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

Won't happen.

 

Maybe someone can get one into a suitcase PC.

 

My 7940X is on its way but I am now undecided on a mobo. I do not feel as good about Asus and their mobos as I once had. My Zenith Extreme (X399) reviewed very well, however, Asus has dropped the ball. When I have my work rig set up and running, I will post information as I did for my Rage Red build.

EDIT: As I stated before in this thread, I will not be delidding my 7940X. I know going into this platform that I will possibly be leaving performance on the table without a delid but I need stability with performance.

 

I should probably take back what I said before (to a degree).
I'm not sure how Intel i9's turbo boost.
What is their turbo boost across ALL cores?

But my point stands. Intel's latest seems marginally better at best and still not worth the cost if you ask me.
Plus, when both Intel and AMD are overclocked to same speeds, they perform either the same or Intel performs just a bit better.
https://arstechnica.co.uk/gadgets/2017/09/intel-core-i9-7960x-review/2/

If you want to OC for the sake of extracting more performance in certain situations (though I really don't know why would people get such a multi-core cpu only to focus on single-threaded tasks which will get done marginally faster), and don't really care about the cost, sure, go with Intel... but if you want to save yourself some money (half the cost), and might actually experience further uplift in performance over time as the industry supports it, AMD is your bet.

 

Its' a good oc'er. Amd TR can't keep up. More like 4.1 or 4.2GHz ?

 

Now, I'll agree to a degree. If you need the extra 10% at a cost of 70% over, then by all means get it, delid it, and get that performance. If you are not delidding, then you see less benefit in Intel over AMD, especially at cost. Here, you have a delid cpu, if I'm correct, or a Hell of a custom loop.

 

But as you said yourself, that's still overclocked (and possibly delid-ed?).
And TR when overclocked to 4GhZ closes the gap between itself and overclocked i9-7960X to about 20%.
The difference is lower if you manage to push TR to 4.2 GhZ.

And TR 1950X costs about $700/£700 less.

So, is the price difference really worth the option to overclock and gain 20%... or none at all if you clock them the same on all cores?
I'd sooner think that people would want to ensure system stability and longevity... which probably means an undervolt on stock clocks with possibly a mild overclock (though, not sure if these high end parts would allow undervolting and overclocking at the same time - they might, X parts on AMD end are usually better binned).

Plus, depending on how Ryzen/TR refresh comes out on 12nm LP, its possible it might clock to 4.5 GhZ and nullify that gap.
Right now, we are comparing a highly refined 14nm++ process for Intel, vs a non-refined 14nm process for AMD - yes I'm aware we have no choice but to compare these products as they are... but still, I think it would be fair to keep that in mind.
One cannot stop to think how impressive is that AMD's giving Intel a hard time with same or better performance on stock... and for 70% less money on a non-refined manuf. process and little or no help from developers in terms of using the uarch properly.

 

Custom water cooling system with an EK Supremacy cooler and a 480 mm radiator hardwareluxx.de
Have both high single and multi threaded performance in one box is nice I wouldn't buy BGA if it was cheaper either.

 

You can't compare TR to bga, that is a ******** statement and you know it! So, for the cost of two extra 480 radiators and a gpu block, plus vrm blocks, you can upgrade to an Intel, before the SL premium for a good chip. Add that in and we are getting close to the cost of my entire water loop. And if talking the 18-core, forget about it.

Sometimes, paying the tax isn't worth it. This may be one of them.

 

I don't compare TR to BGA. Yoo know this. BGA is utter trash and cant't compare it. Put it the other way... I wouldn't save in $700/1000 if I want new tech.

 

As mentioned it is +$1,000 for the CPU, +$300 for a super PSU, +$200 for a super cooling case (480mm rad), +$200 for high end X299, +$300 SL expense. So a minimum of $2,000 extra. I spent $3,800 so it would be over 50% more cost not including the electric bill. Remember too if you live in a hot area you have to cool the house while the system generates all that heat.

 

new laptop out yet? let me know when its out thanks

edit: http://wccftech.com/intel-core-i7-8700k-review/

as expected no IPC improvement just what we would expect from intel. all the performance gain is basically of more cache, higher IMC to feed to cache then feed to CPU. even though memory speed tested is the same, it looks like single core can't process faster than the cache feeding it.

thinking of it similar as samsung's TLC SSD 960 evo, once the SLC cache fills up it gets extremely slow and sluggish and controller/flash whichever is the slowest can't get work done fast enough.

more cache on the cpu would be better for certain workload along with more and faster memory but thats about it. when its all CORE speed, theres no improvement, rather adding more cores adds more latency in ring bus.

 

I don't see it this way... I would still opt for quality psu, MB and Cooling whatever HEDT I look after. And I don't buy new tech every year. A brand new SLI laptop with Coffe will cost a lot as well.

 

I try to at least stagger my upgrades by every other generation (or right now as far as I can possibly make my current system stretch)

 

i9-7980XE & 7960X Review: Delidded Thermals, Power, & Performance

Newegg Studios Live: The New 18-Core CPU in Action with Intel and ASUS

 

I already have a high end PSU and a good WC loop is already planned and budget is allocated. I might as well as get the delidded 7980xe as the only argument not to is price vs performance. I am interested in pure performance and I am willing to committ the budget. Again, not biased towards AMD or Intel. I favor the better product.

 

Yes, but when reducing the skill and comparison, if you get a TR that does 4.1-4.2 achieving 3700+ CB R15, you would have to OC to a speed for a per core return on the 18-core above 4200. Then, you must do the additional license cost for two extra cores, as above 4200 means you are getting more per core, but then have to do the math on wattage draw at that overclock for the TCO calculation. This is in addition to computing the actual expected return on additional work over the period, time under load, etc. If all of that favors the Intel chip, you not only have my blessing, you have my recommendation, so long as that all allows for a ROI higher than the AMD rig after the TCO calculation. There are times where that is needed. But if offloading to graphics cards and each card is a node, allowing you to only use pcie3.0x8 max per card is a hit against Intel, if you have a card that can vendor by bandwidth over 8 lanes. But we've been through this convo before. ;-)

 

We been through this convo before. You know I am buying the CPU for fun and because I really want to play with a HCC chip.

But realistically, if you are using it for a licensing per core program(I am NOT), you have to pull out an excel/google sheet page and start cranking out some math.

Also your GPU argument assume good scaling on multi gpu.

Most importantly: you feeling ok now? Did the med work?

 

The nausea is down and I'm cranking out the work. Read through 200 pages on the final read through of the teacher's manual today. Plan on reading to the half way point, sleep a couple REM cycles, then read about 3 hours worth with a nap in between each 3 hours until finished. Goal is to finish the project by tomorrow, then generate my invoice and move to the next project. But the headache is manageable currently where I can work.

 

IDK, if you look at the various benchmarks and tests for many of them the scaling and higher per core speed of the 7960, actually surpasses the 7980XE CPU results. That 7960 might be a better balance, unless you really need the extra 2c/4t from the 7980XE for specific work.

 

Have fun invoicing!!!

 

I am planning on delidding and overclocking to the edge. Pushing for 4.8-5ghz. I am buying it for the enjoyment for overclocking and having it.

 

(Edited the previous post for naming corrections).

That's my point, the 7960 might OC further per core than the extra core load of the 7980XE.

We are mostly seeing stock speeds so far because various reviewers are worried about their VRM's and cooling standing up to the added power draw, but eventually we will see max OC's for each new CPU, you might want to wait for that before plunking down the green

Who knows the 12c or 14c might actually be the sweet spot for per core OC given the current cooling options and power throughput safely / constantly available.

 

"My favorite part is when they gave me my money." Dr. _____ - The Simpsons

 

With enough cooling, it should be ok. I am planning a minimum of a 480+240mm rad. Worst case, I am looking at expanding to fish tank coolers around 800w.

Enjoy! You sure as hell earned it. Lawyer is very hard work.(Not sarcastic, you guys have brutal hours)

 

“I love deadlines. I love the whooshing noise they make as they go by.” ― Douglas Adams

 

Intel is more power hungry on stock and even when overclocked vs AMD on stock and overclock.
The differential between the two on overclock is more noticeable with Intel (about 30% in it's favor) given it also has 18 cores.

It's your call if you can justify +$1000 more for that much performance (if you are planning on overclocking to begin with) and much more power draw.
On top of that, with Intel, you are forced to replace the entire motherboard with almost every new CPU revision.
AMD sticks to a singular socket... well, in case of TR its' obviously dual-socketed and different than Ryzen, but if you're planning on sticking to such a CPU, then jumping to Ryzen 2 or Ryzen 3 Threadripper equivalent would be a lot easier and cheaper as its probably going to be the same socket configuration, only with probably double the cores (or probably EPYC in 7nm package with Ryzen 2 architectural improvements and manuf. process benefis - that is, if AMD actually releases such a CPU then, and as of yet, such planning/foresight is too far out into the future).

My hypothetical guess is that AMD will probably followup on TR with Ryzen 2 and make an updated version with at least close to what I already mentioned.

But ultimately, its up to you.
I just don't find it justifiable to go with Intel for the price they are demanding vs the performance (even when overclocked - and besides, with such an OC, you need to keep in mind the hardware lifespan... it would be better to undervolt on stock clocks and keep the hardware for as long as it will last - while getting BIOS updates and devs. releasing patches for software to increase performance that way - which is more likely to happen now that AMD forced Intel to release higher core CPU's).

 

The 14c is going to be the sweet spot for per core OC, watch and see. I got some good news for my home workstation (going Xeon), so the 7940X I have coming IS becoming an OC weapon. With a TR and now i9 monster build, I really need to invest in solar panels.

 

http://www.legitreviews.com/intel-core-i9-7980xe-18-core-processor-review_197903/10

I9 7980XE at 4.9ghz with a 4700ish CB15 score. Pretty crazy honestly.

We shall see. I am most likely grabbing the 18 core. We can compare

 

My goal for the 7940X is 4.9 with a delid now that I don't have to worry about my workstation. The good thing is since X299 requires a custom water loop for the crazy OC's, I have everything except for a spare 480 rad. I went a bit overboard when I purchased the loop components for my Rage Red build. We can definitely compare!

 

intel b like tank u com again

 

I wonder how power hungry TR would be be if you could OC much higher with water coooling instead of have to use Ln2. Aka if you could push TR +4075 cb, same as oc’d 16 cores 7960x on water. The last 10-15% Performance increase if this could be manageable will cost. Both in more powa from the wall and higher temp.

 

Nothing is free, best way to find out is take a 7980xe down to stock TR 3,000's and see what the power draw is. probably about the same.

 

Makes me want to contact some OCers to ask their wattage on 5.2-5.4GHz...

 

Thing can be different with 4.5→4.8 vs. lot higher clocks as Ln2 clocks. Same difference with lets say 4.6GHz doesn't mean it will be linear power increase with higher clocks And the coolder you can keep your chips, mean you can hold a bit lower voltage.

 

I do agree. I was more looking at the CB scores being in the proper range. I think 4100-4300 was around 5.2-5.4 and voltage in the 1.5x range. Now, that was before the nuance of TR was learned, not to suggest they do not know what they are doing with TR OC, as they obviously do know. Instead, it refers more to how easy Intel is to OC (fact here, as more people can tune it for good scores, but on both platforms, it takes those truly engaged in OCing to get that extra couple percent at a set multiplier).

 

single threaded software, lemmings, thief gold, total annihilation, reflect macrium, windows 8 in general, somehow ramdisk save/load only uses 1 thread.. and bunch of other things i use from time to time.

multi threaded software i use a lot. steam/newer games, chrome, firefox, adobe flash editor, photoshop, vmware...

we should start a new thread which will have benefits to multi cores or uses new cpu instruction so that its actually worth getting ryzen.

 

Like a "if you do this stuff you should get Intel, if you do this stuff you should get Ryzen" sort of thread? That would definitely be useful.

 

Most users are familiar with if their workflow consists of single or multi-threaded apps. Since Ryzen's inception though this differential has existed.

 

I think a lot make assumptions about it based on historical versions, and may not know if their current version has support.

 

Eight NVMe M2 SSDs in RAID on X399 Threadripper Reach 28 GB/s - Guru3d.com

 

yea make one and find out and keep track maybe lol. more often than not, we can be safe to say multi thread is the future so newer version more likely to have them. then theres also the thing if it actually taking full advantage of cpu rather than dev just simply say, yes they have it but then when you use it for your own scenarios it barely uses cpu at all.

 

That was something else I was thinking about, CPU-heavy vs GPU-heavy applications.

 

pfff gpu heavy. be honest with u gpu heavy seems widely adopted because of the heavy competition. for cpu donno why all these dev mostly on frequency. intel embed doctrine into dev's head.

 

thats a bootable partition? also whats the 4k performance yo

 

There's the odd CAD or rendering program that doesn't seem to have upgraded.

 

@tilleroftheearth From Guru3d.com
New Intel Slides Reveal Product Roadmap up-to 2018
by Hilbert Hagedoorn on: 09/29/2017 08:18 AM | Source | 0 comment(s)

Next week you will see a bunch of reviews on the Coffee lake generation of processors, alongside it the motherboards. in a newly leaked product launch schedule you can see what is coming now and up-to the next half year.

It starts with the 4 and 6-core Coffee lake gen of course with the Z370 chipset. The slides, who have been leaked by website gamersnexus, then reveal H370 Express, B360 Express, and the H310 Express. H370 might be interesting for us consumer pricing wise, it lakcs extreme tweaking options and SLI certification, but would be priced much better. Intel would launch corporate-ready Q370 and Q360 chipsets with enterprise-client features such as vPro, sometime in Q2-2018. Interesting is a mention of two-core procs as well.

Indicated is that in early-November 2017 Intel they could push Pentium and Celeron low-power SoCs based on the "Gemini Lake" platform, consisting of quad-core and dual-core processors, these would have a sub 10W TDP, listed units are Pentium J500S, Celeron J410S, and Celeron J400S.

Surfacing in the slides is also 240 GB and 480 GB sized Optane units. We're really not sure if that is a cache unit or a regular SSD based on Xpoint (we doubt it) after the failed introduction of Optane.

 

if you click the 2nd picture it shows that 8700k previous release date is oct 5th but changed to sept 24th, which today is 29th.. did intel get them out to OEM/retailer asap or what? if thats the case then it aligns what it said in article we'll see review next week, cause 24th would have been 4 business days ago.

also 240gb/480gb optane SSD? that sounds amazing im just hoping its m.2 pcie3.0 x4 lane version, with full 7 channel on its flash. znand product still aint out yet so this is basically the closest thing we've got to ramdisk performance but use it as an actual boot drive.

 

https://cdn.videocardz.com/1/2017/09/Benchmark_04.png

rofl look at that, ST at 4.5ghz is basically within margin of error, possibly due to more cores so higher latency for resources to travel around the ring bus.

also at 7800x vs 8700k, notice how 7800x on x299 is like 177% and 8700k is ~140% compared to 7700k as the base measurement in MT performance. look closely at the table of 8700k vs 7800x in MT and we'll see 8700k literally beats 7800x in every software by a tiny bit, with exception to sandra which is heavily based on memory bandwidth so 7800x quad channel would obviously skew the result.

intel just killed their own 7800x..

 

Yees, as well 7740x. No point in smaller than 10 core if you go for X299.