Leaked AMD Ryzen Benchmarks?

[CSI_PC]

Hey, I did , but I needed m-ATX for htpc use, so no choice as there are no Z370 m-ATX boards. As the ASUS Prime supports 3200 ram, I figured why not for $20 more.

LOL,
well that serves me right I guess
Cheers

 

[os2wiz]

I mean, Skylake has exactly half the latency at very similar settings (looser timings a little, but same MT rate). But a whole 2Gig/sec lower bandwidth.
Another curious thing is that Skylake has way lower latencies than what AIDA has here. But i guess that's the trade-off of cache vs latency.

TRhis test was done on an inferior $50 motherboard that can not process this ram properly. It is absolutely foolish and unscientific to give this latency issue any validity. Out high performance ddr4 ina decent motherboard and then look at latencies. But running wuinrar is bullcrap and dep six seven-zip they are single threaded apps while zip apps can be multithreaded. Nobody these days in the US is using winrar. I use win zip which whilkl kick ass in performance over 7 - zip.

 

[os2wiz]

Biquadratic actually, if you assume that stock Ryzen consumes 9W in idle and it had 3.5Ghz clock in P95.

Let me quote someone who actually overclocks stuff:

Now compare that and what AMD shows here.

Shit motherboard that can not accept the memory. If you accept Juanrga's fantasies you will have egg on your face. Wait until a proper test is done on a proper motherboard.

 

[lolfail9001]

Shit motherboard that can not accept the memory. If you accept Juanrga's fantasies you will have egg on your face. Wait until a proper test is done on a proper motherboard.

We were talking about AMD's tweet done on what looks to be Crosshair VI. If that is a shit motherboard that cannot accept the memory in your book, you already have an egg on your face. So, wipe it off and look at that watermark:

Spoiler

TRhis test was done on an inferior $50 motherboard that can not process this ram properly.

I talk about AIDA64 test with 3200C16. You are so fixated on trying to prove Ryzen perfect, you have stumbled on context twice already.

And I bet those OCers that hit 12-11-11-28-1T/4200 w/Kaby Lake are running a single 4GB stick or something.

People running 2 sticks hit "just" 4000 at these timings with Skylake. Samsung B-Die is a beast when you are not worrying about having to throw it out every month or two because of 2.0V voltage.

I find it funny that those negging the Ryzen based on this review are blissfully ignoring the fact they were running it on a heatsink with half the mass of a Hyper 212 EVO.

You do understand that presence of airflow makes fin area kind of more relevant than heatsink weight? Brick of copper would probably cool stuff worse than my 200g flower.

It is probably the reason for some benches being weird but I am willing to give a new chipset time to teeth in. Not like i will have a motherboard for it anyway until July

Wait, didn't you pre-order x370 mobo already? You posted as much, i recall. But yeah, early issues are to be expected, i wonder if anyone will revisit it after initial reviews though.

 

[OrangeKhrush]

We were talking about AMD's tweet done on what looks to be Crosshair VI. If that is a shit motherboard that cannot accept the memory in your book, you already have an egg on your face. So, wipe it off and look at that watermark:

Spoiler

I talk about AIDA64 test with 3200C16. You are so fixated on trying to prove Ryzen perfect, you have stumbled on context twice already.

People running 2 sticks hit "just" 4000 at these timings with Skylake. Samsung B-Die is a beast when you are not worrying about having to throw it out every month or two because of 2.0V voltage.

You do understand that presence of airflow makes fin area kind of more relevant than heatsink weight? Brick of copper would probably cool stuff worse than my 200g flower.

Wait, didn't you pre-order x370 mobo already? You posted as much, i recall. But yeah, early issues are to be expected, i wonder if anyone will revisit it after initial reviews though.

I have a 1600X sample chip, no motherboard, I am a ITX person so there was no option so I have to wait.

 

[Magic Hate Ball]

You do understand that presence of airflow makes fin area kind of more relevant than heatsink weight? Brick of copper would probably cool stuff worse than my 200g flower.
.

Except it has the exact same structure and design.... with less mass. And it's a 100mm fan so it's a smaller surface area than the cooler I use on my 4770K.

 

[lolfail9001]

Except it has the exact same structure and design.... with less mass.

Fin count, man, fin count.

 

[Gideon]

Oh look the Intel lovers came out and latched on to some unknown site in Iran and treated it's benches like the gospel when it showed Ryzen in a bad light and ignored any at the same site that showed good results.... color me shocked. Always the same names over and over again, spewing garbage all over the AMD forum. Must be rough having literally no life other than scouring the web looking for bad things to say about AMD. Thankfully we should have accurate reviews tomorrow and we can watch you guys troll all over any benchmark Intel wins at and tell us how we made a mistake buying Ryzen, like we suddenly decide to care what your opinion is now. Must be rough being that insecure in life.

 

[sover]

You do understand that presence of airflow makes fin area kind of more relevant than heatsink weight? Brick of copper would probably cool stuff worse than my 200g flower.

Unfortunately that's not actually how it works. Heatsink mass is at least as important as surface area when it comes to dissipating heat. Without enough mass for heat to actually spread through to those fins those fins become expensive decorations. If it was all about fins and air flow then intel would just use less material in their stock heatsinks and make them a cluster of fins.

 

[Comixbooks]

Alot of it depends on what the motherboard takes as a Max Ram

Shit motherboard that can not accept the memory. If you accept Juanrga's fantasies you will have egg on your face. Wait until a proper test is done on a proper motherboard.

Yeah it's usually the motherboard that determines what frequency if Ram you can install.

 

[lolfail9001]

Unfortunately that's not actually how it works.

Yeah, it is a tad bit more complicated because you have to consider like 5 thermal transfers.

Without enough mass for heat to actually spread through to those fins those fins become expensive decorations.

That would be defined by thickness of fin, mass is a byproduct of fin area and fin thickness, after all.

If it was all about fins and air flow then intel would just use less material in their stock heatsinks and make them a cluster of fins.

Actually a very tight cluster of fins can easily weight as much as a proper solid brick of same material, that's common sense, and description of server heatsinks in the same time.

 

[juanrga]

Biquadratic actually, if you assume that stock Ryzen consumes 9W in idle and it had 3.5Ghz clock in P95.

We are talking about different stuff I believe.

I have used the relationship between voltage and frequency to get

Power ~ f^(n+1)

then I used the numbers given in the review (265W @4GHz; 170W @3.4GHz) to solve for the parameter and I got

n ~= 2

Thus power is scaling almost cubically with frequency.

 

[sover]

Actually a very tight cluster of fins can easily weight as much as a proper solid brick of same material, that's common sense, and description of server heatsinks in the same time.

Thermodynamics requires mass and a conducive shape for proper heat dissipation. The more fins there are in a space, the more narrow they have to be, and very narrow fins can't carry heat very far because the amount they can actually conduct already dissipates near their base. That's why for many server heatsinks even when they have lots of fins those fins aren't very tall, and they're on a massive base. For heatpipe tower style coolers the number and mass of the heatpipes is the most important part, not how many fins are attached.

 

[juanrga]

I don't know if this is a joke or serious

 

[Grimlaking]

HA now that's funny and interesting.

 

[os2wiz]

Ah ok.
Well without being able to read the article I could only go by what he said
But still, it means it is not the actual retailer product and context still applies it raises questions where and revision of motherboard they have (which controls the microcode).
Memory config they did/let happen is enough anyway to make me leery.
Cheers

on the engineering sample on Canard pc. Its recent tests directly contradict results seen on production samples for review:

Canard PC guy is stuck on the measurements he made on an early sample, he still has no final chip and is unaware of the definitive numbers.

Indeed he may have been somewhat trolled because it was clear that his sample was well below the server parts for FP in the Sandra submssion, FTR below are the scores of his sample and of a final chip :



AMD Eng Sample: 2D3151A2M88E4_35/31_N (8M 16T 3.14GHz, 8x 512kB L2, 2x 8MB L3)

Score 317,84Mpix/s

Performance 101,09Mpix/s/GHz


Multi-Media Integer 458,38Mpix/s

Multi-Media Long-int 137,10Mpix/s

Multi-Media Quad-int 2717kpix/s

Multi-Media Single-float 348,24Mpix/s

Multi-Media Double-float 201,16Mpix/s

Multi-Media Quad-float 9534kpix/s




Ryzen 7 1800X 3.62GHz

Score 461.13Mpix/s

Performance 127.31Mpix/s/GHz


Multi-Media Integer 557.85Mpix/s

Multi-Media Long-int 168.26Mpix/s

Multi-Media Quad-int 3816kpix/s

Multi-Media Single-float 546.84Mpix/s

Multi-Media Double-float 321.44Mpix/s

Multi-Media Quad-float 14127kpix/s

#54



As for the Iranian drek. We do not know that this anything but an engineering sample as well. Juanrga has claimed the same crap on several other occasions and been proven wrong every time. No verification, no science.

Today, 07:21 AM

 

[lolfail9001]

We are talking about different stuff I believe.

Yes, you forget that the measured power is at the wall.

HA now that's funny and interesting.

10/10 jokes.

on the engineering sample on Canard pc. Its recent tests directly contradict results seen on production samples for review:

Where? I.e., did he actually provide raw numbers? Also, they said as much that their sample either had stability or performance issues, and they did not have fix at the time.

We do not know that this anything but an engineering sample as well.

ES!=QS, but that flew over your head as those 2 memory moments.

 

[lolfail9001]

Thermodynamics requires mass and a conducive shape for proper heat dissipation

Time to dig up that old compute job on the heatsinks.

the more narrow they have to be

Not really, as long as you can decrease the distance between them keeping thickness constant. It will rise the airflow requirements and make it way heavier, obviously, but effect is well specified. Granted, properly verifying myself requires stuff i do not have on hand (the decent model, that is), so i'll take your word for it.

 

[OrangeKhrush]

As for the Iranian drek. We do not know that this anything but an engineering sample as well. Juanrga has claimed the same crap on several other occasions and been proven wrong every time. No verification, no science.

Today, 07:21 AM

Are you sure, I could swear he nailed his projections:

Sandy IPC
No Hexcore
3Ghz impossible, 3.5Ghz if lucky

 

[razor1]

https://videocardz.com/66745/amd-ryzen-7-press-deck-leaked

Full press deck

Just to point out overclocking lol

R7 1800X all core boost 3,7 GHz, XFR single core 4,1 GHz (slide 10)

XFR Single core!

 

[razor1]

on the engineering sample on Canard pc. Its recent tests directly contradict results seen on production samples for review:

Canard PC guy is stuck on the measurements he made on an early sample, he still has no final chip and is unaware of the definitive numbers.

Indeed he may have been somewhat trolled because it was clear that his sample was well below the server parts for FP in the Sandra submssion, FTR below are the scores of his sample and of a final chip :



AMD Eng Sample: 2D3151A2M88E4_35/31_N (8M 16T 3.14GHz, 8x 512kB L2, 2x 8MB L3)

Score 317,84Mpix/s

Performance 101,09Mpix/s/GHz


Multi-Media Integer 458,38Mpix/s

Multi-Media Long-int 137,10Mpix/s

Multi-Media Quad-int 2717kpix/s

Multi-Media Single-float 348,24Mpix/s

Multi-Media Double-float 201,16Mpix/s

Multi-Media Quad-float 9534kpix/s




Ryzen 7 1800X 3.62GHz

Score 461.13Mpix/s

Performance 127.31Mpix/s/GHz


Multi-Media Integer 557.85Mpix/s

Multi-Media Long-int 168.26Mpix/s

Multi-Media Quad-int 3816kpix/s

Multi-Media Single-float 546.84Mpix/s

Multi-Media Double-float 321.44Mpix/s

Multi-Media Quad-float 14127kpix/s

#54



As for the Iranian drek. We do not know that this anything but an engineering sample as well. Juanrga has claimed the same crap on several other occasions and been proven wrong every time. No verification, no science.

Highlighted in orangish and red.
Yes we do

ZD nomenclature of AMD parts are not ES's, those are qualification parts. Qualification parts are final release products, and if you want to look this up please look at Intel's in dpeth documentation on how qualification is done, and how motherboard manufactures are brought into the process early on. AMD as something similar. If you are so bold just call up Intel and AMD and purchase them as well, you will need to buy them in lots.

https://www.pcper.com/news/Processors/Report-AMD-Ryzen-Performance-Ashes-Singularity-Benchmark

It is also interesting to note that the processor listed in the screenshot above is (apparently) not an engineering sample, as TechPowerUp points out in their post:

"Unlike some previous benchmark leaks of Ryzen processors, which carried the prefix ES (Engineering Sample), this one carried the ZD Prefix, and the last characters on its string name are the most interesting to us: F4 stands for the silicon revision, while the 40_36 stands for the processor's Turbo and stock speeds respectively (4.0 GHz and 3.6 GHz)."

So you are saying pc per and tech power up don't know what the nomenclature of the serial numbers are? This is the same thing Juanrga stated, so take it or call BS on everyone that seems to know what the ZD stands for.

 

[os2wiz]

ZD doesn't mean engineering or testing sample, Qualification samples are = to retail samples.

I'm going to go throught this again.

Qualification samples are sent to motherboard manufactures to simulate what the retail samples will be for them to finalize boards, there is almost no difference if any between Qualification and retail samples.

I have here with me, ES of GPU's and had one of Intel CPU's (old P4) till last month when I cleaned out my basement. All of them are very close to retail releases.

This is not something new to me.

If you contact either AMD or Intel and tell them what you are doing, like making a motherboard, they will give you qualification samples which are retail samples without warranty. Qualification samples are the same as Retail. They have to be, because a manufacturer can not base their hardware on a chip that has unknowns.

That leaves one problem unanswered and you can not deny the board he has to have an ancient bios. It is a bootleg motherboard not final retail with an ancient bios . Does that not send warning signals about the memory throughput, since it was known that early bios had severe problems handling memory. We do not know what revision the motherboard is. it most likely is a pre-production board. This just doesn't smell right. But of course troll Juanrga and haters et al jump on this knowing it is far too soon to accept that Iranian report as anything to concerned about, more likely to be laughed off.

 

[razor1]

That leaves one problem unanswered and you can not deny the board he has has to be an ancient bios. It is a bootleg motherboard not final retail with an ancient bios does that not send warnign signals about the memory throughput, since it was known that early bios had severe problems handling memory. We do not know what revision the motherboard is. it could be a pre-production board. This just doesn't smell right. But of course troll Juanrga and haters et al jump on this knowing it is far too soon to accept that Iranian report as anything to concerned about, more likely to be laughed off.

So say you peace and leave it at that, he stated what he though was it.

The press deck that has just been leaked doesn't look too good as it kinda matches up with that review :/, of course not plainly but over all.

Don't expect magic over the press deck benchmarks..... if anything expect 5 to 10% in the direction of Intel.

IPC lower than Boardwell, Multithreading performance AMD is well above Broadwell.

So gaming is going to be a crap shot, win some loss some. Better for the future.

Overclocking well, the press deck says it all, 3.7 on a x1800 max on all cores, XFR 4.1 one single core.

This tells us everything, why they didn't show anything but cinebench, and hand brake lol, cause they couldn't, or wouldn't cause it would show weakness to current Intel Line up, of course price is in their favor Big Time, and the future is better for AMD with their SMT performance, but they can't really show that yet so..... Lets leave a big ? for people figure out. Its not a ? when they leave obvious things out.

 

[sover]

Time to dig up that old compute job on the heatsinks.

Not really, as long as you can decrease the distance between them keeping thickness constant. It will rise the airflow requirements and make it way heavier, obviously, but effect is well specified. Granted, properly verifying myself requires stuff i do not have on hand (the decent model, that is), so i'll take your word for it.

What you seem to be saying is that if you change multiple variables you can end up with a result that seems favorable to your point. To put it simply, heat transferred is directly proportional to the size of the conduit and the conductivity of the conduit. More material shaped appropriately will transfer more heat. More material shaped inappropriately won't transfer more heat. Less material will transfer less heat if all other variables (shape, air flow, temperature delta, material, etc) are kept constant.

Sure you might be able to make a smaller heatsink with more fins dissipate more heat if you can push a few hundred cubic feet per minute over it, but that's a ridiculous idea when you can form a heatsink in a more optimal way and have a quiet lower flow fan.

 

[Riptide_NVN]

IPC lower than Boardwell, Multithreading performance AMD is well above Broadwell.
So gaming is going to be a crap shot, win some loss some. Better for the future.
Overclocking well, the press deck says it all, 3.7 on a x1800 max on all cores, XFR 4.1 one single core.

Yah doesn't look like a slam dunk decision over a 6850K. Couple more cores is about all it has over it. Worse overclocking potential. Less IPC. Gonna trade blows with a 6850K in some apps.

Given the fact the 6850K isn't exactly much better it doesn't make me feel like I've made a mistake on ordering the Ryzen.

Final judgement reserved for tomorrow AM.

 

[lolfail9001]

What you seem to be saying is that if you change multiple variables you can end up with a result that seems favorable to your point.

Good thing there is way more variables either of us can control for in our heads, eh.

. To put it simply, heat transferred is directly proportional to the size of the conduit and the conductivity of the conduit.

Yeah, i know of that PDE.

Sure you might be able to make a smaller heatsink with more fins dissipate more heat if you can push a few hundred cubic feet per minute over it, but that's a ridiculous idea when you can form a heatsink in a more optimal way and have a quiet lower flow fan.

But we do not care about noise here, do we? After all, we do not know what noise that gammaxx knock off produces.

 

[sover]

Good thing there is way more variables either of us can control for in our heads, eh.

Yeah, i know of that PDE.

But we do not care about noise here, do we? After all, we do not know what noise that gammaxx knock off produces.

Good to know we're all in agreement here now that heatsink mass matters.

 

[os2wiz]

Yah doesn't look like a slam dunk decision over a 6850K. Couple more cores is about all it has over it. Worse overclocking potential. Less IPC. Gonna trade blows with a 6850K in some apps.

Given the fact the 6850K isn't exactly much better it doesn't make me feel like I've made a mistake on ordering the Ryzen.

Final judgement reserved for tomorrow AM.

IPC is just a notch below Broadwell much closer to Broadwell than Haswell.

 

[lolfail9001]

Good to know we're all in agreement here now that heatsink mass matters.

Not mass as much as volume and properties of material used. Yeah, nitpicking i know, but mass still has formally nothing to do with it.

 

[razor1]

Good to know we're all in agreement here now that heatsink mass matters.

Mass has nothing to do with heat transfer........

Thermal conductivity, only has direct correlation, with the ability of a material to absorb a certain amount of heat quickly and that dissipating that heat quickly.

This is why water is one of the best thermal conductors, its specific heat capabilities are above most metals.

So the mass of water doesn't matter, even with little amounts of it as long as its being transported away from the heat source, can cool much better than a block of copper. Granted can't use water lol, but there are ways to "water cool", well submersion techniques that rival or get better results than LN2, by using organo-phosphate liquids.

 

[sover]

Mass has nothing to do with heat transfer........

Thermal conductivity, only has direct correlation, with the ability of a material to absorb a certain amount of heat quickly and that dissipating that heat quickly.

If you want to get into semantics it's often characterized as a function of component sizes and thicknesses, which of course is the shape by the mass, but if you want to argue that I can understand where you're coming from, but I don't subscribe to it.

 

[sover]

Mass has nothing to do with heat transfer........

Thermal conductivity, only has direct correlation, with the ability of a material to absorb a certain amount of heat quickly and that dissipating that heat quickly.

This is why water is one of the best thermal conductors, its specific heat capabilities are above most metals.

So the mass of water doesn't matter, even with little amounts of it as long as its being transported away from the heat source, can cool much better than a block of copper. Granted can't use water lol, but there are ways to "water cool", well submersion techniques that rival or get better results than LN2, by using organo-phosphate liquids.

What? So it doesn't matter if you have a drop or a pint of water in your water cooler? Man, I hadn't a clue.

 

[razor1]

If you want to get into semantics it's often characterized as a function of component sizes and thicknesses, which of course is the shape by the mass, but if you want to argue that I can understand where you're coming from, but I don't subscribe to it.

Size, thickness all that have nothing to do with it.

 

[mdaniel]

It's nice to know I can come here and have my geek meter recalibrated down to the low end of the spectrum. I've built my own computers for more than 30 years and have been internationally certified in computer forensics...but some of these discussions leave me thinking I'm reading Greek.

 

[razor1]

What? So it doesn't matter if you have a drop or a pint of water in your water cooler? Man, I hadn't a clue.

if you have a metal or alloy that has a high or higher specific heat, that is all that matters, its size, thickness, shape, doesn't matter as long as its fits flush with the component you are trying to cool. You may need more size or different shapes to dissipate the heat accordingly by what ever means, liquid or air.

Also depending on the alloy or metal, it needs stability for itself so it might need to be bigger or heavier.

Now lets get into phase changes, do you know why its worse to get burned by 100 c of water vapor than putting your hand on 100 c metal plate?

Guess which one is carrying more potential energy.

 

[Simplyfun]

Not that that this thread hasn't been radically derailed multiple times, but, no one gives a shit about the mass of air coolers for the cutting edge of processing. I still use a cascade sometimes and I think that's pretty old school of me.

If my CPU is air cooled it's a cherry trail and I'm pissed about the fan noise.

 

[Schmave]

So actually that leaked slide deck looks to be about what I was expecting. In multi-threaded programs Ryzen can win or be up there with the 6800K and 6900K, and it is a little behind in single threaded performance compared to some of the newer Intel processors. It's not bad at all, especially since I would consider the results in gaming "good enough" for my standards.

I do a fair amount of photo and video editing on my computer in addition to gaming, so I would likely choose Ryzen if I were putting together a new build. The price/performance ratio in heavily threaded programs is very good, especially for the 1700 and 1700X.

I bet this Leynar Khayrullin person is getting a phone call from AMD right about now...

 

[Simplyfun]

I bet this Leynar Khayrullin person is getting a phone call from AMD right about now...

The thing about marketing is that Leynar doesn't have to actually exist. If it seems that a leak sprouted out of nowhere AMD can shrug their shoulders when asked by folks like Kyle and say "out of our control" even when it costs him hits.

 

[OrangeKhrush]

There is some good news tomorrow

 

[Grimlaking]

Mass has nothing to do with heat transfer........

Thermal conductivity, only has direct correlation, with the ability of a material to absorb a certain amount of heat quickly and that dissipating that heat quickly.

This is why water is one of the best thermal conductors, its specific heat capabilities are above most metals.

So the mass of water doesn't matter, even with little amounts of it as long as its being transported away from the heat source, can cool much better than a block of copper. Granted can't use water lol, but there are ways to "water cool", well submersion techniques that rival or get better results than LN2, by using organo-phosphate liquids.

Ok you are ALMOST on the money here but I need to dispel some items or put in some corrections.

1. Water is a great agent in a closed loop because it is a non compressible fluid with a small specific surface area that can flow through a internal water block and get maximum contact on the internals and transport that up to flow through a radiator depositing the majority of it's heat to the cooler surfaces present that in turn dissipates the heat to the air.

2. A good "air" cooler is still transporting the heat away from the CPU and allowing it to be transported into the air. BUT the potential for a negative heat soak situation is greater.

The reasons behind this are a couple.

A. The Air Cooler typically is exhausting it's heat into your case. There by creating a potential to recycle that same hot air through itself, AND transfer the heat to surrounding components. If it manages to do this long enough with poor enough air flow in the case you will have heat soak. The disability to exhaust heat.

B. Water blocks are considered better not because they dissipate heat better but because they dissipate it outside of the case.

This is one of the reasons even in modern server farms they use standard heat sinks along with fins aligned to air flow. Metal to air in a controlled environment is still the best Dollar to degree methodology to dissipate heat.

It's so efficient that many facilities are being built to allow for only outside air flow even in what we would consider very hot environments. No active cooling at all.


I hope that explains things a bit more.