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

This isn't trying to talk you out of it, but two points:

1) 8 core optimization has gained A LOT for both sides since Ryzen and should only get better.

2) Prema is taking a break right now because of Jack offs.

But, you've made your choice which is why I stopped discussing the virtues of AMD and have moved to asking your setup.

Sent from my SM-G900P using Tapatalk

 

Jagoff...

 

ole!!!, agree 100% with your quote below. Even as I wish AMD success/profit so that they can continue to prod the giant and us mere consumers of both company's benefit greatly too.




When highest real world - not 'scores' - performance is needed 'today' and in the near future (~18 months to 2 years) and absolute cost (i.e. saving a few dollars a day over the expected life expectancy of the system) doesn't matter - and a slightly higher power bill isn't a deterrent too;

And for anything less than a server class workload (i.e. anything up to or below an workstation class workload)...; Intel is still firmly in the lead.

You buy hardware/software for the workloads you have now and for the expected life cycle of the hardware. Not based on industry wide expectations of software developers working tirelessly 24/7 in the next few months to offer something to match the new (HCC) hardware we now have access to.

Yeah; those developers will work as hard as they can and those benefits will be coming too. But the current platforms (I predict all of them) will also be obsolete by then...

In the meantime; if you are being paid for the computer work you do - or; if you simply want to spend as little time as possible doing it (paid or not); then settling for lower (real world) performance (in most consumer workloads) does not sound like a reasonable stance.

No matter what it points to in the medium/far future...

Right 'now', is all we have (ever).

And, if you sacrifice today's performance for tomorrow's promises; more than likely, life, time and hindsight will prove us wrong in the 'meantime'...

Buy the performance you know for 'sure' and hope the (performance) promises don't remain 'wishes' forever. After all. This isn't the last platform we'll ever buy.

 

So, couple points. First, needs vary. Depending on the need, AMD already has the optimizations (although I still don't recommend the mainstream Ryzen if what you need is HEDT performance, which is also why I don't talk about the 6 and 8 core Intel's as much, because they are headed into the mainstream market quickly-- literally 6 months or so).

I agree to a degree. For home market, power isn't a deterrent, in most cases. In servers, that adds up quickly if you have multiple racks or an entire floor.

For the anything less than server class workload, I disagree. Single core - yes, but that may change on the higher core count chips (14-18; 12 still being like medium core count, but looking to be a drop of at least 200+MHz, starting to wash away the single core advantage). But, many do not need 12+. I'll still get the 16-core, but I'm a specific type of person.

Now, for the expected lifetime, AMD will see more and more optimizations. If your life cycle is only two cycles, then you have a point. But, because people buy for longer (think of those still on SB), you may look for the platform where the socket can support the future CPUs, which makes Zen more compelling. You keep saying betting on promises, but these are more than simple promises (and in that regard, Intel's been over-promising and under-delivering for awhile now).

Also, Intel doesn't have any HCC out yet. Period. Many of the optimizations made for Ryzen benefit the Intel chips with HCC as well, just to a lesser degree. So, those tireless months translate to working to optimize both platforms for better multi-core scaling, otherwise both would get horrible performance above a certain core count.

Once again, real world performance depends on task and INTEL DOES NOT WIN IN ALL TASKS. We wouldn't be discussing this if they did, or if the performance difference was starker.

Also, what you argue for is instant gratification versus delayed rewards. We all know you have to do the latter A LOT in life.

 

Hardware unboxed:
AMD Ryzen 5 1600 vs. Intel Core i7-7800X: 30 Game Battle!




Sent from my SM-G900P using Tapatalk

Edit: it took a custom loop to hit 4.7GHz on the 7800X, whereas a coolermaster was used with Ryzen 5. So, no AIO to hit that speed.

 

@ajc9988 had AMD glofo provide better silicon and chips allow say 4.6-4.7ghz OC i'd have probably go for AMD. the performance between intel and AMD at that point isnt all that much different.

in the case what @tilleroftheearth has mentioned above, its more to do with what we have now and what i'll be getting. companies do compromise a lot so waiting on optimization isnt exactly what i wanted. dont get me wrong 16c at $1000 plus water cooler is great, an excellent value at the performance almost the same as intel at that core count.

 

Review the hardware unboxed video comparing the Intel 7800X and 1600 in gaming, with the 7800X at 4.7 vs 1600 at 4.0. It is a wash in gaming, varying by game. So check the list and see what your needs are, but your statement is incoherent as it stands on actual performance! (Don't confuse this with trying to convince you, it is only correcting false statements of fact)

Also, Intel's performance at that core count is unknown, so stop blowing BS up our butts. From what is known, Intel may have lower performance on their 16 core overclocked. So just stop with false statements and unknowns.



Sent from my SM-G900P using Tapatalk

 

the 12-core i9 base clock is confirmed at 2.9Ghz vs 3.4(?) on TR4?

 

The 12-core 1920X is 3.5GHz base. The 7900X (10-core) has a base clock of 3.3. So, the base clock on the 12-core dropping 400MHz means we could see more than 200MHz lost on the OC. Now, the 18-core has a rumored base of 2.7, but that doesn't mean the drop in OC isn't there, just that they clocked the 18-core closer to the limit.

Sent from my SM-G900P using Tapatalk

 

If I may, even with 4.0 GhZ OC on Ryzen, by how much does AMD fall behind?
Intel can perhaps reach higher clocks (and some of its CPU's ARE clocked higher at 4.5 GhZ), but how much of a difference does that produce really? Extra 5 - 10%?
The IPC difference doesn't seem to exist in any drastic account really between Ryzen and Intel's Skylake when you take into account that once software is optimized for Ryzen the overall differences are maybe a few % in Intel favor (in some cases they don't even seem to exist)... and even then, Intel needs to be clocked higher to produce another 5% maybe 10% difference at much higher power expenditure.

Nope... not worth it to go with Intel in my case at all. Intel is way too expensive and doesn't offer anything that AMD doesn't have performance wise beyond a few %, which in the long run don't matter all that much really.
Even if you're doing 3d rendering, are you really going to go for less cores and higher clocked cores for the same price?
Hardly. The sensible idea is to go for higher cores with AMD, OC and spend less money in the process while practically negating any performance advantage on Intel end.

Ryzen already has plenty of optimizations behind it and more optimizations ARE incoming.
AMD 'finally' managed to stir up developers so they begin writing their software for higher cores and optimizing instruction sets for them as well.

People are basically betting on existing Intel optimizations that the industry produced. Looking long term, and considering how the system is based on cost efficiency... AMD simply trounces Intel in price, efficiency and puts up a fight in performance both at stock and OC.

 

We do not know the turbo clocks, but yes this does not look good for Intel. The problem here is we were all assuming only 100-200 MHz lower clocks per added 2 cores from the 7900x. What we may be getting is even lower than that.

This is for whatever reason, be it they just can not get there to the fact of not wanting to compete with their own Xeon's. The problem in the end is even at 18 cores a lowly 3.3 GHz at Turbo chip will have trouble with a TR 16 core clocked at 4 GHz. This and the fact the TR will be out for months ahead of the now unsure Intel offering could be devastating for them.

 

Еxactly - same with Coffee Lake - the 6-core 12-thread i7 will come out 6-8 months after the Ryzen 5 1600 and, by the looks of things, offer only 10-20% more performance at a higher price tag. I mean, it will be competitive with Ryzen, sure (unlike current 4-core variants) but will the extra horse power justify the price gap? Not to mention Ryzen has 4 more PCIe lanes for expansion and such.

 

Plus, let's not forget that AMD managed to spur creation of Asus ROG Stryx Ryzen laptops with 6 and 8 cores and rx 580.
It was AMD's doing after all, and infinity fabric is apparently a cheap way to increase computational power of a conventional multi-core CPU with smaller nodes while using silicon.
In all honesty, AMD should really diverge some focus towards using composites of carbon such as graphene, carbon nanotubes and synthetic diamonds as a baseline material for after Ryzen.

Infinity fabric already underwent some improvements with B2 stepping if I'm not mistaken to reduce latency times.
A monolithic approach as Intel uses it doesn't seem to work as well - yes its doable and possibly it can be optimized better for increased efficiency, but doesn't seem to offer the same ease of assembly like IF (let alone the costs associated with a monolithic approach).

I wonder what Intel's response will be. I mean, are they continuing with the monolithic approach or are they going to implement something similar to infinity fabric (doesn't seem likely considering they described it as 'glue' - but they might not have any choice going forward - and If they do introduce something akin to IF, they'll simply call it differently).

Makes you wonder what might happen if AMD, Intel and Nvidia decided to forgo competition and just decide to freely share their ideas between each other to create far superior hardware for everyone and keep innovating (real innovations, not minor increments) for the sake of innovation, while offering recycling programs to harvest old technology/hardware for raw materials so new ones could be synthesized and preserve/protect/repair environment (oh wait... that's unlikely to happen in the current socio-system - the best you could hope for is maybe support from all 3 companies involving same open source options such as Vulkan and DX12).

 

not much behind, 4ghz is good enough for most people really. but coming from a 4.4ghz 8c ivy going to a 4ghz 8c with slightly better ipc and power consumption, not my thing really. im all about cpu and storage performance upgrading for almost no improvement is a no go.

let say, 10c at 4.8ghz for a decent chip
12c prob around 4.6ghz
14c 4.4ghz
16c 4.2ghz
18c 4.0ghz

if scaling is similar to that from 10c onward then at around 16/18c no point going for intel. unless one would want to take advantage of turbo boost max 3.0 or other features i dont see any reason going intel for that many cores.

 

If you want Intel you should probably wait till the next cycle of HEDT, as it's a mess.

Ryzen 7 might not be a bump up, but the 10c/12t ThreadRipper 1920x and above should be a nice jump up.

If you really want fun for home, get Epyc

You're dreaming with these numbers. Already the 10c is overheating when people get to 4.6ghz.

My estimate for safe daily all core OC, by your list:

10c 4.3ghz
12c 4.1ghz
14c 3.9ghz
16c 3.7ghz
18c 3.5ghz

 

im talking about top chips though, silicon lottery ones. i have no interest in getting crappy chips aint my style

 

I think more like Lottery delided ones being;

10c= 4.8 GHz
12c = 4.5 GHz
14c = 4.2 GHz
16c = 3.9 GHz
18c = 3.6 GHz

Now with improvements to x299 and maybe process of the chips this may get higher, but where I see it now.

 

intel needs to use solder for higher end 14-18c chips or ryzen destroy them

 

If you already have 8c/16th ivy bridge, there would be little point jumping to ryzen 1700 or 1800x.
The better option would likely be threadripper at 16c/32th as it seems that it will turbo boost higher than current intel i9.

Alternative is to wait for 7nm+ on AMD side early next year and see where it ends up in terms of performance etc.
I have to agree with hmscott that the current i9 lineup is best avoided and if you want Intel that badly (although I really can't understand why you'd want to opt for a more expensive option that would give maybe a few % performance more) just wait until they release their next hedt lineup.
Otherwise, Threadripper 16c/32th or Epyc seem like better options.

 

thats why i gotta go for highest ipc cpu with highest frequency possible. 8-10c 4.8-5ghz seems a decent jump really. threadripper only gives me the jump in multithreading but not single threads you see.

 

A 12-core @ 4.8-5GHz from Intel currently is, simply put, a fantasy. The VRM modules will melt when you start pulling 4-500W. Sorry but 10% more performance is not in any way worth 300% more power draw.

 

Not in single threaded. Ryzen and TR are what, maybe 5% behind intel in IPC?
Also if you're shooting for 4.8 and 5 ghz on Intel, I think you might be overshooting really. Sustaining those clocks without burning the cpu will be next to impossible if Intel current lineup is any indication. Also, you are assuming that Amd won't be able to clock that high with Ryzen 7nm+ or Ryzen2.

Well, it's your decision

 

@ole!!! if you are at 4.4 GHz on an 8c IVY then even 4.8 GHz in Sandy-x will not yield anything of significance. Only if you are thread limited and a 16c TR would benefit you should you upgrade. TBH if I were at your stage I would not even think of an upgrade as I would not get enough extra from a system to justify it.

 

Agreed with TanWare. That's why I said that if he's already using 8c/16th ivy bridge CPU, jumping to ryzen equivalent wouldn't be much of an upgrade. But jumping to Threadripper 16c/32th would probably be of great benefit, and it's B2 stepping offering great performance for much lower cost.
It's either that or wait for more efficient Intel hardware or Ryzen 2, but the current iteration of Intel simply isn't worth extra 10% performance at far higher power draw.

 

So, it has to be remembered that Silicon Lottery is delidding, putting liquid metal, then resealing them for customers.

"This CPU includes our delidding service! Temperatures under an overclocked load (1.2V-1.3V) typically decrease anywhere from 10°C to 20°C when delidded, ensuring greater stability and a longer component lifespan.

• IHS is removed from the CPU.
• Components near the CPU die are coated with liquid electric tape.
• Stock thermal paste is replaced with Thermal Grizzly Conductonaut.
• IHS is sealed back into place, so the CPU can be treated just as if it were stock."

So, an average person purchasing cannot expect those speeds, but Lottery ones will. I don't see much improvement on the daily driver side.

Your timeline is wrong. H1 2018 is 14nm+ and not all lines are on it. Then 7nm is Q3 2018 at the earliest (Q4 2018-H1 2019 is more likely, depending on the product). The 7nm+ is not likely to make an appearance until Q4 2019 and throughout 2020. Intel's Ice Lake will not appear until 2H 2018 into Q1 2019, starting with the Server chips (and likely HEDT) and is on 10nm+. You then wait for 10nm++ to battle with 7nm++. So, it is actually waiting for late next year to early the year after that for a decent Intel product, but 7nm Zen 2 will likely have equal to higher density on logic and have a speed boost, meaning it could be the victor in this situation.

That makes no sense. Higher speed does not mean higher IPC. Look at the benches for the 6-core. With current optimizations, half show Ryzen 5 winning, half show Intel, all with a 700MHz spread between the two. Only if you have more programs directly benefiting from the single core speed and IPC does it get close to being a choice. If you use newer games or the programs will optimize for multithreaded soon, only if you plan to upgrade soon after those upgrades does it make sense. But you can do your own calculations on that. It is just humorous that it takes 700MHz higher on Intel to only win half of the fight on the 6-core. Think long and hard about the ramifications of that.

Edit: didn't word that last part correctly. The IPC advantage is completely wiped out in anything that can use multithreaded goodness is my point, so much so that you need a custom loop and 700MHz to overcome it on Ryzen. Even if that gets slightly narrower (like to 500MHz because of two dies on TR), it means that a lower clocked 1920X could be equivalent in most cases (if it overclocks to 4GHz) to a 7920X running at 4.5GHz. It also means the 16C 1950X will beat a 16C Intel chip if the 1950X hits 4GHz all cores and Intel's does 4.3GHz or less. It is a bit telling, but these are still guesstimates.

 

Shamelessly borrowed from reddit ( source)

 

you are assuming things again yo. higher speed doesnt mean higher ipc. but in this case intel has both higher speed and higher ipc so which is why im going intel. donno how many times i have to keep telling you that lmao.

 

somehow i said 8-10c, but it turns out to be 12c at 5ghz. you have conveniently added 2 more cores

skylake-x 8c at 5ghz would be around 30% faster than ivy 8c at 4.4ghz. thats a big jump honestly in terms of cpu performance. further more the new chipset allows other benefits like usb 3.1, more usb 3.0, speed shift, turbo boost max 3.0, NVMe storage etc. much more than just benefit from CPU.

i have waited forever for DMI 2.0 to change to DMI 3.0 along with a CPU that comes with 6c or more. coffeelake is first mainstream that satisfy this, that or i go skylake-x

 

Not that I mean to burst your bubble, but sites such as Guru3D and TechSpot (Hardware Unboxed) have already presented evidence that Intel's new design method has drastically reduced IPC performance. To the point where at the same clockspeed (in this case, I believe they tested at 4.2GHz) the R7 CPUs performed better than Skylake-X in Single-threaded tests

Here is a summary.

 

Refer to post above - Skylake-X IPC is actually lower due to the newer Intel design that is replacing the older ringbus-style.

 

Don't know how many times I have to tell you this, it is specific to software and as has been shown the higher speed and IPC is beat half the time with the lower clocked chip, meaning you only benefit if the majority of what is used is in the specific half that loses.

Sent from my SM-G900P using Tapatalk

 

It may be that with Intel deprecating their speed advantage starting with 12c CPU's, and assuming it's going to be the same or worse as the core count increases, that AMD will remain steady at their speeds vs core count.

Look at the Xeon line parts enlisted in the i9 stable, they all run at slower than "enthusiast" speeds.

Also, those Xeon parts typically don't allow overclocking, and IDK if Intel should unlock them for i9 - or if it will do any good.

I'd wait until the whole line is out, rev's in MB's Power Sections can be made, and let 6 months of MB ECU's and BIOS updates go by before jumping into it. Maybe even a subsequent stepping of the CPU's.

Again, you might as well let Intel get their chairs set up and let the orchestra run through a few numbers before purchasing a season performance package.

Or, wait until the next generation, probably better.

Hey, why don't you get a ThreadRipper in the meantime, pull up a chair and sing along with the "Intel sucks" quartet for a while, and you might be happier than ever

 

Quite correct that my timeline is wrong.
Q1 2018 is 14nm+ with regular Zen, and Zen 2 probably comes in late 2018 at 7nm.
What I meant to say was that 14nm+ will likely give AMD current Zen lineup a clock boost at same or lower power draw with B2 stepping (possibly) and Ole could go for that.
However, in all probability, and if I was in his position, I wouldn't swap out an 8c/16th cpu (even an Ivy Bridge) for another 8c/16th cpu as the differences would be too small to justify performance wise.

I'd rather go with Threadripper 16c/32th at 4.0 GhZ which will perform admirably in multi-core software and decently in games (comparable to current Zen lineup at 4.0 GhZ really - and that's no slouch considering that optimizations now reduced performance differences between Intel and AMD to nothing with IPC going to AMD favor - at this point, if I was in Ole's position, I'd settle with 4.0 GhZ at a nice/efficient power draw with 16c/32th because its a lot better deal compared to Intel in performance, efficiency and power draw) and then see what happens with Zen 2, Zen 3 and Intel's responses.

 

rather vague and incorrect, but i'll let u look into it.

certainly speed is lower compares to say vs 4c at 5.2ghz. well, theres always turbo boost max 3.0 to boost 2c to say possibly over 5ghz or higher which is yet another premium feature AMD doesnt have at least for now.

how turbo boost max 3.0 works and how it benefits me? i donno because im gonna be the one to test it.

 

Intel is cannibalizing their own products after a K.O. from Ryzen.

Shamelessly borrowed from reddit, again ( source)

 

Look, diminishing and dismissing doesn't mean the statement is wrong. In fact, because you have shown zero facts nor a retort, it shows you lack the understanding to respond.

 

Thanks for being the GP in the Intel i9 disaster, I look forward to your reports on details we've discussed... the only problem is, who's gonna buy it when you want to sell it?

 

No one after Ryzen numbers come out!

 

Those who started with 7740X The main reason for this <optimized> 7700K Ala meant to be the first step into HEDT

 

https://www.facebook.com/video.php?v=10156463662414942

 

Yeah, the CPU might have a place to go if he breaks up the build. But, who is gonna want the 1st generation x299 motherboard? And, what else can he do with the rig? The 2nd generation CPU's will likely need a 2066v2 motherboard.

He'd need to find someone willing to take on the whole mess. Maybe if he sells it for less than an AMD matching build would cost.

 

I'm sure he will getting rid of it. Mind you... People buy BGA as well And we have Max-Q!!

 

yea about 5-7% behind in terms of ipc at the same frequency and intel can hit possibly 4.8-5ghz depending on the chip where as ryzen is stuck at 4 ish.

 

well thats what you'd think and you can keep thinking that lul.

people will buy based on what their needs, looking at overclock.net, quite a bit of people getting skylake-x. intel isnt as weak as people making them out to be.

for those who bought 7740x LMAO, well each to their own LOL

 

IPC varies by program and task. Considering the slower running chips, granted in games in varying degrees of multithreading, can now perform the same as the Intel chips heavily overclocked, it says those programs that have been optimized are able to get a higher IPC out of Ryzen. The fact you still don't get IPC varies is a bit disturbing...

Sent from my SM-G900P using Tapatalk

 

i'd disagree with you. now since you have made the claim, you can provide your results and evidence or you can choose not to, honestly i dont care lol. i'll believe what i've learned and what i know, and what i will purchase will benefit on my usage scenarios.

 

We've already provided multiple sources but for some reason you're ignoring them all even though Guru3D have basically proven that Skylake-X's IPC performance is lower than that of the Zen cores. In any case, things can only get better for Ryzen.

 

i'd take tomshardware and anandtech over guru3D simply because i know the techies there are more knowledgeable. but fk i dont even want to go into details explaining things on where to find it its why i dont even want to reply LMAO.

i dont even wish to know where you find your sources lol if i need help i'll google search or i'll ask for it man. if you disagree with me simply say you disagree, dont have to bash and say i dont site sources etc. if you want source simply ask for it rather than say i donno what im talking about rofl

 

http://www.ideal.ece.ufl.edu/workshops/wiosca08/slides7.pdf (page 3)
https://en.wikipedia.org/wiki/Instructions_per_cycle

I could go on and on, but it is a mountain of data and EVERYONE ELSE IN THE KNOWN UNIVERSE UNDERSTANDS IT AS TRUE!

Edit:

Calculation of IPC[edit]
The number of instructions per second and floating point operations per second for a processor can be derived by multiplying the number of instructions per cycle with the clock rate (cycles per second given in Hertz) of the processor in question. The number of instructions per second is an approximate indicator of the likely performance of the processor.

The number of instructions executed per clock is not a constant for a given processor; it depends on how the particular software being run interacts with the processor, and indeed the entire machine, particularly the memory hierarchy. However, certain processor features tend to lead to designs that have higher-than-average IPC values; the presence of multiple arithmetic logic units (an ALU is a processor subsystem that can perform elementary arithmetic and logical operations), and short pipelines. When comparing different instruction sets, a simpler instruction set may lead to a higher IPC figure than an implementation of a more complex instruction set using the same chip technology; however, the more complex instruction set may be able to achieve more useful work with fewer instructions.

Factors governing IPC[edit]
A given level of instructions per second can be achieved with a high IPC and a low clock speed (like the AMD Athlon and early Intel's Core Series), or from a low IPC and high clock speed (like the Intel Pentium 4 and to a lesser extent the AMD Bulldozer). Both are valid processor designs, and the choice between the two is often dictated by history, engineering constraints, or marketing pressures. However high IPC with high frequency gives the best performance.

 

woah caps, calm down