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

Hopefully there will be a point where the single core clock is enough to satisfy legacy software while offering more in the MT area. Devs not optimizing will hopefully solve itself in the long run.

 

This has started with Intel now increasing the core count on i3s, increasing the core count on all mainstream, AMD increasing mainstream core count, and HEDT and Server chips increasing core counts. The software industry will be forced to adapt, like going from dual to quad cores.

Sent from my SM-G900P using Tapatalk

 

Intel didn't bothered sending out any hardware for review in Europe when they launched HEDT.

 

yea, some of the software are really popular and a lot are used but they work like junk, interms of taking advantage of latest advanced hardware. example would be reflect macrium.. has very good UI, easy to use, an excellent tool but poor storage/cpu performance. uses 1 thread almost 99% of the time.

all the people who got 7700k i7 now its like they payed full price but got an i3. starting 8000 series main stream the old i7 is the new i3 lmao.

 

EPYC™ memory bound HPC performance vs Intel Xeon – Extended Edition

 

u know i was hoping amd could just gave us 8 channel of memory in TR just like their server side but i guess that would make half of their server SKUs obsolete.

 

It's possible, lower the core count to 16, with 4 per CCX, raise the clocks to ThreadRipper speeds, and keep the 4 x 2 = 8 channel memory support along with the 4 x 32 PCIE = 128 PCIE, then get a nice single socket motherboard and you've got it.

AMD already has an 8-core ThreadRipper that matches the Ryzen 7 8-core that supports 2 x 2 = 4 channel memory and supports 64 PCIE lanes.

AMD seems willing to meet the demands of users interests, hopefully they will continue to do so

 

http://www.tomshardware.com/news/samsung-flash-memory-summit,35180.html

LOOKS LIKE COMPETITION TO OPTANE FINALLY. if we wait for intel to release 256GB optane m.2 SSD it'll take yrs. hope this will force intel to push them out faster

 

AMD says no current plans to unlock a 1P chip on epyc when asked. But, that same video is where they said they will unlock more features once ready.

There is a Toshiba on a similar design. M.2 but with an advanced controller offering 12GB/s, which may have been misreported as m.2, which is limited to 3.9GB/s by pcie 3.0x4

Sent from my SM-G900P using Tapatalk

 

I've been expecting a new IO socket with far greater throughput, 12GB/sec is a good start.

PCIE x8? xN?

 

3.9 is a hard theoretical limit on throughput for 4 lanes. They basically put 4 nvme pciex4 on one m.2, but to get that to the CPU you would need 16 PCIe lanes (if slightly under, you need 12 lanes). It is like raid, but on the m.2 controller itself, instead of getting a raid card and putting 4 nvme drives on it.

Sent from my SM-G900P using Tapatalk

 

8700k august 21 whew fast release. 6c 5ghz clevo thanks

 

That's the announcement date. Availability is expected in October-November (depending on the country)

 

ewww
well even if it comes in a laptop that will take awhile. clevo usually dont release the laptop till several months after so prob wont expect anything till end of the year or early next yr.

 

Threadripper 1950X vs Core i9 7900X - Gaming & Rendering Benchmarks!

Threadripper 1950X vs i9 7900X Benchmarks! $1000 CPU BATTLE!

 

I'll say this much... the massive size of TR alone gets me very stimulated and makes me want it without even knowing if it performs better. I want that in a laptop and I want it to be bigger and heavier than the P870DM3... -iwannitandiwannitnaayyyoooww

 

It's absurdly potent in multi-threaded applications. In single-threaded, it's akin to Skylake-X at 4Ghz

 

Well, hopefully it can be overclocked like a banshee on the verge of an acid overdose.

 

From what I've seen, AIO cooling gives 4.0-4.1. In custom loop cooling, EK said they have 4.2. Also, a bios update is coming that will reduce the voltage needed to overclock to achieve higher speeds, meaning it may be pushed even further on the all core count OC.

 

OK, well then I should wait and see what happens. 4.2GHz won't cut it. Would have to be at least 4.5GHz at the bare minimum before I would want it, and 4.7GHz or higher would make it a must have. Perhaps I spoke in haste because of how gloriously massive it is. I really love that part.

 

64 PCIe lanes tho.

 

Yes, that's definitely some major icing on the cake, but higher clocks are a must-have. Not interested in stock clocks or baby-step overclocks for CPUs or GPUs. As long as it can do that I don't care if it turns out to be a massive power pig. Actually, that would make it even more exciting if it is a massive power pig at 4.7GHz.

 

You should go to the AMD Ryzen thread and watch the reviews. After seeing the power (and power consumption of the system when OC at 400W) in the different tasks, you may change your mind. You have to remember, Intel's all core boost on the 16 core upcoming chip is 3.6, just like AMDs. So, the all core OC of Intel's, at $1700, may be the same, considering the 10-core OC being around 4.7-4.8 daily driver, with binned chip and proper cooling. I'm starting to think, even with the tweaks they made to the HCC chips, we'll be seeing similar clocks out of Intel. But you have to wait to find that out. http://forum.notebookreview.com/thr...zen-tr-epyc-vega-polaris-gpus.799348/page-275


https://www.pcper.com/reviews/Proce...-1920X-Review/Perf-Dollar-Pricing-Conclusions

 

Exactly. Why rush? Big ain't always same as fun.

See. Ain't fun without LN2 http://hwbot.org/benchmark/cpu_freq...Id=processor_5392&cores=8#start=0#interval=20

 

http://hwbot.org/hardware/processor/ryzen_threadripper_1950x/
You mean like these LN2 scores?

 

My next new setup (desktop) ain't based on LN2. 4.0-4.1 or 4.2GHz sucks!!

 

Yeah, absolutely. 4.0-4.2GHz does suck. Homey don't play dat. That's got turd written all over it even if there are lots of cores and threads to try to make up for being slow.

Cinebench is looking good at 5.0GHz and higher. The score at 4.0GHz is not particularly exciting.

Man, what is the deal with the horrible wPrime 32m scores? That's crazy slow, LOL. http://hwbot.org/submission/3622686_theoverclocker_wprime___32m_ryzen_threadripper_1950x_15sec_6ms

But, wPrime 1024m is looking nice. http://hwbot.org/submission/3622684...___1024m_ryzen_threadripper_1950x_39sec_358ms

 

Intel's Threadripper(16 cores)
http://hwbot.org/hardware/processor/core_i9_7960x/

 

Watch those bar graphs closely, this one I thought was especially amusing:

ThreadRipper sucks in games! (From TechRadar)

https://www.reddit.com/r/Amd/comments/6st8zf/threadripper_sucks_in_games_from_techradar/

Huge difference in bar length, with only a real 1 FPS difference in scoring

" Der_Heavynator 908 points 2 hours ago
At first I thought this was a ****post.... Then I saw it on their site..."

Update: [Editor's Note: the original version of this review included charts that inaccurately represented the benchmark data. This has been now corrected, but the data remains unchanged.]
http://www.techradar.com/reviews/amd-ryzen-threadripper-1950x

 

Yeah, now that's more like it. 7960X kills 1950X in wPrime 32m and 1024m as well as Cinebench.

http://hwbot.org/submission/3622700_splave_wprime___32m_core_i9_7960x_1sec_264ms

http://hwbot.org/submission/3622648_hicookie_wprime___1024m_core_i9_7960x_29sec_499ms

Waiting is always the smartest move. It sucks to spend money on something just because it's new. Bad idea every time. Show me the numbers first (from people that know what they're doing, not 'professional reviewer' numbers) is the best approach in today's tech world filled with marketing hype, lies and deception.

That's a really good example of what I am talking about. Showing me the numbers and skewing them are two different things. And, this worthless crap is from a "professional" LOL. The kiddos sure do like it, though.

 

There's gonna be a lot of "defense" sent up for Intel today, so watch out for that as well.

These are new CPU's, old software - not optimized for AMD ThreadRipper, so there will need to be a lot of scores compared from many sources - throwing out the obviously skewed results before we can get a good read on the current state, and then we watch the improvements over the next few months, just like Ryzen.

Focusing on daily usage apps, tools, games, build environments instead of tiny specialized (ancient) benchmarks is going to be a better mix to help you decide if your usage will benefit from an AMD or Intel platform.

Besides at 2x+ the price, x299 power issues, OC thermal runaway cooling requirements, Intel had better hope some of their CPU's are faster in some things

 

See Wprime 1024M scores vs. clocks +5.3GHz on both Intel and Amd 16 cores. 29 vs 39 sec and not the big difference between clocks.

And yeah, I don't want be GP.

 

What are you talking about?, of course you'd be an Intel GP getting one of their new CPU's.

They both are new processors with new architecture elements, including Intel with their new Mesh Architecture, compared to AMD's Infinity Fabric showing Intel is suffering in performance compared to the previous generation.

Intel better hope for software rebuilds with optimizations too

 

I'm too old for this. Same for hardware and software(updates). I like seeing where my money goes. Haven't too much of them, yoo know

 

I'll just watch for now. It's too early to make any decisions. Not spending any money right now any way. It would be really stupid to spend it if I had it to spend until all of the smoke and mirrors shenanigans have run their course. Not interested in contributing to the furtherance of either brand at the moment. Not being a fanboy of any brand and only caring about results at the end of the day is very liberating... no pressure or reason to care about who wins... just wait and buy whatever turns out to be the winner. I still predict that clock will remain king once all of the smoke has cleared.

But, I'm not sure I agree with the "old software - not optimized for AMD ThreadRipper" part of that comment. Pretty simple time and speed measurement tests (wPrime and Cinebench) and the bottom line is old software still matters. Those benchmarks have been out how many years now? Maybe it would be better to say that they are not the best measurement of ThreadRipper's potential. (I know that's not your intent, only saying that is probably not the best way to say it because it sounds more like an excuse than an explanation.) Cinebench and wPrime obviously were not optimized to be biased against ThreadRipper because they existed long before ThreadRipper was even a concept on paper.

I still like how massive the AMD ThreadRipper CPU is. That's pretty awesome even if it doesn't end up winning the performance crown. Intel needs to do that, too... for no reason other than big is the opposite of small, and that's awesome in and of itself.

 

So - to compare top scores on HWBot ATM:

AMD INTEL PERCENT
wPrime - 1024m - 39sec 358ms 29sec 515ms 133% to intel
Cinebench - R11.5 50.37 points 52.33 points 104% to intel
Cinebench - R15 4514 cb 4931 cb 109% to Intel
HWBOT x265 Benchmark - 1080p 84.97 fps 164.6 fps 193% to Intel
HWBOT x265 Benchmark - 4k 22.43 fps 38.45 fps 171% to Intel
HWBOT Prime 9810.08 pps 14751 pps 150% to Intel
Geekbench3 - Multi Core 79144 points 80192 points 101% to Intel

But, most of this software is also not made for a 2P test, meaning AMD can carry a deficit, as less difference is seen in certain other tasks for actual workloads. But, it should be noted Intel will likely keep the crown. But that is what we have on benches for HWBot ATM...

 

We can only use what we have for measuring tools. And, only an Intel fanboy/shill would refuse to give credit to AMD for breaking their trend of only selling worthless garbage CPUs. I am proud of them for doing this even if I don't end up actually wanting one after all the smoke has cleared. Intel needs to sweat and worry, and I hope they have reason to continue to sweat and worry for a very long, and have to work harder and harder to stay on top. I am very glad their unhealthy monopoly has finally faced a serious threat. Way past due!!

 

Intel is doing that on FPGA, and will likely do the monster socket on the Ice Lake-X platform, as the 2011/2066 pin LGA was not designed to handle that many cores with the power nearly as well as a larger socket (in other words, it was coming, just later).

As to the software, all software has been optimized for icore, and not much has changed from intel in a decade. So software will work better with it with zero optimization than AMD. Also, the TR is closer to a 2P server setup on the same chip, especially in regards to memory and cache use and latency, apparent in these benchmarks. Even Intel's mesh gives it a hit over the ring bus. Because of this, neither platform is as good as it will be until software engineers fully scale for both companies.

 

Why not put your sentiment in the form of an AMD purchase, make it count for something instead of turning around and spending 2x as much to reward Intel

It's one thing to say thanks to AMD for doing this verbally, or in a post, if you plan to get a new desktop then help AMD continue the fight by buying their stuff, tuning and optimizing it for better performance - and give feedback to AMD to help improve the product.

Even if the Intel CPU at OC is boiling over the water cooling and providing higher benchmarks, unless you get noticeable bankable improvements in day to day operations it's not worth feeding Intel with more money.

Now AMD has given us a chance to feed them what they need to keep improving.

 

I don't put + $5K for charity. I don't buy computers every 1-2 years.

 

Please stop phrasing it as "giving charity", you know it's not charity - please stop being silly.

ThreadRipper and Ryzen / Epyc are products that perform better than or close enough to be a complete replacement for Intel.

Intel costs 2x+ as much and has enough other problems so as to not recommend it anyway - Intel x299 is problematic with it's own problems to resolve.

My point is we don't just sit there and cheer on AMD for sticking it to Intel causing Intel to start doing better, instead take action and reward AMD - buy their stuff instead of Intel!!

At this point buying x299 is more charitable to Intel than buying x399 + ThreadRipper is being charitable to AMD, with AMD you're getting more exciting new stuff!!

And, don't forget to buy some Vega stuff too!!

 

The truth is, the gap will likely be completely gone with 7nm, which you can use with the same AMD MB. Same is not true of Intel, which will require a new board for Ice lake. Cannon-X won't be much but incremental by comparison, AMD will have the same logic density as Intel (7nm AMD, Intel 10nm+), roughly, and potentially much higher clocks. But, that gets into the argument of buying for now versus buying for late 2018 to 2019 for a CPU upgrade without knowing the resale market yet. Also, only if AMD gets enough adoption will companies be more willing to optimize for the change (or if Intel abandons the monolith).

 

If and when I buy something. I will go for best performance. Same as I went for P870. Overclocking will be a big part of it. Offer me what I want... I buy it. Computer is a hobby for me, same as well a working tool.

 

On charity, it depends. If you are buying in at a certain price point (I only wanted to spend $1K-1,200 on the CPU), then the decision becomes easier, depending on your purpose and task. If you are buying at core level/performance king, then Intel is the choice. So, it isn't charity if you buy at a price point, it is if you are buying at core count or pure performance ATM.

But I do think @hmscott got a bit advertisey there for a second. (edit: the vega add on)

edit: I just cannot justify 70% price premium on a 1-10% performance jump in my main use scenarios. Meanwhile, other software needs optimized, and until it is, there are times you may get the much higher performance jump. But, the benches cited above that show 70-100% performance jump are not real world use (but great for numbers).

 

Well, at least the Devil knows how to buy your soul, just a .001 improvement in a benchmark, and you're in

Really though, stop giving Intel money for being a little faster in some things, it's sad really, not cool.

 

See latest words in my previous reply.

With Phase cooling as before I will be a happy man with high clocks

 

I'll be happy on water and cruising with, eventually, 4 GPUs. Now the software companies need to step it up...

Edit: I'd like to eventually, if they scale well enough, do 8K downsampling to 4K on gaming, when not used for its intended purpose.

 

It's charity if you are overpaying to give them more money that it's worth.

If it's a fair price it's a straight deal. AMD is a fair price for performance, enough for my use and it's a nice vote against Intel and for AMD.

Overpaying Intel is a rip-off, I wouldn't dignify it by calling it charity - but it is overpaying for what you are getting.

Advertisey? I shouldn't need to be pointing out what should be obvious and acted on without being "advertisey", but unfortunately without it people often will just plod along in their rut not realizing there are other options.

Continuing on buying Intel is a rut. Buying AMD is getting out of that rut and into another world of opportunites - either purchase is a risk with them both being new - AMD is a less costly risk.

I can't think of a good reason Intel has given us to buy from them instead of AMD, not one.

Can anyone else think of a reason to not buy AMD and to fund Intel once again?

 

Phase change would be fun for either one, and I look forward to seeing your results

 

Mind you, I don't give No one my money every year. Seee... Dell doesn't get a dime from me anymore!!!