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

rofl those 2 cpus are a complete joke intel should be ashamed. spending extra for platform PLUS CPU thats like $700 combined just for a chance to upgrade in future to only more cores but not new architecture, further more only able to access 16 lanes and dual channel memory, all for like 100-200mhz over 7700k honestly no point.

400w might be too unrealistic for my usage. i have like 4 software that uses avx and maybe 1 that uses avx2 that are NOT benchmark, and out of these software only 1 or 2 i use constantly one being avx2 firefox which are not even close to being 100% optimized for multi processing during normal usage, though it can be even tougher at times than XTU memory test and run hotter, only during process restoring of the older version 45 or prior.

i'll prob hit a peak of 350w and not that often unless i do something else while process restoring but during that time my entire CPU is being used by firefox and it LAGS on current 1680v2 which is ivy 8 core at 4.3ghz.

i'd say corsair H100i is probably enough, but i am already spending big bucks on cpu + delid so i'll prob just go for 3x360 rad as its also good to have airflow for VRM too.

 

Reason I'm bringing that up is that even 350W pull is a monster, and they were talking that around 4.8-4.9. So, with that and even delid, it will be pushing it. I have decent results with the TT water 3.0 AIO, able to reach 4.8/4.7 @ 1.43 on my skylake 6700K, bench at 5.1/4.8, but that is over 1.5V, even though temps only ever got to around 82-84C. But, the 7900X has IVR, so you can't do a direct voltage comparison. It would be closer to voltages seen on haswell, which has fivr.

But, if the temps are there or able to work it... The VRM may need a direct fan, as that one video showed. But that is a lot of PSU supply being ate up.

Sent from my SM-G900P using Tapatalk

Edit: I use noctua iPPC 3000 fans in push/pull though.

Edit 2: also, I haven't discussed all bios settings and my stability testing fully here. I did in part in the P770ZM thread and P750ZM thread, but have a slightly different process, more rigorous, for desktop.

 

The thing to be concerned about is that the will be a 20% idle tax on the other i9 CPU's as well, 40% tax at load, as compared to previous generations.

It will get lost in the overall measurements, as we don't have x270 platform numbers for the new i9's for comparison.

It will be nice to know some of those *huge* power load measurements will be inflated with the x299 platform overhead vs. all due to the new i9 CPU's alone.

 

The one point I'd like to focus on as most important is that water cooling, either open or closed loop should be used with the Ryzen CPU's like newer hot Intel CPU's.

I saw one review where they used air cooling, and it wouldn't go past 3.8ghz, and they actually used another air cooler (looked the same or worse than what they had) and were surprised that they didn't have any improvement

They also mentioned that they had been almost exclusively using water cooling via CLC's for 3 years... on the Intel's, why would they switch to air cooling for the higher core count Ryzen, makes no sense to me:

It should be Titled "...Failed due to not using Proper Water-Cooling when OC'ing"

Above all else be realistic in expectations, if it's not going past a certain OC setting, then don't kill yourself because the number you wanted is 200mhz higher, be happy with what you get, and if you have a higher capacity cooling option, use it.

Overclocking Ryzen to 4GHz... I FAILED

 

The second fail wasn't due to temps but to the Ryzen chip-limitation itself.

There's nothing to do but wait for new software/bios releases.

 

My point was don't have unreasonable expectations, if it's not working - it's not working, quit flogging a dead OC.

The lower grade Ryzen's are running at 3.8ghz-3.9ghz, while the lucky few reach 4.0ghz on low grade, while it's standard on the 1800x - pay for it!!

The 4.1ghz is a dream out of reach for all but the very fortunate, or SiliconLottery customers.

All of these OC's are arrived at more reliably with water-cooling, and careful assembly and tuning. Don't throw in your first air-cooled build in 3 years on your first Ryzen OC.

I thought they gave the wrong impression overall.

Ryzen is no more difficult or frustrating than any other build to OC as long as you set your expectations within reason.

None of this is difficult, but some of it is "impossible" or "unreasonable", try not to venture into that realm, and it's all fun and fluffy bunnies.

 

I haven't reviewed it all yet. Date of test? Mb? Firmware version?

Sent from my SM-G900P using Tapatalk

 

Finish watching the video, then follow the video link to his upload, read details, ask questions

 

Just don't have time. Reviewing PP to be included with a teacher's manual. I'll check this evening or tomorrow morning.

Sent from my SM-G900P using Tapatalk

 

To anyone watching any video.

Get Video Speed Controller for Chrome.

It allows you to speed videos up, I usually cruise between 1.30-1.40x - so a 30-40% increase in speed, to get slow talkers through their words faster, and to see youtube videos faster, as they usually tend to extend their material.

 

Might try that. Still, can't do much else while focusing on cross-referencing and explaining securities statutes. Hate to tell someone the wrong statute or regulation required for specific disclosures. I literally have my laptop and a 50" 4k screen littered like my desk used to be long ago...

Sent from my SM-G900P using Tapatalk

Edit: a slice of my life, from the Dodd-Frank Wall Street Reform Act:

"[T]he issuer is required to prepare an accounting restatement due to the material noncompliance of the issuer with any financial reporting requirement under the securities laws, the issuer will recover from any current or former executive officer of the issuer who received incentive-based compensation (including stock options awarded as compensation) during the 3-year period preceding the date on which the issuer is required to prepare an accounting restatement, based on the erroneous data, in excess of what would have been paid to the executive officer under the accounting restatement."

 

well we all know ryzen has poorer silicon quality than intel, ontop of that being first gen from glofo. broadwell silicon quality wasnt even as good as what it is now. basically wohever is getting eypc or TR will have to be prepared of not able to clock past 3.8 or 3.9, assuming heat is not an issue, if one can accept that and wishes for only multi threading performance as well as cheaper option, amd destroys intel here hands down, theres no argument intel simply cant compare on price point.

@ajc9988 honestly i wouldnt worry too much about it. they torture those cpu to highest end which majority of software we use will never reach. heat wont be a problem as long as delid + water cooling and just making sure VRM also gets some air and it'll be fine. people are making big deals out of it much worse than what it seem to be. granted, z270 platform was better and it was explained because bigger surface area and less power/current passes through VRM, with some air flow to VRM on x299 boards and not doing prime95 avx512 that will do 500w system power draw is the way.

 

I wouldn't be so sure on Ryzen being at that, fully, yet. It is a new stepping which may yield slight overclocking improvements, but I wouldn't expect much more than 4.0 all cores (so in line with your thinking, but slightly more optimistic). I'm looking water chiller for it and running gaming VMs off of it on a LAN, as well as media, etc. Home server BoH workhorse. It will only have a 4Gbps connection to start, but that will change with later investments.

But, this was the first GloFo 14. Some have speculated 14nm+ to be the same process, but transistors, etc., done in the style to be used on 7nm Zen 2, following the IBM style. It is still process limited, but could give a good boost.

As to what they do, I do overclocking as a hobby, and have a decent ranking for non-exotic cooling. (HWBot same handle; haven't checked my rank in awhile, though). So, I plan on pummeling it and testing worse case scenario to guarantee 24/7 uptime and no hidden surprises...

Sent from my SM-G900P using Tapatalk

 

i know you wouldnt be so sure. i would be very sure 4ghz will be its limit just like the 1800x. very rarely occasionally you'll get 4.1 or 4.2, TR will be around that area assuming heat doesnt limit it's OC. i do not doubt myself once i believe it .

 

I'm taking a water chiller, which gets Ryzen 7 to 4.3, then backing off 200mhz in the event of heat and any unforeseen issues. An extra 100 if they don't give the MB enough attention (power delivery, as with Intel, but also factoring in a VRM block, which should deliver some clean current), etc. So, my expectations are 4.0 on a water chiller, anything else is gravy! at half the cost of Intel's 16-core, which I'd expect not too much higher (with chiller, 4.5 tops, but pulling like 550 watts+), I'll take it for my purposes.

Sent from my SM-G900P using Tapatalk

My last chiller was damaged and had to be returned, so this is the new one to be used:
http://www.aquatuning.us/water-cool...tan-2000-hc10001650watt-kaelteleistung?c=6569

 

AMD not overclocking to 5.0 GHz causes another unseen issue. that is at 4.0 GHz it draws no where near the power of a 7900x at 5.0 GHz. This no matter what you do with power the chip clocks no higher. Because of this the boards are not really being tested at these higher power demands for extended periods.

This could cause an issue down the line where 14nm+ or other enhancements that allow the chips to clock higher start to show unforeseen power issues on the boards. Though there can be another issue here too.

That is now with the beefier x399 boards may deliver cleaner power allowing for slightly better stable overclocking of the TR chips. Epyc is not overclockable so we have no idea from that right now. Unfortunately the opposite could be true as well if the boards have poor power delivery as well.

 

Well, we have to remember, 40% transistor performance at same power or 60% power reduction at the same speeds. That is just on process. Architectural tweaks with the new Zen cores can also increase IPC. So, being lower power now means that it may do very well at Zen 2 without needing the 550W to get those extra clocks at 7nm.

But, I also predict X399 and Epyc will go straight to 7nm without the 14nm+ refresh. So I wouldn't be nearly as worried on the future power consumption so long as they do like Intel boards. Just look for VRM reviews and hopefully two 8 pin connectors (which they may do because of the extra 6 cores on release).

Time will tell, though. But, if you plan two graphics cards, NICs, your pumps and hard drives (ssd and hdd), and a 16 core Intel maxed, with the graphics cards overclocked, you could max out a 1600W power supply with that rig on Intel. If AMD draws less, even in the 250W range on a 4.0-4.1 overclock, you'll be looking at plenty of room on PSU.

Sent from my SM-G900P using Tapatalk

 

Another issue is that Intels 16 and 18 core variants will most likely no be able to see that high of an overclock. I doubt you will see 5.0 GHz with the added cores heat and power draw. This means at 4.0 GHz with these high core count CPU's there will be less of a performance differential. Eventually I do see though Intel taking the performance crown for now, but maybe by not as much as everyone thinks.

 

Which is why I know my performance will still get lots of points!!!

I'm currently #1465 out of 112419 worldwide on HWBot.org. Really cannot wait for seeing reviews... http://hwbot.org/user/ajc9988/#Points

 

As mentioned several times, something is off with those benchmarks.

 

At the risk of repeating myself, I think these all are early incarnations using stepping 1. I think that early on AMD tried it but found with that stepping there were issues using multiple core clusters on a die. This issue with the infinity fabric is why early on Intel was not worried and on HEDT only planed a 12 core. Even Epyc was not a concern for them. With stepping 2 though the landscape has changed a bit and it sent Intel into a tail spin, not that it was one they could not handle.

 

Could you be so kind to expand on this? I am not sure I completely understand what you mean.

 

Original Ryzen CPU's are all stepping 1. When they first started creating muti-core CPU' (multiple silicon clusters) like the TR and Epyc they would have at first started using these stepping 1 cores. I think they found issues not addressable by AGESA updates with the infinity fabric where addressing those multi-cores was either slow or non existent. Intel probably knew of this issue and therefore was not too worried as they themselves had similar issues with the C2Q CPU's of old.

Once they realized that the CPU's were now working well with stepping 2 it became time to shift gears as they say. Intel has the gears to shift but in the meantime it has forced fiercer competition and further price cutting than originally anticipated.

Edit; a side note is Intel is looking to make the next CPU. If they re-looked at the C2Q doing a high efficiency interconnect like AMD's with 4, or more, 8 core (7820x) CPU's on a die at 5GHz each or more? well forget the Epyc as it is.

 

That's the wet dream Intel has every night, what if we could smoosh together our CPU dies just like AMD? We could win!!!

Instead of lose, badly... "wake's up", NOooooooo!!

 

Again, not saying it is doable just something I am sure they will look into.

 

That isn't how a corporation like Intel thinks. Lol...

But you did paint a funny picture.

I would guess they 'looked into' this a long time ago - and discarded it. Something new is needed to keep up with the rest of their improvements and added features in their timelines...

Again; I have no doubt that Intel has an actual plan going forward. Even if right now they're doing a power slide around the curves AMD seems to be throwing at them.

 

Actually, they already have the bases in the works, and I'm not talking the C2Q. They have the mesh. It isn't the same as IF, but is to lower the overall chip latency for big core count CPUs (originally implemented on the Xeon Phi). Then they have the new embedded chip design allowing for multiple nodes to work on the same chip. This was likely to speed up transitions so non-essential components hard to shrink did not need shrunk. But, all they would need to do is a couple core complexes, that mesh, the embedded portion for the controllers, and bring any advances in MCM design, and they could do it. Nvidia already announced taking an MCM approach using multiple dies in the future on their video cards. Scalability with heat distribution.

I wouldn't rule out a return to it in future chips for the reasons I said above. Instead of two core complexes, just breaking it out to two dies may perform better and increase yields while decreasing time to market. But, I'm sure they will do it in their own way, if done.

 

Agreed, never say never. I just think that if it was feasible for them to do so; we'd have seen examples already.

 

Why show examples to the competition, and why advance, when they had the whole market cornered for a decade.

There's a reason they've accelerated almost all of their R&D outcomes once AMD awoke from their hibernation.

Competition breeds innovation.

 

Not even close.

See:
http://forum.notebookreview.com/thr...s-plasma-vs-lcd-history.806427/#post-10555163


Can't call the competition stalled while your fav is 'hibernating'. Can't have your cake and eat it too.


Yeah; competition does breed innovation. But the opposite isn't necessarily true either.

 

what tanware said, there are more issues than just temperature. ryzen cant overclock higher simply because of other limitation. now if you use LN u might be doing more.

 

give me a new cpu with 50%+ IPC ty intel amd ibm thanks

 

Linking your previous post isn't necessary, instead just quote it here.

Furthermore, comparing the CPU market to the LCD/display market is outright useless, your argument didn't even sail past the port.

Now for the other part of your post. Intel has been holding back if you look at their architecture and roadmap, the launches have been expedited beyond belief.

However, I do agree on this sentence of yours:

But keep in mind, Intel had no reason to do anything, they are already milking the market, all they had to do at that time was to produce better silicon, and yield. That particular ability has since been lost as AMD has caught up to them. LUCKILY Intel has had loads of time working on R&D - so it is not all lost.

Intel had the market HANDS down in their backpocket for a significant amount of time; advancing past quad-core CPU's would never have accelerated had AMD not entered with their newest architecture.

 

That is what I said! That Ryzen is limited even on a water chiller! As to what is achieved on later refinements, like a 7nm built to deliver 5GHz, I'm still not saying overclocking will be crazy, but rather that the base frequencies reached will be much higher. 4.3-4.4 I think is the most I saw with a chiller on HWBot. Here are two links to single stage phase change and cascade phase change scores on it:
http://hwbot.org/submission/3531687_true_monkey_superpi___32m_ryzen_7_1800x_8min_14sec_131ms
http://hwbot.org/submission/3482797_fugger_superpi___32m_ryzen_7_1800x_7min_38sec_313ms

Notice, even with the sub-zero, you are looking at 4.625 and 5GHz, respectively. So, with a water chiller below 20C, but above 0C, looking at 4.3 falls right in place. I already acknowledged the limit of the chip on frequency. Temp control is but one aspect, and often is reached before other limits (as is seen with the Intel chips, especially if not delidded). So, what is your point. I gave the exact reasons why I estimate the frequencies I did, while also mentioning the new stepping. I don't expect any major change from that stepping. If 14nm+ adopts the transistor style used for 7nm (which makes sense, refine the specific transistor on the larger node before the shrink to work out architectural anomalies with the transistor type, then work out the anomalies after the shrink during the period before tapeout and volume production), then you might see a change in frequency and ability to overclock, while admitting and saying expectations should remain low as we do not know it is the transistor type and implementation that is preventing the overclocking as there are so many other variables at play. So, what is your point?

 

There is no doubt they'll all deliver that; how long can you wait?

 

Love it!!!

 

ajc9988, am I understanding you correctly that you will be pairing a ~$900 'chiller' with your AMD build (or is that an £900 'chiller')?

Sorry, if that is true - what is the point of 'saving' with an AMD platform?

 

Intel has the market not because they're 'holding back', but because they give their customers what they needed.

Tech is tech and my statement stands; if you can't make the connection that real advances are made at a pace that is slower than your liking, from the company you want, I can't make you see it.

I agree (once again...) that the launches may have been moved up... but hardly beyond belief. Nobody has a platform waiting to go and sits on it. Nobody. Ever.

Intel already had greater than QC offerings. Uhm...

 

I will be. There is a limit to the overclock, but that also will be used to chill multiple GPUs, VRM, and I'm debating on ram (just a custom air ram cooler with the ram I bought is enough to run 1.5V on and keep it under load at or below 34C on any given DIMM; so it would be a minor change at best and mainly bling factor if I do it). But, this purchase of the water chiller (with the amount saved on the AMD over Intel purchase) is for not just this build, but future builds over the next 5-10 years. Don't get me wrong, it is only because I'm betting long on Zen 2 and 3 that getting on this platform, other than the price/performance which I am not pushing on this because it is HEDT, makes the most sense. I acknowledge I may need to get another MB in late 2018-2019 for DDR5, although if I wait for late 2019 to 2020, revisions may include DDR5 and PCIe 4.0, but that is a ways out. You have to remember, some hardware is reusable with future builds. A water chiller is one of those items. So, it is keeping with my budget to maximize my overall performance over a period of years. Even if I switch back to Intel with Ice Lake-X or Tiger, I then have the benefit of competition further driving down prices by then AND a handy water chiller to use with it. Does that make more sense now? Plus, my graphics card OCing skills need sharpened and a water chiller is a great way to hone it (by removing one of my bottlenecks). Pascal works best under 50 or 55C, at which point it starts throttling. So it is more than just looking to cooling the CPU (especially since that chiller is built for 1650W of heat).

 

Thanks for the additional details of your 'timeline'. I sincerely wish you luck with that long term strategy!

The only thing(s) I know for a fact that I care to upgrade anymore (with any expected ROI) is RAM and Storage on any of my platforms (notebooks or desktops).

If I need more productivity? A second/new(er) platform will always make me more productive than dismantling my current working system and hoping the new will be better/faster/more (or just as) stable. If the older platform is actually phased out? Donating it is again more worthwhile to me than saving a few $$'s in hardware but taking a significant chunk of my time (which = $$$) in lost productivity while I validate the new setup/frankenbuild.

I also don't O/C, RAID(0) or run benchmark after benchmark on my systems. I learned all that I needed to know many years ago. What I learned is that what BM 'scores' showed had no/little correlation to what my productivity levels were. And that is after ignoring the time and effort to do the O/C'ing, RAID and BM's too...

I wouldn't trade that knowledge I gained for anything: but I can understand the need to go through that yourself.

I have a very hard time envisioning any of my systems running with water nearby. I am positive that that is not the future, but I've been wrong before.

Have fun and I hope things work out as you plan.

 

amd 7nm 5ghz sounds like a dream to me, not that its NOT possible. with the way they do thing now looks like they are ignoring that and go straight for ipc design first so they can catch up to intel clock per clock.

1. i certainly dont wanna wait 2 yrs for it lol
2. if they can do it great. let say 4ghz cpu can almost match a 5ghz cpu. i'll take the 4ghz due to any possibility oc will bring more performance per 100mhz and run cooler with less power/voltage. amd cant do that right now but might in the future, let me know when they can and i'll buy it!

 

Exactly. If you look at the lifespan of radiators, etc., you would spend how much to water cool everything, some can go for more than one build, but for the cost over the period, you get more performance from the chiller AND, depending on life span, would cost $90 a year for 10 years, $180 per year for 5 years, not including electricity costs (setting the temp so that it isn't ran constantly and buying a unit a bit over what is needed for the entirety of the rig to increase longevity and decrease time the compressor is going as hard). For the potential extra performance over a custom loop for the period, you are getting a really decent value on it.

But, this is also for a home server and is my hobby. I LOVE tuning new hardware and benching, then coming to the eventual place where I set it for everyday use. I have other systems I can use for work in the interim (plus, my expenditures in upgrading my home network is were the real cost for work is going, even though I will use the server for work as well, it is also a quality of life purchase; networking and security, though, that is work and necessary for my clients). But you are correct in considerations that need factored into how the IT infrastructure deployed for any business, especially small and medium businesses (large ones have the dedicated IT staff with thousands to handle; different world). But, sometimes selling or donating the equipment is the best way to go and just purchasing what you need. But, as I said, this is my hobby, tuning computers.

I will OC, but it has to have extreme vetting (hence why I pointed out how OC3D said that OCCT caught errors that P95 did not). When I'm going for top score, I won't risk destroying the hardware, but I'll push those limits. After I know those bounds, I back it off and do the same things I do for overclocking to test stability for every potential scenario that will come up, settling at what is going to work and keep the hardware safe for years. But, that is also for my home network. For an office environment, you buy what you need and that is guaranteed to work as sold. I upgrade thermal paste and pads after purchase on some devices just for better longevity or cooler operation if the building might experience season temperature fluctuations, if it doesn't come with a service contract, of course. Raid 0 isn't worth it. Raid 1, maybe, but raid 5 is nice, as is other commercial raids.

For BM, I usually just create a temporary ramdisk, put the system on it, and use that. Much better way to ensure that the storage isn't a bottleneck. But, I also use them to tune ram, etc. I get a little extra performance from that, which is nice (increasing 15-30 seconds plus on a SPi32 run does give you awesome increases over the year, but you need access to more than just primary timings to achieve it and validating the ram overclock for use literally takes a week, running P95, memtest86+, windows memtest, etc., for long enough to ensure it isn't kicking errors and is operating at proper temps, etc.).

As I said, this isn't "going through that," it is a hobby. Saying that to a person that loves to tune their car instead of going to a mechanic is no difference. NO ONE SHOULD BE BELITTLED OR DERIDED FOR DOING WHAT THEY LOVE!

Water is but one way they have given better performance with little cost. That is why AIOs have become more prevalent, even in builds for certain systems. It is also why it is being used on graphics more often, including prosumer cards. But, that is but a stop gap and most commercial builds prefer air so that it just works. No worry of pump failure, just fans which are cheap and can be swapped out readily. Point taken there.

I do as well.

This isn't a 2 year wait, it is 1.5 years. That is the same time for Intel to put out a decent 10nm product. LOL! If you look at the GloFo marketing information, they said 3GHz at 14nm for HPC. They hit that on most of the Epyc CPU line (takes more than a process to achieve that, architecture is in play as well). They are advertising 5GHz for 7nm HPC. IBM has made 5.5GHz chips for this segment for a long while. So, if the architecture can take advantage of the process, seeing the HEDT and consumer chips hit that isn't unreasonable in the slightest.

So, with that, 7nm at 40% transistor performance at the same energy consumption would mean 25% would be the jump from 4 to 5GHz. If over 5GHz, then it eats into the 15%, while the rest is attributed to the process. Architecture improvements and adaptations to the process can improve, or at times make worse, the IPC, or stay neutral. So, let's say you take the 25% in performance jump, meaning you took 62.5% of the potential improvement if you kept the energy use the same. That leaves 37.5% of the 65% energy reduction in play. That is a potential 24.375% reduction in power. Now, the voltages used are reduced to 0.6-1V (stated in the literature). So, that means, if everything went perfect, you could see a 25% or more increase in performance, get a 5GHz processor, and reduce the energy needs by almost 25% at the same time. Of course, it still needs executed so is just theory, but that looks like some good potential moving forward!

 

Another example of being misunderstood on the 'net...

I am not belittling or deriding you for doing what you love. Like I said; I went through that phase myself, years ago.

Enjoy what you love - I just thought by sharing something from my perspective would help you understand me/my posts a little better too. Take care.


P.S. In the future, I think it may help if you assume the best of people and not the worst. At least when you're conversing on the 'net.

 

We both did it here and in different ways. The larger emphasis was more for a general statement, but this is the problem with language, it, by its very nature, is imprecise. It is exactly why Socrates refused to write down any of his teachings, he understood it would be misunderstood. Here, we both could not hear the tone of each other, and on rereading my statement, especially broken out like that (I meant to create a new paragraph just because of proper structure and change of the topic being discussed slightly), it reads with WAYYYYY more tone than I meant to convey. On that, I do apologize.

As to assuming the best in people, I wish I could, I've just seen too much. Now, I hope for the best in people, always! But hope and trust are not the same, so I expect the worst. Also, I hope you and ole!!! know that another reason I will interact with you guys in a more challenging manner is that you guys have been around this forum for awhile. I kicked up participation a couple years ago, but have been a member for awhile as well. If it is someone new that I'm afraid it may scare them off of the forum, or that is wanting to learn, I try not to (although admittedly I come off rough with many people, or short, when I'm just trying to get the point across quickly). I challenge you guys as you challenge me because we are all knowledgeable, we all can take it, even if we get frustrated constantly (edit: often from miscommunication or misunderstandings of tone). My work makes me a stickler for detail, which also makes me very scrutinous of what is said. So, try not to take it personally. I apply what I would in how I was trained to analyze in everything. I attack facts and sources like a lawyer on cross(-examination). You look for the bias, the sample size, etc. I'm sure you've seen it a lot with me questioning sources, articles, picking out the points that are weak, etc.

Here, I did read that with the wrong tone, which definitely changed its meaning, then responded without seeing how that response would have come off far more combative than if we spoke in person (and it likely wouldn't have even been said if speaking in person (edit: because of hearing each other and the way it was said)).

So, cheers!

 

Something interesting, I post this here as the benchmark scores also show against competition. There are the Threadripper 1998x ones. CPU Monkey apparently got a hold of an engineering sample and someone got the score from 25 days ago and it was 3141. This is the first link, now if you go to the second current link on their website the score for R15 is now 3416, quite a jump if true.

https://www.reddit.com/r/Amd/comments/6gzgcd/amd_ryzen_threadripper_cinebench_scores_leaked/

http://www.cpu-monkey.com/en/cpu-amd_ryzen_threadripper_1998x-724



Edit, here is a link to i9-7900x at stock clocks it scored 2195, at 100% scaling and the same clocks at stock a 16 core would only do 3512. Just food for thought.

https://hothardware.com/reviews/intel-core-i9-7900x-and-core-i7-7740x-cpu-review?page=6

 

If the scaling is the same, then a TR with same clocks and 10 cores/20 threads would essentially score 2129 on R15 (about 6% performance difference in Intel favor for half the price on AMD side).
But, let's allow for 10% differential if those numbers are accurate... double the cost still doesn't justify going with Intel.

 

Well;


Edit; Throw a little more math, if the Intel skews reduce by 100 MHz and 200 MHz for each 2 cores added.

 

This was posted in the Anandtech forum thread discussing GloFo 7nm process.

Here is linkedin profiles just for reference:<br /> https://www.linkedin.com/in/hyucksoo-yang-04068711<br />- 14nm Process Integration (14LPE / 14LPP / 14LPP+)<br /><br /> https://sg.linkedin.com/in/gyana-sahoo-33294015<br />- Working on Different GF technologies like 65LPe, 55LPe, 40LP, 28 (HPP, LPQ, SLP, LPS), 22FDX, 14(LPP, HPP, HP)<br /><br />14LPP+ and 14HPP are the same node.<br /><br />The Cobalt MOL, SSRW, etc was always planned for 14nm nodes... via the whole 20nm and beyond. (14nm is essentially 20nm @ GlobalFoundries)<br /> https://www.linkedin.com/in/jiehui-shu-jeff-shu-63455b26<br />- Process owner of <i>Cobalt</i> Cap process for 20nm and beyond BEOL EM improvement<br /><br />SSRW is a bit harder to find:<br /> https://www.researchgate.net/public...SSRW_for_20nm_and_below_semiconductor_devices<br /> http://www.rit.edu/kgcoe/eme/sites/...tract_ Challenges in Manufacturing FinFET.pdf<br />^-- patents of this (The present disclosure enjoys industrial applicability in any of various types of highly integrated semiconductor devices for 22 nm technology products and beyond, particularly for 14 nm and 20 nm technology bulk CMOS-based products and beyond.)<br /> http://www.eetimes.com/document.asp?doc_id=1262902 (Instead of 14XM replace it with 14LPP/FX-14)<br /><br />GlobalFoundries and AMD have done this secret node stuff before...<br />-&gt; 28 SHP(Kaveri) / 28 A(Carrio) / 28 HPA(Bristol/Stoney)<br />For example, when the website only details 28SLP/28HPP, etc.<br /><br />If anything the 14HPP node is aimed at the 12nm FinFET node @ TSMC.

Sent from my SM-G900P using Tapatalk

Edit: this is what was said that prompted that explanation:

GlobalFoundries has only access to 14LPe and 14LPP. The Globalfoundries 14nm+ node is HPP. Not to be confused as a better version of 14HP, which is SOI FinFETs. 14HPP which is Bulk FinFETs and is basically 14LPP with 7LP-styled transistors, MOL, etc, and such has higher performance.

 

ASUS Strix X370-F - The Perfect Ryzen Motherboard?

 

wtf that color.. lmao. whats up with those lighting nowadays lol when i use computer i dont even look at it. its not like a good looking laptop rofl