Leaked AMD Ryzen Benchmarks?

[os2wiz]

Tweaktown if yeah want 4.0 base clock, ya need water cooling.

PS I am sitting down and listening to the video again, I posted the video earlier in this thread! Great to know people can't seem to get the info when its under their noses.

I totally disagree with that. A high end air cooler will allow 4 GHZ base clock. Obviously not a Coolermaster hyper 212 EVO. Try a Master Maker 8 or Noctua 15D, or a Thermalright Silver Arrow. No way those air coolers won't allow a 4.0 GHZ clock.

 

[os2wiz]

Lol, my post was sarcasm, there are a few doom mongers here and there that I am tired of wasting my data on.

I know Juan well, after 3 years of his non stop garbage on Toms, I lost my chizz and laid out on him, got a perma ban for hostile and offensive abuse. It was worth it, all those badges and trophies were so worth it.

Same thing happened to me on Toms. I did nothing more than ask if he was an Intel employee or stockholder. Toms staff are disgusting supressors of user expression. I never once called anybody a name except after they banned me I called the moderator a Nazi. But that is exactly what he acted like. They coddle Juanrga.

 

[ecmaster76]

Same thing happened to me on Toms. I did nothing more than ask if he was an Intel employee or stockholder. Toms staff are disgusting supressors of user expression. I never once called anybody a name except after they banned me I called the moderator a Nazi. But that is exactly what he acted like. They coddle Juanrga.

There are certain flavors of trolling that are perfect for keeping people logged in and arguing about stuff without being so offensive as to make people leave. Anything for page clicks.

Calling out said trolls is rocking the boat

 

[OrangeKhrush]

The dipshit is EX Bubblebath

 

[cageymaru]

That link seems to be password protected now...

http://www.hardocp.com/news/2017/02/15/ryzen_spy_pics

 

[buttons]

So we can buy this thing in 12 days yet we dont know anything really. Amd has all ryzen info as tight as area51.

Is it because of intel or does it suck?

If i remember correctly, this is the opposite of the bulldozer launch... then again my memory is not the greatest...

 

[OrangeKhrush]

So we can buy this thing in 12 days yet we dont know anything really. Amd has all ryzen info as tight as area51.

Is it because of intel or does it suck?

Benchmarks leaked show it doesn't suck at all and that was mostly at low frequency and locked clocks. It won't be the fastest in anything but it will be extremely fast.

 

[CSI_PC]

If i remember correctly, this is the opposite of the bulldozer launch... then again my memory is not the greatest...

This is contentious for some here, IMO since the division-team split it seems to me the CPU division is more professional than the Radeon one in context of managing presentations-data-official leaks.
At least their approach fits more with the bellweather enterprise tech companies (broad statement and yeah there will be bad examples but plenty of good ones), Radeon/GPU division most of the time makes me frown with their approach.

Cheers

 

[razor1]

Hi Razor.

I have been following chips, TDPs, and the silicon lottery since before the Athlon was released...

Unexpected market disruptions occur, often quickly. Sometimes, despite superior implementations of certain features (Tile Based Rendering in the Kyro II at a fraction of the cost 17 years ago?) technologies don't pan out due to bad timing or bad luck.

What F are you talking about, I was talking about thermodynamics, actually physics and why certain things happen, and you sit here and pull that out of your ass?

Currently, AMD has a customer in Google for their Deep Learning products for 2017 http://www.forbes.com/sites/aaronti...g-machine-learning-in-the-cloud/#3d5983e14181

I don't know if you've heard of this company, but Google was co-founded by a kid who graduated from the same high school I did. That's not important, or a claim to intelligence, just a fun bit of trivia. Just wanted to make sure you realized this small garage-based company was interested in their silicon.

Yes, Nvidia was there first, but ultimately that isn't a guarantee for anything. But, if you remember 3dfx was also a dominating market force at one point.

Did you read the freakin article

But again, Nvidia is way ahead of AMD here. It's been providing software tools for deep learning to developers for years now. AMD has a lot of catching up to do.

The funny thing about your statement is that you fall into this logical fallacy constantly in your posts when I read them. "Intel (or Nvidia) has succeeded in the past, they must surely succeed in the future!" "They have the money/contracts, they must surely dominate!"

It's rather entertaining.

What is entertaining is that you can't even fuckin read the articles that you post lol

 

[razor1]

I kind of agree,you're just throwing terms out there to scare him away. None of that will tell you how well this thing will overclock over its base clock. Obviously they can't hit 220w at 14nm, but Intel chips have shown they can go up to ~130w on a 91w spec quad at 14nm, and we know the core sizes are similar.

Fwiw, I have a degree in microelectronics and wrote a paper on maximum clock speed at a given node for my semiconductor physics 2 class and none of those terms were in my paper. Maybe that's why I only got an 80 on it? The electromagnetic field on the gate insulator was the limiting factor in my paper as exceeding a certain amount would damage it. Electromigration is a longer term reliability problem,not an out of the box clock limiter.

That was theoretical though,you'd have to know the layout of the chip and test thermals throughout the chip layout to see where localized hot spots are that can limit performance,which all of the research papers in the world won't help you with. It's why you get higher clocks with hyper threading turned off on an Intel vs on,even though everything else is exactly the same. Hyper threading isn't causing more electromigration or even higher gate leakage, it's causing more localized heat somewhere on the chip causing slightly lower clocks. You can read all the papers you want and it won't tell you how or if that will affect clocks without testing it as it's architecture related, not node related. Given the ryzen chips are a completely new architecture that nobody has flogged yet,there's no real way to determine it. Maybe they've figured out how to enable hyper threading with no overclocked penalty? Maybe it's worse? Who knows?

Then I suggest both of you go back to high school physics and have your teachers spank you!

https://www.google.com/search?q=leakage+temperature+voltage&ie=utf-8&oe=utf-8

wow your paper didn't but so many others do, why is that? Cause people that write these papers must not know as much as you do.....

Where did you go to college again, I'll tell people never to go there for the type of degree you got.

http://www.ruf.rice.edu/~mobile/elec518/readings/DevicesAndCircuits/kim03leakage.pdf

I guess people at rice know about hotspots and leakage and voltage, hmm maybe they should forfeit their degrees and and have you teach them?

I agree with you about I don't know unless I know the architecture bit, I suggest you go look at 3 or few posts or so prior to that post I made, cause I stated the same shit you did lol.

Yes no one bothers with reading anymore but they take crap out of context because of laziness and convenience.

Well as you stated 4 core Ryzen at launch looks to be 3.6 ghz base clock for the highest end one, so not sure if their are any ones above that. Not sure if later on they will come out with higher clocked parts, cause if they can't do it right now, by the time they will be ready with higher clocked cores, Intel might have a response. It seems like the process is holding back the clock speed, its unusual to see a half disabled chip or half the size chip not being able to clock more than the full chip. Seems like voltage limits are close. But it could be the architecture forcing those voltage limits too.

Oh forgot, AMD isn't selling Ryzen to OEM's yet at least for another quarter or maybe two, so either OEM's aren't going to pick up Ryzen because well diversifing product lines actually hurt sales in the same segment when market share of a particular product is extremely high or AMD is trying to get rid of old CPU stock, I think its a combination of both.

People its important to read!

Take your freakin emotions out of your pants and read.

 

[os2wiz]

Yeah good luck with that will wait for reviews to see if they can over clock like that. Maybe very few gold samples for the sounds of it from tweak town.

Not implying anything btw, never do, my posts are very straight forward, no BS around the edges, if you haven't noticed.

Don't know how long you are in the hardware game, but tweaktown was one of the main leakers 5 to 10 years back, their info is usual very good.

Beta blockers can do that but I don't need those, don't have any memory issues, although lack of sleep and a few drinks will , ps. you should ask your doctor about calcium channel blockers, they might help with the memory problems.

My point is that no one can legitimately make that statement unless they list all the high end fans they used to attempt an overclock. It is purely hype that statement. We have to be scientific in our investigations or its all medieval superstition. My guess is somebody used a far from top shelf air cooler and then like a fanboy lied like hell.

 

[razor1]

My point is that no one can legitimately make that statement unless they list all the high end fans they used to attempt an overclock. It is purely hype that statement. We have to be scientific in our investigations or its all medieval superstition. My guess is somebody used a far from top shelf air cooler and then like a fanboy lied like hell.

And why is Tweaktown, Why Cry, and a couple of leak sites saying the same things or similiar, frequency is going to be an issue?

I stated this earlier in the thread!

4 different people with different connections inside AMD or connections to people that have the chips (motherboard manufactures) saying similar things?

 

[os2wiz]

Yeah good luck with that will wait for reviews to see if they can over clock like that. Maybe very few gold samples for the sounds of it from tweak town.

Not implying anything btw, never do, my posts are very straight forward, no BS around the edges, if you haven't noticed.

Don't know how long you are in the hardware game, but tweaktown was one of the main leakers 5 to 10 years back, their info is usual very good.

Beta blockers can do that but I don't need those, don't have any memory issues, although lack of sleep and a few drinks will , ps. you should ask your doctor about calcium channel blockers, they might help with the memory problems.

You and I are cut of the same cloth. If you have noticed I have two accounts here. I am os2wiz on my desktop and Ira Wechsler on my tablet. I had problems due to short term memory issues while logging on with os2wiz on my tablet. I had to create a new account. I already sent a message to Kyle about it. If he can merge my Ira Wechsler into my os2wiz account I will have 100 posts in the forum. I am very forthright but a little less belligerent than you. I respect the fact that you are seeking truth,, because that is all that I am about. I just won't take Tweaktown's word for it unless they fully describe what cooling they used. That is the only verifiable methodology. I do make mistakes, but I am not a dogmatist. So we will have to see on this. I respect knowledge. At one time I respected Juanrga. Unfortunately it became very apparent he was unable to deal with truths that contradicted his own notions. He is very knowledgable , yet very flawed due to his hubris. He has descended into trolldom.

 

[razor1]

You and I are cut of the same cloth. I had problems due to short termmmeory issues with logging on with os2wiz on my tablet. I had to create a new account. I already sent a message to Kyle about it. If he can merge my Ira Wechsler into my os2wiz account I will have 100 posts in the forum. I am very forthright but a little less belligerent than you. I respect that fact that you are seeking truth,, because that is all that I am about. I just won't take Tweaktown's word for it unless they fully describe what cooling they used. That is the only verifiable methodology. I do make mistakes, but I am not a dogmatist. So we will have to see on this. I respect knowledge. At one time I respected Juanrga. Unfortunately it became very apparent he was unable to deal with truths that contradicted his own notions. He is very knowledgable , yet very flawed due to his hubris. He has descended into trolldom.

Sorry for being belligerent, that was not my intention, it gets tiresome around launch time for new products, when there are so many people that come out of the wood works and old crazy fanatics that do the same,

physics, math, logic just go out the window, come on guys we are a bit better than that I think.

 

[os2wiz]

And why is Tweaktown, Why Cry, and a couple of leak sites saying the same things or similiar, frequency is going to be an issue?

I stated this earlier in the thread!

4 different people with different connections inside AMD or connections to people that have the chips (motherboard manufactures) saying similar things?

I am not saying the 14LPP will give the same high clocks as Intel's process. But the X iterations of Ryzen will overclock a bit more . Also do we know what chipset they used on the AMD motherboard? If it was not at least a B350 or X370 and if it did not have at least 8x2 phase control it will not overclock well at all. Some of these B350 and even x370 motherboards have inadequate vrm power phase control to do much with overclocking although they claim they can. I even saw many in the Bulldozer days who bought boards with the 970 chipset instead of a 990FX and cried they could not get a decent overclock. They had 6x2 phase control or 4x4 phase control which was wholly inadequate. So there are several variables that have not been fully revealed publicly. Too early to have confidence in the conclusion they gave. I do agree the overclocks most likely won't go to 4.5 on water for all cores on this intial sku of chips. I do believe there is more room in refining the process to get higher clocks without the current thermal wall

 

[razor1]

I am not saying the 14LPP will give the same high clocks as Intel's process. But the X iterations of Ryzen will overclock a bit more . Also do we know what chipset they used on the AMD motherboard? If it was not at least a B350 or X370 and if it did not have at least 8x2 phase control it will not overclock well at all. Some of these B350 and even x370 motherboards have inadequate vrm power phase control to do much with overclocking although they claim they can. I even saw many in the Bulldozer days who bought boards with the 970 chipset instead of a 990FX and cried they could not get a decent overclock. They had 6x2 phase control or 4x4 phase control which was wholly inadequate. So there are several variables that have not been fully revealed publicly. Too early to have confidence in the conclusion they gave. I do agree the overclocks most likely won't go to 4.5 on water for all cores on this intial sku of chips. I do believe there is more room in refining the process to get higher clocks without the current thermal wall

That I can agree with to a degree, the amount of phase controller not so much though, phases don't affect the end results of the chip itself, phases are there to help stabilize the voltage and power draw across all the motherboard needs.

 

[os2wiz]

Sorry for being belligerent, that was not my intention, it gets tiresome around launch time for new products, when there are so many people that come out of the wood works and old crazy fanatics that do the same,

physics, math, logic just go out the window, come on guys we are a bit better than that I think.

Thanks. I totally agree with the frustration of all the hype of the launch. I know you are a good guy, I behave very similarly under stress and fatigue. By the way I know about calcium channel blockers but my cardiologist insists they do not do the job as efficiently. he saved my life 5 years ago and I feel I have to cut him some slack.

 

[lolfail9001]

Oh, for pete's sake, in 2 weeks we will already know just how far the first batch can be pushed in. My personal expectation is the first batch Broadwell-E level. It is fucking bad for 4Ghz boost chip, but it is still something.
The only worry i have is that i still have not figured out why Passmark's prime numbers and physics test is so bad, and this could badly misfire.

 

[Magic Hate Ball]

Sorry for being belligerent, that was not my intention, it gets tiresome around launch time for new products, when there are so many people that come out of the wood works and old crazy fanatics that do the same,

physics, math, logic just go out the window, come on guys we are a bit better than that I think.

With you, it seems, basic manners go out the window.

Will you post a retraction if you are incorrect regarding issues with Ryzen 8C hitting 4ghz on a good air cooler?

 

[os2wiz]

Oh, for pete's sake, in 2 weeks we will already know just how far the first batch can be pushed in. My personal expectation is the first batch Broadwell-E level. It is fucking bad for 4Ghz boost chip, but it is still something.
The only worry i have is that i still have not figured out why Passmark's prime numbers and physics test is so bad, and this could badly misfire.

You can not accurately determine IPC from one synthetic benchmark. You have to take an average of all nechmarks and single thread apps. Some apps are coded in such a way that they defeat the logic used in a cache. The next singel thread app may not act the same way. So do not get discouraged. I suspect Haswell+ IPC Braodwell may be a little bit of a stretch. But at Haswell levels it is still good enough to run all those single-threaded games that are written by Neanderthal software designers. Really no excuse for stone age programming these days. It is either they get subsidized by Intel or that they are purely lazy. Game performance improves when you incorporate more threads into the equation.

 

[os2wiz]

With you, it seems, basic manners go out the window.

Will you post a retraction if you are incorrect regarding issues with Ryzen 8C hitting 4ghz on a good air cooler?

No need to press it. I am sure Razor1 will be the first to admit it if he is incorrect. We do not want bombastic accusations nor a witch hunt here so lets keep it as cordial and tolerant as possible. We are all human and all flawed.

 

[razor1]

Thanks. I totally agree with the frustration of all the hype of the launch. I know you are a good guy, I behave very similarly under stress and fatigue. By the way I know about calcium channel blockers but my cardiologist insists they do not do the job as efficiently. he saved my life 5 years ago and I feel I have to cut him some slack.

ah yeah cerrtain heart problems and calcium blockers have contra indications! Your doctor made a good call

 

[razor1]

With you, it seems, basic manners go out the window.

Will you post a retraction if you are incorrect regarding issues with Ryzen 8C hitting 4ghz on a good air cooler?

Yes I will, don't need a retraction, this isn't the new york times lol, we are speculating on rumors right? The rumors are saying clocking issues due to so factor in thermodynamics in the silicon. I'm sticking with the rumor because its from multiple sources, you are not. Simple as that.

 

[Magic Hate Ball]

Yes I will, don't need a retraction, this isn't the new york times lol, we are speculating on rumors right? The rumors are saying clocking issues due to so factor in thermodynamics in the silicon. I'm sticking with the rumor because its from multiple sources, you are not. Simple as that.

Alright then. We'll agree to disagree on how reliable the rumor sources are.

I'll check back in at launch.

 

[OrangeKhrush]

Oh, for pete's sake, in 2 weeks we will already know just how far the first batch can be pushed in. My personal expectation is the first batch Broadwell-E level. It is fucking bad for 4Ghz boost chip, but it is still something.
The only worry i have is that i still have not figured out why Passmark's prime numbers and physics test is so bad, and this could badly misfire.

On Anandtech they showed RAM speed and latency used, add that the CPU running at its max turbo ie: 3.7ghz ZEN had about 30-40% performance in prime/physics left out there. Not really like prime and physics are the be all and end all.

 

[razor1]

On Anandtech they showed RAM speed and latency used, add that the CPU running at its max turbo ie: 3.7ghz ZEN had about 30-40% performance in prime/physics left out there. Not really like prime and physics are the be all and end all.

What about the single threaded performance?

The leak came from a system builder, Boost Boxx, they used a motherboard without boosting capabilities.

As far as synthetics go, Physics is important as its testing game like physics simulations so that should also factor in single thread performance quite a bit.

Integer math is multi threaded dependent, and that should be good for Ryzen.

Prime numbers the amount of data needed, shuffling between the memory and cpu is going to be an issue with the memory type used, not sure if its the memory or motherboard (cause we don't know what are the requirements of Ryzen when it comes to memory) or just something not so good with Ryzen when it comes to memory, AMD in the past has had these issues with memory too.

 

[Magic Hate Ball]

What about the single threaded performance?

The leak came from a system builder, Boost Boxx, they used a motherboard without boosting capabilities.

As far as synthetics go, Physics is important as its testing game like physics simulations so that should also factor in single thread performance quite a bit.

Integer math is multi threaded dependent, and that should be good for Ryzen.

Prime numbers the amount of data needed, shuffling between the memory and cpu is going to be an issue with the memory type used, not sure if its the memory or motherboard (cause we don't know what are the requirements of Ryzen when it comes to memory) or just something not so good with Ryzen when it comes to memory, AMD in the past has had these issues with memory too.

I've been following the Anandtech thread on the Passmark scores people are testing out. The benchmark for primes seems very off, as the most recent result from testing showing a 5820K going from 57 primes with HT enabled to 80 primes with it disabled....

I'm not sure how it is programmed but apparently HT is a major liability for that test.

 

[razor1]

I've been following the Anandtech thread on the Passmark scores people are testing out. The benchmark for primes seems very off, as the most recent result from testing showing a 5820K going from 57 primes with HT enabled to 80 primes with it disabled....

I'm not sure how it is programmed but apparently HT is a major liability for that test.

I think its the cas latency, that will affect the physics score too. There is no way an 8 core Ryzen with ~haswell level IPC, should be getting half the physics and prime number scores of a 4 core 4.0 ghz skylake which that is what it seems to be getting.

But then again gotta ask the question why did they use ram with such loose timings and latency for such a CPU.

 

[lolfail9001]

I'm not sure how it is programmed but apparently HT is a major liability for that test.

Of course it is, HT destroys cache in that test, when it is [performance] depends on it. Memory's impact on it is still minor, though.

Yeah, i am positive that whatever is it that causes sample to suck in Primes test here if it will be reproduced in actual testing will be fairly noticeable.

I think its the cas latency, that will affect the physics score too. There is no way an 8 core Ryzen with ~haswell level IPC, should be getting half the physics and prime number scores of a 4 core 4.0 ghz skylake which that is what it seems to be getting.

Definitely not just latency, on Sandy it looks to have fairly minor effect from even doubling it.

 

[razor1]

Of course it is, HT destroys cache in that test, when it is [performance] depends on it. Memory's impact on it is still minor, though.

Yeah, i am positive that whatever is it that causes sample to suck in Primes test here if it will be reproduced in actual testing will be fairly noticeable.

Definitely not just latency, on Sandy it looks to have fairly minor effect from even doubling it.

Interesting yeah you are right!

 

[OrangeKhrush]

Th

What about the single threaded performance?

The leak came from a system builder, Boost Boxx, they used a motherboard without boosting capabilities.

As far as synthetics go, Physics is important as its testing game like physics simulations so that should also factor in single thread performance quite a bit.

Integer math is multi threaded dependent, and that should be good for Ryzen.

Prime numbers the amount of data needed, shuffling between the memory and cpu is going to be an issue with the memory type used, not sure if its the memory or motherboard (cause we don't know what are the requirements of Ryzen when it comes to memory) or just something not so good with Ryzen when it comes to memory, AMD in the past has had these issues with memory too.

The single thread is fine, my 5960x at 3.5 gets about 2100 at 3ghz about 1850 so single thread is about the same

 

[OrangeKhrush]

We tried to work that out, a good set of 13 cas 2400's are available

 

[razor1]

Th

The single thread is fine, my 5960x at 3.5 gets about 2100 at 3ghz about 1850 so single thread is about the same

So ~haswell level IPC going up against Skylake with some issues with physics and primes.

You think those aren't used much, physics is used in games and primes are used in HPC. Is that why AMD stated HPC not so much for Ryzen?

 

[OrangeKhrush]

So ~haswell level IPC going up against Skylake with some issues with physics and primes.

You think those aren't used much, physics is used in games and primes are used in HPC. Is that why AMD stated HPC not so much for Ryzen?

Or can just wait for a high end X370 board with top notch memory and get a better representative number then decide where it is at. ~40 on a crap board and cheap ram is rather good.

Physics was not an issue on 3dmark

 

[razor1]

Or can just wait for a high end X370 board with top notch memory and get a better representative number then decide where it is at. ~40 on a crap board and cheap ram is rather good.

Physics was not an issue on 3dmark

Just interesting that you opted to take out the things that Ryzen doesn't look too good in...... now you are down to lets wait and see, which is what I've been saying all along.

 

[Pieter3dnow]

Oh, for pete's sake, in 2 weeks we will already know just how far the first batch can be pushed in. My personal expectation is the first batch Broadwell-E level. It is fucking bad for 4Ghz boost chip, but it is still something.
The only worry i have is that i still have not figured out why Passmark's prime numbers and physics test is so bad, and this could badly misfire.

Answers here



It is some testing with Passmark on same memory and cpu clock speed

 

[OrangeKhrush]

Just interesting that you opted to take out the things that Ryzen doesn't look too good in...... now you are down to lets wait and see, which is what I've been saying all along.

FP, integer, IPC is all good and that will drive performance. aMD also tried giving enthusiast level on mainsteam die size with less than what intel e's dies have and with 30% lower clockspeed than kabylake. All in all an impressive showing given that the goals have been moved from ipc to some lousy synthetic. I would rather run superpi. Two pages before that a user posted a list from 1333 to 3200 and how spd/cas/ timings affect results, lowest to highest was over 30% -40% that is major. I dont think the bench was exactly well optimized like toms hardware getting 20% higher 3dmark scores than official record scores.

 

[Cali3350]

Im expecting around 4GHZ stable to be the 'normal' overclock, and im expecting Broadwell IPC. If I get those, Ill take the dive. Ill give up the extra Intel performance for more cores and to help AMD (since thats more than enough performance for me anyway, and Intel needs the kick in the balls).

 

[lolfail9001]

It is some testing with Passmark on same memory and cpu clock speed

You know my opinion on that guy, what does he justify the disrepancy in primes/physics.. or rather, Ryzen matching 6700k on those in spite of having twice the thread count.

mainsteam die size

That die is fairly close to the size of Broadwell LCC die, actually. The one with 2 more cores, 2 more channels, 8 more PCI-E lanes. Hardly "mainstream".

 

[razor1]

FP, integer, IPC is all good and that will drive performance. aMD also tried giving enthusiast level on mainsteam die size with less than what intel e's dies have and with 30% lower clockspeed than kabylake. All in all an impressive showing given that the goals have been moved from ipc to some lousy synthetic. I would rather run superpi. Two pages before that a user posted a list from 1333 to 3200 and how spd/cas/ timings affect results, lowest to highest was over 30% -40% that is major. I dont think the bench was exactly well optimized like toms hardware getting 20% higher 3dmark scores than official record scores.

You don know Intel's iGPU is like 40% of its die right? I'm not impressed with Ryzen's die size, but again, people that need only CPU's because they use discrete GPU anyways its better for them.