Official AMD Ryzen Benchmarks, Reviews, Prices, and Discussion

[JDG1980]

It kind of is though. I remember the pre-BD times in this forum when hype for AMDs new architecture was just as fevered. AMD even had someone from their marketing dept. (JFAMD) who posted regularly (he let everyone know about his job so it wasn't something shady) that further fanned the flames.

The leaked pre-release Bulldozer benchmarks accurately reflected just how poor a CPU it really was. A lot of people didn't want to believe it, since it was hard to believe that AMD would release an architecture that was in many ways worse than its predecessor. This was exacerbated by the fact that JF-AMD was going around telling people not to believe the leaked benchmarks and to wait for release.

With Ryzen, the leaks are far different and indicate a CPU that, by and large, is competitive with Intel's Broadwell-E offerings. Possibly worse OC depending on who you believe, and probably inferior AVX performance based on what we know about the architecture compared to Intel's newer chips - but competitive overall in a wide variety of applications, both single and multithreaded. Sure, it potentially could be a disappointment - but history teaches us that credible-appearing leaks that correspond across multiple sources are generally correct and that people who try to explain them away are usually wrong.

 

[rtsurfer]

The same sort of situation happened with AMD and Nvidia. AMD had significantly better memory controllers than Nvidia during the mid 2000's, but starting with Kepler Nvidia was able to suddenly bolt past AMD in memory speeds for a few years. It's not unreasonable to think AMD built a strong memory controller on top of the other massive improvements they are rolling out over their previous architecture.

Agreed. Too early to call it.

Haswell-E did 3000 on 8 DIMMs right out of the gate. With some tweaking, i would not lie, but it did.

So, seems okay i guess, all the more reasons to wait for Zen 2.

Well, that sort of crashes the entire hype train. We'll see on March 2nd if it'll recover. Timing detection looks right though.

I had Haswell-E, I know you are not lying. But you have to remember that those were 4GB DIMMs. We don't know what Ryzen does with 4GB DIMMs. Haswell-E has sort of broken Mem dividers ( to this day). To run 3000Mhz you needed to use 125 strap.
On 100 strap you'd be better off running 2666,3200Mhz.

You could do 3000Mhz on 100 strap too, it would just be significantly harder.


I don't see how this crashes the hype train for people other than the ones who wish to see it crash. Two reasons for this

1) Who cares if the IMC can't do 4000Mhz as long as the performance at 3200Mhz is still good . I don't see the BW-E guys complain.


2) IMC isn't really "weak", AMD is just taking a safe approach with their launch, don't want to give people access to features that are not fully fleshed out.

Don't believe me? Here's a statement from Asus engineer that says the same

http://www.overclock.net/t/1624058/...-fix-expected-in-1-2-months/180#post_25872953


I bought Skylake day 1 & couldn't get my Corsair kit to run 3600 XMP. While some Gskill kits worked. It was not smooth at launch, but slowly through BIOS updates, things settled down. I expect the same with Ryzen.


But you can sing DOOM if you wanna. It's just not logical that's all.

 

[USER8000]

Because it is not sold with CPU in any retail package. You know, just like very few people will pay for the Wraith Max.

Then if that is the case they should not need to mention it - instead of you trying to defend what they said. You are annoyed I mentioned the fact that Ryzen is an SOC and you want to bury that.

Yes, 3600Mhz, but at what timings? To me it looks fairly obvious that since they run 1T command rate they have way looser timings than required.

Which again is not relevant to a different design - you are trying to look at Intel design and make assumptions on a different CPU.

I have seen people like you and your mates in the last few months - first it was AMD will not support 2667MHZ,then when it was they could not do above 3000MHZ. When it was evident it was higher - then its latency E-PENIS.

Its one moving target after another.

JEDEC standards, look these up.

Intel warranty standards - look it up. No amount of distorting things,changes Intel only officially supports 2400MHZ DDR4 like AMD.

Intel like AMD does not official support IMC overclocking(unless you buy that plan from Intel) - these are all out of spec changes.

AMD helped them with that: they compare 1700 against 7700k themselves.

Not the R7 1800X - but I know you really don't want HEDT CPUs to be compared to Ryzen since you are scared of the price drops,so you must try and bury anybody saying otherwise.

Everyone who talks about weak IMC does mention that bandwidth is good (i mean, 1T command rate, of course it is good, even Broadwell-E hits 91% efficiency with 1T). But latency is not. And believe it or not, games like their latency, though faster cache should solve that one.

Which again its not relevant since it is a different CPU design. It was first MHZ E-PENIS,since Intel was higher. When it was evident AMD could run higher clocked RAM,and had decent bandwidth too,then bandwidth was not important but latency was,etc. It was the same when some were saying,but,but quad channel HEDT>>>>>>>>>>>>>>>>>>>Ryzen since it was dual channel. Then when the leaks hinted it might not be such a big deal it moved to Kaby Lake.

The problem is that all the excuse making on forums about how Intel is XYZ better in marginal out of spec situations like the use of expensive RAM sticks is not really what the vast majority of people who Skylake and Kaby Lake people are running their systems with.

Maybe on tech forums - I don't know anybody running massively overclocked RAM on either.

3GHZ RAM yes,not really any higher.

Those very high speed RAM kits are expensive at least in the UK.

But excuse making is what will give Intel marketing a good set of tips to not drop prices.

Own goal on forums.

Anyways, 3 days to go for fanboys on both sides to realize that Zen is not the next coming of Jesus or the next coming of Bulldozer. It is Phenom III.

It was the same marginal arguments during then about memory speed,and whether one overclocked slightly better than the other too.

In the end it meant a BIG FAT ZERO. Moronic people arguing whether their Phenom II X4 or C2Q would be longer lived. All the people I knew who had them used them as long as each other.

Going for the cheapest option was the best one.

 

[GroundZero7]

I think all these power sensors and smart clocking are in there for maximum mobile power savings (low end clocking) to make their chips competitive in the biggest market segment. It will also pay dividends in the server market where power consumption is a MAJOR selling point.

Could also be something AMD developed for XB2/PS5.

 

[beginner99]

Ryzen is a SOC which has the southbridge as part of the CPU itself..

That's the interesting part. Ryzen has 16 pcie 3.0 lanes for the GPU and another 4 lanes for a NVMe SSD (this depends on the motherboard, could also by 2xpcie and 2x SATA). Plus tons of USB 3.0 from the SOC directly. Then the SOC uses another 4 pcie 3.0 lanes to connect the "chipset". The chipset offers SATA, USB 3.1 (= 3.1 Gen2), 3.0 (3.1 Gen 1) and 2 and pcie 2.0 lanes.

The advanatge over Z270 is therefore pretty huge as in z270 everything! expect the 16 GPU lanes go through the chip set which then is connected to the CPU by 4x PCIe lanes. So all I/O devices on Z270 share these 4xPcie lanes. If that is actually a bottleneck in some situations remains to be seen but would at least be interesting to test for in a review, hello AT?

 

[lolfail9001]

With Ryzen, the leaks are far different and indicate a CPU that, by and large, is competitive with Intel's Broadwell-E offerings. Possibly worse OC depending on who you believe, and probably inferior AVX performance based on what we know about the architecture compared to Intel's newer chips - but competitive overall in a wide variety of applications, both single and multithreaded

Yeah, Ryzen as arch is great, but certainly has room for improvement.

You are annoyed I mentioned the fact that Ryzen is an SOC and you want to bury that.

Ryzen being a SoC does not affect my point one bit, it is just creating excuses to explain whatever you want to explain with it.

Which again is not relevant to a different design - you are trying to look at Intel design and make assumptions on a different CPU.

Nope, i have literally came off from hwbot memory board, looking at 90% bandwidth efficiency on 1T on legitimate quad channel DDR4-3333 with 11-ish timings on Haswell-E. Memory overclocking, you say? That's what memory overclocking looks like.
And yes, Ryzen is not as memory hungry as Skylake, with beefier/faster caches. But not everything fits inside the cache. Far from anything, even.

I have seen people like you and your mates in the last few months - first it was AMD will not support 2667MHZ,then when it was they could not do above 3000MHZ. When it was evident it was higher - then its latency E-PENIS.

We just follow the rumors and leaks, man, while you are here trying to pull off an JF-AMD.

Intel warranty standards - look it up. No amount of distorting things desperately on your part,changes the fact Intel only officially supports 2400MHZ DDR4 like AMD.

Implying i contest that fact. Of course it does not support DDR4-2666, it did not exist as JEDEC standard at the time. And yes, i know of their warranty policy too, whataboutisms do not work well here.

Not the R7 1800X - but I know you really don't want HEDT CPUs to be compared to Ryzen since you are scared of the price drops,so you must try and bury anybody saying otherwise.

See, the projection is here, too! But yes, even AMD knows that their line-up is really meant to be compared against HEDT CPUs because Broadwell-E sucks like that. So, yes, if i want to detect someone cheering/shilling for AMD i check if they compare with 7700k. If I want to detect someone cheering/shilling for Intel, i check for BDW-E comparisons.

Which again its not relevant since it is a different CPU design

No amount of difference in CPU design won't change the fact that CPU wants the data to be delivered as soon as possible. Trying to put it under the rug solves nothing and help you not even a little bit.

But your excuse making is what will give Intel marketing a good set of tips to not drop prices.

And what? I was not going to buy a CPU that would consume more than 100 watts in peak anyways. That's why i looked forward to Ryzen in my rig: i had the stupidity to believe AMD to not lie about power consumption for once. Now i will just have to match it against real measurements and see how far that can be pushed.

The advanatge over Z270 is therefore pretty huge as in z270 everything! expect the 16 GPU lanes go through the chip set which then is connected to the CPU by 4x PCIe lanes.

16 lanes from CPU go directly to the board on Z270 platform, the only x4 lanes bottlenecked stuff is everything on the chipset. So, all the block devices, audio, NICs and I/O. In comparison, on Ryzen it means you can have some Gen 1 ports and 1 NVMe x4 drive not bottlenecked by the chipset. OTOH on Z270 you can have 2 NVMe drives that may not bottleneck each other unless they are used together, but on Ryzen one of them has to be on 2.0, cutting the speed in half or more.

 

[USER8000]

That's the interesting part. Ryzen has 16 pcie 3.0 lanes for the GPU and another 4 lanes for a NVMe SSD (this depends on the motherboard, could also by 2xpcie and 2x SATA). Plus tons of USB 3.0 from the SOC directly. Then the SOC uses another 4 pcie 3.0 lanes to connect the "chipset". The chipset offers SATA, USB 3.1 (= 3.1 Gen2), 3.0 (3.1 Gen 1) and 2 and pcie 2.0 lanes.

The advanatge over Z270 is therefore pretty huge as in z270 everything! expect the 16 GPU lanes go through the chip set which then is connected to the CPU by 4x PCIe lanes. So all I/O devices on Z270 share these 4xPcie lanes. If that is actually a bottleneck in some situations remains to be seen but would at least be interesting to test for in a review, hello AT?

Agreed,which is going to give it a big advantage over X99 when mini-ITX motherboards are released due to cost.

Also, power consumption and TDP measurements need to also take the SOC nature of the system into consideration. Some review sites try to isolate power consumption for the CPU only and in Ryzen this will cover both CPU and chipset power consumption and cooling unlike X99 and socket 1151.

 

[Vaporizer]

12 core SKX-X would have no competition, so it's a "good" value if you really need the cores (i.e. time is money for your use case).



Nah. Bring in a lower-clocked 8 core/16 thread chip with 28 PCIe lanes at $499 to wreck AMD's party, then put an 8 core with 44 PCIe lanes at higher clocks at the $799 price point. Then at the $999 price point, Intel can put out a 10 core/20 thread + 44 PCIe lane part, which would be uncontested. Then, of course, for the really "hardcore" people put out the 12 core/24 thread part with 44 PCIe lanes at $1700.

This way Intel can have its cake and eat it, too. It's really actually straightforward and should be good for the consumer as 6/8 core prices should come down nicely.

This would leave the 7700k in a akward position and would wreck the complete normal desktop line party of Intel.
Considering a 6c Skylake x at ~350?
I7 7700k would have to be discounted to sub 300.

 

[lolfail9001]

This would leave the 7700k in a akward position and would wreck the complete normal desktop line party of Intel.

Only if X299 boards get cheaper. There was already a time when 5820k was cheaper than 6700k (early in Skylake launch). Boards still made the overall price very in favor of 6700k.

 

[USER8000]

snip

What? JF-AMD?? I am not making claims about anything - I find it hilarious how some are trying to over-emphasis this "memory" advantage Intel has.Considering both IPC and clockspeeds seem much better than what many forum experts predicted,it makes me wonder why they think they are 100% correct on the memory front too.

In fact the IMC was barely mentioned until it appeared Ryzen IPC might be good,and soon it became the "most important" thing which some people are trying their best to bump it up in threads.

Its nothing but typical tech forum E-PENIS one-upmanship.

What is the likelihood that the vast majority of Kaby Lake systems don't run above 3GHZ RAM anyway?? In many countries the fast kits are not cheap and its not helped memory prices have shot up in many countries. My mate bought a 3GHZ kit for his system last year and it now costs nearly 70% more. Higher speed RAM costs even more.

Anyway the core is different - you are still trying to look at an Intel design and then thinking the AMD design will be EXACTLY the same.

Plus my comments about memory E-PENIS is that. First it was AMD Ryzen was doomed since it was not quad channel. When leaks hinted it might not be an issue,then it was memory speed. AMD does not support fast RAM over 3GHZ. Now it appears that might be the case,and memory bandwidth is decent,now it is latency.

If memory bandwidth was all that,then a triple channel Core i7 920 setup should always defeat a Core i7 2700K running the same speed RAM.

HEDT quad cores with triple and quad channel memory controllers should always defeat their dual channel consumer quad core compatriots in everything.

Its the same with graphics cards - a GTX980 beat a GTX780TI with less memory bandwidth.



You cannot just compare different designs like that,and you are the ones making claims about how important it is - in fact I would argue that is very JF-AMD like.

 

[Magic Hate Ball]

This would leave the 7700k in a akward position and would wreck the complete normal desktop line party of Intel.
Considering a 6c Skylake x at ~350?
I7 7700k would have to be discounted to sub 300.

I think he's also kind of ignoring the fact that if Intel hasn't already been working on these chips, they won't have them out until Zen+ hits. If it's not in the pipeline now, don't count on it until, what, late 2018 at the very earliest?

You can't just finagle chips out of thin air. Can they cut down Skylake-X to fit a dual channel memory board? Why would they eat their own HEDT margins even if they could?

X299 mobos won't be cheaper than X370 boards simply due to complexity, so there's a slightly higher barrier to market.

If the AMD chips have similar yields to Polaris, and a 8C/16T chip is smaller than the P10 (which it looks to be) then what's stopping them from dropping their margins even lower by the time Intel fires back?

Income at a lesser profit is still income, if it's money you wouldn't otherwise have as the underdog.

 

[Vaporizer]

I dont know. Skylake X are relabeled Xeon chips arent they? And there are 12c+ Dies out there. So adding a 12c to the skylake x should be easily possible. But it will somewhat have impact on Xeon prices?
Also disabling pci lanes is possible.
So the scenario seems quite possible and makes some sense. At the end of the day Intel will look at the market how Ryzen perfprms for the next quarter and will then decide if they introduce price cuts with skylak x or not.
If market stcks to high clocked 4c they will laugh (at the sheeps) and call it a day.

 

[lolfail9001]

What? JF-AMD?? I am not making claims about anything - I find it hilarious how some are trying to over-emphasis this "memory" advantage Intel has.

Find a single one. We talk about it because that's the only thing... well, aside from overclocking not being a thing, to talk about.

In fact the IMC was barely mentioned until it appeared Ryzen IPC might be good,and soon it became the "most important" thing which some people are trying their best to bump it up in threads.

Because we started to get leaks about IMC together with leaks about IPC, genius.

What is the likelihood that the vast majority of Kaby Lake systems don't run above 3GHZ RAM anyway??

Excuses, excuses...

In many countries the fast kits are not cheap and its not helped memory prices have shot up in many countries.

Yes, DDR4 prices shot up lately because mobile phones. So what? 3GMT/s still costs about as much as stock 2.1GMT kit.

Anyway the core is different - you are still trying to look at an Intel design and then thinking the AMD design will be EXACTLY the same.

Different in core uarch does not mean Ryzen can work without any data to work with.

If memory bandwidth was all that,then a triple channel Core i7 920 setup should always defeat a Core i7 2700K running the same speed RAM.

HEDT quad cores with triple and quad channel memory controllers should always defeat their dual channel consumer quad core compatriots in everything.

Its the same with graphics cards - a GTX980 beat a GTX780TI with less memory bandwidth.

Nice strawman, amigo.

You cannot just compare different designs like that,and you are the ones making claims about how important it is - in fact I would argue that is very JF-AMD like.

It is important for my workload because i do stuff that requires jumping all around the gigabytes of allocated memory. Not so important for you, but acting as white knight does not help your cause.

I dont know. Skylake X are relabeled Xeon chips arent they? And there are 12c+ Dies out there. So adding a 12c to the skylake x should be easily possible. But it will somewhat have impact on Xeon prices?

Well, their 2640v4 update should be priced at $1k, and should logically be a 12-core. So, Intel can actually transplant price structure of Broadwell-E, but add 2 cores to each SKU, if they want.

 

[3DVagabond]

Sp anything that shows Ryzen in a positive light is hype and anything that diminishes it is fact? Why do people want to defend $1000 8C/16T chips? Do you get a cut or something?

 

[tamz_msc]

The vast majority of Kaby Lake desktop systems don't run anything above an i5 with bargain-basement 2133MHz DDR4.

If people are so worried about Ryzen's memory performance, then they shouldn't espouse doomsday scenarios and just wait for the reviews in two days' time.

 

[3DVagabond]

The vast majority of Kaby Lake desktop systems don't run anything above an i5 with bargain-basement 2133MHz DDR4.

If people are so worried about Ryzen's memory performance, then they shouldn't espouse doomsday scenarios and just wait for the reviews in two days' time.

It's called "Grasping at Straws". AMD haters do it all of the time.

 

[deasd]

Pardon? Memory problem? Would that lead to performance downgrade? If yes then how much? If not then why you keep talking about it? Sorry but I don't get it.

 

[cytg111]

Find a single one. We talk about it because that's the only thing... well, aside from overclocking not being a thing, to talk about.

Because we started to get leaks about IMC together with leaks about IPC, genius.

Excuses, excuses...

Yes, DDR4 prices shot up lately because mobile phones. So what? 3GMT/s still costs about as much as stock 2.1GMT kit.

Different in core uarch does not mean Ryzen can work without any data to work with.

Nice strawman, amigo.

It is important for my workload because i do stuff that requires jumping all around the gigabytes of allocated memory. Not so important for you, but acting as white knight does not help your cause.

Well, their 2640v4 update should be priced at $1k, and should logically be a 12-core. So, Intel can actually transplant price structure of Broadwell-E, but add 2 cores to each SKU, if they want.

Genius? Why always with the namecalling... This is beginning to ruin my Ryzen reading(i am not the only one).. any chance you two could take it to PM?

 

[lolfail9001]

Sp anything that shows Ryzen in a positive light is hype and anything that diminishes it is fact? Why do people want to defend $1000 8C/16T chips? Do you get a cut or something?

No, the hype are claims of 4.4Ghz on air as some someone made recently after finding out about that knob in MSI manual. The fact is that Ryzen is damn good uarch. The other fact is that Ryzen's IMC is not nearly as good as the core. But see the reaction to the former and the latter.

It's called "Grasping at Straws". AMD haters do it all of the time.

In fact, you do not even need to walk far to see it, it's right here.

Pardon? Memory problem? Would that lead to performance downgrade? If yes then how much? If not then why you keep talking about it? Sorry but I don't get it.

For 99% of you, it will not matter. For me, it kind of does, just like the rumors of AMD misleading with power consumption. Not that i have a choice of "cheap" 8 core CPUs on the market, anyways.

any chance you two could take it to PM?

Sure.

 

[looncraz]

I dont know. Skylake X are relabeled Xeon chips arent they? And there are 12c+ Dies out there. So adding a 12c to the skylake x should be easily possible. But it will somewhat have impact on Xeon prices?
Also disabling pci lanes is possible.
So the scenario seems quite possible and makes some sense. At the end of the day Intel will look at the market how Ryzen perfprms for the next quarter and will then decide if they introduce price cuts with skylak x or not.
If market stcks to high clocked 4c they will laugh (at the sheeps) and call it a day.

A 12-core die will not help Intel compete against Ryzen or Ryzen+. It's not the entire HEDT market that's in danger, it's the top of the mainstream market which AMD has chosen to redefine as having 8 powerful cores.

Intel has no way of competing with that. We are two Intel generations away from when Intel planned to introduce mainstream SIX core CPUs with a 6-core Cannon Lake die. AMD will have Zenver2 (Zen+) with its 8 cores on a more mature process (assuming they don't go for broke and jump to 7nm - I hope they don't, despite the obvious potential rewards if they manage it).

Coffee Lake is just another minor refresh to Kaby Lake, which is just a refresh to Skylake, which is a refresh to Haswell... Higher clocks, lower power, slightly higher IPC is all we're going to see from that at the end of the year.

That will push Ryzen into a lower tier for competition, but not by much... and AMD will be ~6months or less away with Zenver 2 which should re-level the field, if not turn it more in favor of AMD. Intel will have little choice but to use its large die (cut-down) and its enthusiast platform to fight AMD's small die and mainstream platform.

Intel needs an 8-core SKU for the mainstream market, otherwise they will bleed some market-share.

AMD is in the best position it has been in since Phenom II was fighting against Core 2 Quad.

 

[rtsurfer]

No, the hype are claims of 4.4Ghz on air as some someone made recently after finding out about that knob in MSI manual. The fact is that Ryzen is damn good uarch. The other fact is that Ryzen's IMC is not nearly as good as the core. But see the reaction to the former and the latter.

Actually both of the things you said are not "facts" till the official reviews are out or forum members have their chips.

 

[Tup3x]

It would be brilliant for AMD to send RAM kits out that their processors can't run at advertised speeds, wouldn't it?

That was not the point. The sisoft sandra benchmark showed totally different clock speeds so who ever did that benchmark probably did not use the xmp profile.

 

[WhoBeDaPlaya]

Sp anything that shows Ryzen in a positive light is hype and anything that diminishes it is fact? Why do people want to defend $1000 8C/16T chips? Do you get a cut or something?

Same reason they'd defend a supposed $1K+ compute vid card that doesn't even do double-precision any better than regular consumer cards, I guess...

 

[bjt2]

TDP != power consumption
TDP = Typical Dissipation Power

That is to say how much heat energy, in Watts, the product can dissipate.

A device can "consume" 125W and dissipate only 95W of heat. The balance is eventually turned to heat (kinetic energy), just not within the CPU... the balance will be dissipated in the ground plane of the motherboard, PSU, or even the earth ground outside your house.

TDP stands for Thermal Design Power.
When an electrical apparatus is powered with X watts, the sum of kinetic energy, electromagnetic energy and thermal energy MUST be X, because of law of conservation of energy.
Kinetic energy is zero, because AFAIK a powered PC does not move.
Electromagnetic energy is negligible because of FCC rules and because otherwise all of us would be sterile due to testicle overheat.
All remaining energy (almost all) does turn in heat.
Regarding rated TDP of the cooler: i have a 130W sandy bridge 2011, with a cooler rated 220W, so the rating of the cooler does not mean anything...

 

[Veradun]

Never mind that by the time the future is here any current chips are going to be outdated.) don't scale well beyond 2-6 threads and even some of those that do are typically are bound by a single thread that's running some important task such that a high clocking 7700k is a better choice if all you care about is gaming performance.

https://www.computerbase.de/2017-02/cpu-skalierung-kerne-spiele-test/

AMD helped them with that: they compare 1700 against 7700k themselves.

And they sent 1800X s to reviewers, which is compared to the 6900X, by them