing.

That's why Ryzen win's, out of the box it's gonna be within a few percentage points of it's best, with an OC not gaining much.

Intel *needs* OC at the same price point as AMD to even have a chance of getting on the board

 

Ryzen 3 Unboxing, Massive GTX 1080 Ti, Expensive X299 Boards!

 

I am not among those 99%. And Ryzen ain't something I want from the OC numbers I already have seen. Maybe Threadripper, if they can OC properly. No in Hell I want run my chips stock clocks 24/7/365. Ain't happen, bruh. Who want...

 

I know, it's the concept of "doing real things" with the PC that elude's you

PC's were made to do creative work, solve problems, improve your life.

Not to be an end all be all unto itself, that's called being reductive, missing out on it's purpose as a tool to do other things.

That's why going around and throwing LGA high power CPU's in the faces of people looking to game or do real life things with their laptops looks so silly to 99% of the people on the receiving end.

Pushing benchmarks in front of their noses doesn't help either.

If someone wants a nice Car for their commute and weekend outings, they don't want the Tank salesman calling on them quoting it's overwhelmingly superior City crushing abilities.

 

AMD Threadripper Round-Up! [IS X299 ALREADY DEAD?!]


WORLD'S WORST X299 BUILD - ONLY 10FPS?!?

 

If you just have to have a 5 GHz machine there are no AMD products for you. At 5 GHz the selection of Intel chips will be dwindling as well.

 

Yup, this one didn't even make a good showing at 4.6ghz...

" Um yeah, we aborted overclocking as at 4600 Mhz / 1.30 Volts the 10 cores reached 110 Degrees C. We did finish one CB15 run though, close to a 2500 CB score."

ASUS ROG STRIX X299-E Gaming review - Overclocking
https://www.guru3d.com/articles_pages/asus_rog_strix_x299_e_gaming_review,25.html

And, now we hear the 12 core is 400mhz down in clock speed... so 4.2ghz OC?

I wonder what the >12 core Intel i9's will clock?

 

sane people would know without delid, they cant expect good temp. if they still want to force their way through a 4.5ghz+ overclock on an average chip without delid they are going to fry their equipment. i must say they are silly to even try that but sure!

 

I agree with Ole!!! Here. But, you get 200-300mhz at most from delid and heat remains an issue. So, you need 4.3 w/o delid and 4.5+ with it, for the product to be worth my prior recommendation. If the 14+ core is lower than 4.2, you just lost the speed advantage and IPC cannot overtake the multi-core advantage.

Sent from my SM-G900P using Tapatalk

 

pitiful intel, they using their biggest real advantage which is ipc and clock speed to beat AMD, but then they are also trying to control the clock speed by using crappy tim so we dont OC. they are afraid if giving consumer too good of a chip with high clock their future node can't keep up and their future stock chips will look like a pos vs say, a 5ghz 10c.

even now, they are trying to give us gimped CPU and still holding back because denial. a good business would immediately see the issue and use their resources for a brand new architecture right away yet they dont do it, poor business practice.

on high core counts, an intel large die could be the cause of high power draw vs AMD just put 2-4 dies together. going on this, it might be better to do 12c+ for AMD because intel's 14-18c probably cant overclock that much due to heat anyway.

 

So, what you are both saying is the Intel i9 is an abject failure, completely failing to provide an out of the box solution for high clock rate OC's, burning down the house instead.

I agree, it's a shameful display by Intel, and the x299 VRM / power failure's even with lower Core count CPU's just add's fuel to the fire, so to speak.

No "sane" person is going to risk killing their $1k++ CPU by delidding, like literally averaging down to the next lowest %, it will be 0% of the people will do a delidding, themselves or as ordered through a service.

0%.

So that means Intel i9 is a 100% failure for OC'ing!!!

Ryzen!! Ryzen!! Ryzen!!

 

naw the cpu itself is pretty good. if you ignore fact of the non-soldered IHS and just cool bare die or replace thermal paste, there are no other chips allow 8-10c 5ghz overclock, even going as far back as sandy-E & ivy-E 6c 5ghz under 1.4v was extremely rare. the quality of silicon on skylake-x is unreal to be capable of under 1.3v for 4.9/5ghz with 8 or 10c and people should not turn a blind eye to it.

 

I don't agree entirely, although delidding is not going to be very prevalent (if I was buying an Intel, I would, but I have a delidded 6700K in the desktop and delid 4790K in my laptop, so I'm in a rare class). But, HT sucks compared to AMD SMT (not sucks, but is lower performance). They have higher IPC, but NOT by much. Because of that, as the clocks get closer, like within 300mhz, the SMT multithreaded advantage overtakes Intel's IPC advantage in multithreaded applications, which becomes more disparate on larger core count chips.

But, Intel has used the icore for a decade now, almost. Software is optimized for it. So Intel still rocks most tasks. That is changing, especially with the coming n-scaling on cores. Now, the SB chips lose to CMT chips. That shows what software optimizations can do. That puts Intel in a precarious place. But that doesn't mean Intel doesn't have a solid product. It just means AMD has a true game plan now.

When 7nm drops, you will have the same logic density (one person calculated GloFo process to be slightly denser under traditional analysis, although on other, newer analyses, produced by Intel, Intel keeps the lead). With that, we could see a large shift toward AMD.

Also, what is known is the same x299 VRM are on the X399 boards. So we have to wait for info on cooling them!

Sent from my SM-G900P using Tapatalk

 

That is the IVR, which reduce it like HW, while making the chip hotter.

Sent from my SM-G900P using Tapatalk

Edit: also, under that logic, you give AMD the win using 0.6-1V reaching large speeds once they get to 7nm. They must have better silicon, because they will be in the 4-5GHz+ range with such low voltage.

 

The compelling part about the new Intel chip is it does appear to overclock well - temps and board issues aside. You can bet overclockers will be delidding that puppy - especially if there's an Rockitcool type kit. Without a delid and a careful eye on the other components heating up they're looking to be a hot mess though. I'd be a little paranoid having an overclocked Intel setup right now. Especially hearing how the power cords were heating up too.

I'm actually quite impressed how AMD are absolutely delivering here to the general masses looking for more cores, and at a sane price. Add to that the chips seem to be really holding their own against Intel and rapidly getting better with different MB and BIOS updates. They've done an awesome job with the resources available to them. Unfortunately it doesn't look like there's much left to squeeze out of them in terms of core overclocking.