AMD Ryzen 5 1600 & 1400 CPU Review @ [H]

[KazeoHin]

you answered your own question.. more ryzen sales = more reasons to fix their drivers so that those buying ryzen processors don't also buy AMD GPU's..

ah yes, but you see, If Ryzen is seen as "bad for gaming", less people will buy it.

Thus Nvidia wants people to perceive AMD's CPU as "Bad for gaming".

 

[raghu78]

Nice to see a recent resident migrate back to his watering hole called Semi accurate, his IPC ~ Sandy arguments are getting no love at all there, won't be long before the permaban hits, then he will be back here sadly. I mean for a guy that was so high on cinebench for finding IPC he now avoids it like a cancer patient, now he tries to find IPC on games, oh dear.

Excuse me sir but, Cinebench keeps showing IPC is Haswell/Broadwell level, what AMD actually have an issue iwth is clock speed.

very true. lesser said about that person. the better it is. AMD needs a better process to hit higher clocks. Its disappointing that the first gen Zen max out at 4 - 4.1 Ghz. I think we will see some custom tweaks to 14LPP to allow much higher frequency. I am thinking Zen 2 based Pinnacle Ridge will hit 4 Ghz base and 4.5 Ghz turbo along with some IPC improvements. Some very obvious areas for improvement are fabric speeds and cache latency.

 

[Olle P]

I'd definitely consider the 1500X to be better than the 1400 for a non power user.
For just $20 more you get:

• Better (quieter) cooler.
• Higher clock speed. (Without OC)
• Twice as much L3 cache.

In the comparison 1600 vs 1700 one must consider that the 1700 offer 33% more theoretical performance at a 45% price premium.
(The relative price difference does go down if you include motherboard and after market cooler and whatever else.)

 

[Nihilus1]

CPUs are hitting a wall. AMD just reached the wall now with Ryzen, but Intel been stuck here.

Intel has been stuck at that wall and has kept the prices sky high. Ryzen 7 is the reason we will not see an 8700k for $350+. Ryzen 5 is the reason we will not see an 8600k for $250. It takes competition to move the industry forward.

----------------------

Can't wait until they release some quality ITX boards.

 

[SighTurtle]

Late to this conversation, but what incentive does Nvidia have to optimize their drivers for Ryzen? What ACTUAL monetary benefit does that action entail? Spoiler-alert: Nvidia has no reason to or even a responsibility to optimize for their direct competitor.

More Ryzen sales = more revenue for AMD.

More revenue for AMD = More potential R&D money for RTG

More R&D money for RTG = only bad things for Nvidia's bottom line.

Nvidia should be doing everything they can to make EVERY AMD product flop if they want to keep their position.

We don't know if it's a Nvidia problem, only that's it's cropping up but it could just be a game related problem.

Intel has been stuck at that wall and has kept the prices sky high. Ryzen 7 is the reason we will not see an 8700k for $350+. Ryzen 5 is the reason we will not see an 8600k for $250. It takes competition to move the industry forward.

----------------------

Can't wait until they release some quality ITX boards.

Oh no doubt about that, if we are stuck at a wall, better Intel and amd than just Intel.

 

[juanrga]

Nice to see a recent resident migrate back to his watering hole called Semi accurate, his IPC ~ Sandy arguments are getting no love at all there, won't be long before the permaban hits, then he will be back here sadly. I mean for a guy that was so high on cinebench for finding IPC he now avoids it like a cancer patient, now he tries to find IPC on games, oh dear.

Excuse me sir but, Cinebench keeps showing IPC is Haswell/Broadwell level, what AMD actually have an issue iwth is clock speed.

On average the IPC of Zen is closer to Sandy than to the on-pair-with-Kabylake predicted by residents at SA forums. Even in a favorable workload as CineBench, Zen IPC is a noticeable 8--9% behind Broadwell

 

[OrangeKhrush]

On average the IPC of Zen is closer to Sandy than to the on-pair-with-Kabylake predicted by residents at SA forums. Even in a favorable workload as CineBench, Zen IPC is a noticeable 8--9% behind Broadwell

Sandy 2600K at 3.5Ghz = 124

Differentials

Ryzen to 2600K = 15

Ryzen to 6900K = 12

Yip closer to Sandy alright, this is probably why you need a 4.5Ghz+ Sandy just to match a stock Ryzen

 

[Master_shake_]

On average the IPC of Zen is closer to Sandy than to the on-pair-with-Kabylake predicted by residents at SA forums. Even in a favorable workload as CineBench, Zen IPC is a noticeable 8--9% behind Broadwell

that graph is weird.

it looks like they held the 6900k at 3.5 ghz with no turbo boost

but if we look at this graph with the 6900k definitely hitting it's turbo of 3.7 it's lower than the one you posted.

 

[OrangeKhrush]

that graph is weird.

it looks like they held the 6900k at 3.5 ghz with no turbo boost

but if we look at this graph with the 6900k definitely hitting it's turbo of 3.7 it's lower than the one you posted.

Very true, well spotted, since CB15 measures top of turbo score you have to factor in 3.7 or even as proven in the AMD live show that the 6900K was hitting 4 on turbo 3.0

 

[OrangeKhrush]

Looked up anandtechs database, the 6900K scores 155 at 3.7ghz per CB15 ST, so at 3.5 it only scores 146.62 making the argument that it is closer to Sandy pretty much a dead end argument. This is the reason why in quite a few IPC / clockspeed games the Ryzen 7's can match or beat the 5960 and 6900K at stock. Greater SMT off sets the IPC limitation given up to Broadwell.

In COD the 1600 was per computerbase.de the top chip for that game, beating the 7700K so GPU limits are not a valid argument. It is actually more how the games are coded as to whether you get real representative performance or broken performance ie: Battlefield is really good for Ryzen but Tomb Raider is broken, COD is good, Hitman is not.

 

[juanrga]

Sandy 2600K at 3.5Ghz = 124

Differentials

Ryzen to 2600K = 15

Ryzen to 6900K = 12

Yip closer to Sandy alright, this is probably why you need a 4.5Ghz+ Sandy just to match a stock Ryzen

8--9% behind Broadwell on CB15 implies RyZen IPC is closer to Sandy than Skylake

131 / 116 ==> 13% gap Skylake over RyZen

116 / 108 ==> 7% gap RyZen over Sandy.

 

[juanrga]

that graph is weird.

it looks like they held the 6900k at 3.5 ghz with no turbo boost

but if we look at this graph with the 6900k definitely hitting it's turbo of 3.7 it's lower than the one you posted.

The graph you give is weird. It puts the 6900k single-thread score behind the 1700X, when it is just the contrary, with the 6900k beating not only the 1700x but the 1800X as well

Moreover, your graph is internally inconsistent for RyZen chips. Ryzen @3.7GHz scores 145 whereas @3.8GHz scores 154, which cannot be correct because 145 * 3.8/3.7 = 149.

 

[FrgMstr]

Have to disagree. The 7700K is the fastest in games today and by the time the market catches up in a few years (at the earliest) I'll just build another system then. It took forever for software developers to get on the x64 bandwagon after the original Athlon 64 came out and I'm betting it will take even longer for stuff to take advantage of 6+ physical cores. There's still a ton of software and games that don't even use 4!

Yeah. Been a member for almost 17 years, but don't read the articles.

I just came to a different conclusion than you. I'm willing to admit I might be wrong. Time will tell.

You sure about that? You disagree...and come to a different conclusion? Here is what I wrote in our first Ryzen review, and it still stands.

"If you are building a PC today that is going to be used for nothing but desktop gaming, I would suggest you buy a 7600K or 7700K and overclock those and enjoy the performance you will be getting with those."

That said, for an all around forward looking machine, I would not purchase a 4-Core CPU if I was going to be keeping it for a while.

 

[Master_shake_]

The graph you give is weird. It puts the 6900k single-thread score behind the 1700X, when it is just the contrary, with the 6900k beating not only the 1700x but the 1800X as well

Moreover, your graph is internally inconsistent for RyZen chips. Ryzen @3.7GHz scores 145 whereas @3.8GHz scores 154, which cannot be correct because 145 * 3.8/3.7 = 149.

and then here is anandtechs

 

[OrangeKhrush]

8--9% behind Broadwell on CB15 implies RyZen IPC is closer to Sandy than Skylake

131 / 116 ==> 13% gap Skylake over RyZen

116 / 108 ==> 7% gap RyZen over Sandy.

Your math is so wrong its embarassing, you posted that two months ago and ryzen cores got 119-120 now down to 116. You are being underhanded about your totals to try skew Results.

8--9% behind Broadwell on CB15 implies RyZen IPC is closer to Sandy than Skylake

131 / 116 ==> 13% gap Skylake over RyZen

116 / 108 ==> 7% gap RyZen over Sandy.

Your numbers are dishonest. Taking direct numbers off Anandtech database gives you single thread with max single turbo unaffected by turbo 3/XFR. The numbers i gave are direct off database.

Sandy 106
Ivy 110
Ryzen 119
Haswell 120
Devils canyon 123
Broadwell 125
Kaby 128
Skylake 129

Either way you slice it you are wrong. Your numbers look blatently alterered.

Eg 6700K 181

181 * 3 / 4.2 = 129

4790K 180

180 * 3 / 4.4 = 123

It bugs me, the need to be dishonest. IPC is not AMD's issue, its low base clocks. It competes with similar clocked intel parts but lags behind anything from 4790K and newer K chips.

Also it is 2-3 games that exhibit bad behaviour which implies coding issues, call of duty showed a clear win for the 1600X,. So again games that scale to clock speed AMD lags but all scorea are well over the game line so gaming is still very smooth. I can organise a friend brings a few AMD parts to go up against my 4790, i have no doubts those Ryzen parts will be better

 

[Master_shake_]

The graph you give is weird. It puts the 6900k single-thread score behind the 1700X, when it is just the contrary, with the 6900k beating not only the 1700x but the 1800X as well

Moreover, your graph is internally inconsistent for RyZen chips. Ryzen @3.7GHz scores 145 whereas @3.8GHz scores 154, which cannot be correct because 145 * 3.8/3.7 = 149.

talking about inconsistencies.

how is it you found a benchmark where a 6900k scores a 171 and 154?

that is just not possible.

especially on this page alone there are 4 that say it's not possible.

who's disingenuous now?

sure it's slower than a 1700x.... if you skew the results.

 

[juanrga]

Your math is so wrong its embarassing, you posted that two months ago and ryzen cores got 119-120 now down to 116.

What? This is kindergarden-level math

139/151 * 126 = 115.98...

which I rounded to 116.

Your numbers are dishonest. Taking direct numbers off Anandtech database gives you single thread with max single turbo unaffected by turbo 3/XFR.

LOL I am giving you performance at fixed 3.5GHz (PcPer graph) and at fixed 3GHz (Anandtech graph), turbo and XFR play no role here. Stop from looking for weird excuses.

 

[juanrga]

talking about inconsistencies.

how is it you found a benchmark where a 6900k scores a 171 and 154?

that is just not possible.

Your graph is internally inconsistent because the same site is giving two incompatible values for RyZen.

One site giving a score of 154 for the 6900k and another site giving 171 is perfectly compatible because one site is using turbo 2 and the other is using turbo 3. Indeed

154 * 4 / 3.7 = 166.49

which is only 2.7% behind the score of 171. And a 2.7% discrepancy is within the margin of error of measurements.

 

[OrangeKhrush]

What? This is kindergarden-level math

139/151 * 126 = 115.98...

which I rounded to 116.



LOL I am giving you performance at fixed 3.5GHz (PcPer graph) and at fixed 3GHz (Anandtech graph), turbo and XFR play no role here. Stop from looking for weird excuses.

148 * 3 / 3.7 = 120

Per pcper graph of the 1700 max turbo 3.7ghz single core.

Again a far cry from your random number which conveniently lowers the number. MOE numbers usually upscale not down scale as you cannot theoretically lower what the baseline IPC is capable of at any clock.

151/139 * 126 so you this equation for ryzen but the universally accepted equation for the 6900 to show MOE, kudos.

I have a huge issue with people that go out of their way to skew one side.

 

[FrgMstr]

Your graph is internally inconsistent because the same site is giving two incompatible values for RyZen.

One site giving a score of 154 for the 6900k and another site giving 171 is perfectly compatible because one site is using turbo 2 and the other is using turbo 3. Indeed

154 * 4 / 3.7 = 166.49

which is only 2.7% behind the score of 171. And a 2.7% discrepancy is within the margin of error of measurements.

One of the reasons we always lock in clocks on cores. With dynamic clocks and cores how can you know what you are truly looking at?

 

[FrgMstr]

Actually that statement and the benchmarks in your original R7 review is *one* of the reasons I built my new 7700K system, as I use my home PC for gaming and general internet stuff. I'm not sure why the sudden change of heart, but it doesn't matter. If anything, I believe coding lazyness and general market resistance to change will prevent an overnight shift to 6+ core code optimization. Like I said, I might be wrong, am prepared to eat crow, but right now that's the way I see it.

Fair deal on that, brother! Neither of us are wrong as the entire point is subjective. For a purpose built gaming machine, a 7600k or 7700k at 5GHz is not going to be toppled any time soon.

 

[juanrga]

148 * 3 / 3.7 = 120

Per pcper graph of the 1700 max turbo 3.7ghz single core.

Again a far cry from your random number which conveniently lowers the number. MOE numbers usually upscale not down scale as you cannot theoretically lower what the baseline IPC is capable of at any clock.

151/139 * 126 so you this equation for ryzen but the universally accepted equation for the 6900 to show MOE, kudos.

I have a huge issue with people that go out of their way to skew one side.

What is so difficult to understand? Using the benchmark you mentioned, the IPC is 8--9% lower than Broadwell

And 8--9% behind Broadwell means that the IPC is closer to Sandy than Skylake in this bench

 

[raghu78]

What is so difficult to understand? Using the benchmark you mentioned, the IPC is 8--9% lower than Broadwell

And 8--9% behind Broadwell means that the IPC is closer to Sandy than Skylake in this bench

You have to have something constant to compare two different graphs. In the pcper bench 7700k at 3.5 Ghz shows itself to be 10% faster than the Ryzen at 3.5 Ghz. In the anandtech bench the 6700k which uses the same Skylake core and is pretty much the same chip is 22% faster than 2600k Sandy at 3 Ghz. A Ryzen at same clocks will effectively sit right in between at 120 (120 * 1.1 = 132). This proves Zen to be exactly mid-way between Sandy and Kabylake and pretty much on par with Haswell. When comparing cores for IPC we need to try and restrict the number of differing parameters like memory clocks, memory channels etc . There is room for a lot of creative manipulation when comparing different processors from different charts when trying to arrive at IPC conclusions for a workload.

Anyway here is another bench which locks Kabylake and Ryzen at 4 Ghz and the difference is even smaller. Kabylake is just 7.5% faster. The only caveat is this is a single CCX config arrived by disabling 1 CCX from Ryzen. Anyway we are looking at single thread performance here. Memory speeds were matched at DDR4 2400 CL15.

http://www.zolkorn.com/en/amd-ryzen...i7-7700k-mhz-by-mhz-core-by-core-en/view-all/

Its quite clear that Ryzen is very much a Haswell class core with few weak points like gaming performance (due to slow fabric speeds and higher L3 cache latency) and AVX performance (due to 128 bit load/store). You can expect Ryzen performance to improve with platform maturity as Ryzen loves high speed memory especially in games.

https://www.pcper.com/reviews/Proce...Core-i5/Memory-Speed-Scaling-and-Windows-10-P

 

[udneekgnim]

Late to this conversation, but what incentive does Nvidia have to optimize their drivers for Ryzen? What ACTUAL monetary benefit does that action entail? Spoiler-alert: Nvidia has no reason to or even a responsibility to optimize for their direct competitor.

More Ryzen sales = more revenue for AMD.

More revenue for AMD = More potential R&D money for RTG

More R&D money for RTG = only bad things for Nvidia's bottom line.

Nvidia should be doing everything they can to make EVERY AMD product flop if they want to keep their position.

if someone buys a Ryzen CPU and they're looking to get a GPU in the 1060 / 480 range, then the 480 is the better choice

anything higher than that is Nvidia because AMD does not have a good answer

this hopefully changes whenever AMD gets around to releasing Vega and then we'll see if Nvidia fixes their drivers for Ryzen

 

[cybereality]

Plus cost of HSF. Get a crappy fan then I think the 1600 could match with its included. Buy a real HSF then you are in 1700 territory. If you are willing to spend bucks on a HSF the buy the 1600 and go 4Ghz. I don't understand the "X" models at all. Pay extra for no HSF? They all seem to hit the same mark. Am I right?

Oh man! I didn't even realize the 1600X is missing the HSF. I should try reading more. Knowing that may have affected my decision, but now I guess I can get a better fan. You saved me!

 

[OrangeKhrush]

You have to have something constant to compare two different graphs. In the pcper bench 7700k at 3.5 Ghz shows itself to be 10% faster than the Ryzen at 3.5 Ghz. In the anandtech bench the 6700k which uses the same Skylake core and is pretty much the same chip is 22% faster than 2600k Sandy at 3 Ghz. A Ryzen at same clocks will effectively sit right in between at 120 (120 * 1.1 = 132). This proves Zen to be exactly mid-way between Sandy and Kabylake and pretty much on par with Haswell. When comparing cores for IPC we need to try and restrict the number of differing parameters like memory clocks, memory channels etc . There is room for a lot of creative manipulation when comparing different processors from different charts when trying to arrive at IPC conclusions for a workload.

Anyway here is another bench which locks Kabylake and Ryzen at 4 Ghz and the difference is even smaller. Kabylake is just 7.5% faster. The only caveat is this is a single CCX config arrived by disabling 1 CCX from Ryzen. Anyway we are looking at single thread performance here. Memory speeds were matched at DDR4 2400 CL15.

http://www.zolkorn.com/en/amd-ryzen...i7-7700k-mhz-by-mhz-core-by-core-en/view-all/

Its quite clear that Ryzen is very much a Haswell class core with few weak points like gaming performance (due to slow fabric speeds and higher L3 cache latency) and AVX performance (due to 128 bit load/store). You can expect Ryzen performance to improve with platform maturity as Ryzen loves high speed memory especially in games.

https://www.pcper.com/reviews/Proce...Core-i5/Memory-Speed-Scaling-and-Windows-10-P

He has played the wall of ignorance card, no logic can penetrate it. At this point i feel like i am being trolled

What is so difficult to understand? Using the benchmark you mentioned, the IPC is 8--9% lower than Broadwell

And 8--9% behind Broadwell means that the IPC is closer to Sandy than Skylake in this bench

So lets do this again

Sandy 106
Ivy 110
Ryzen 120
Haswell 120
Devils Canyon 123
Broadwell 125
Kaby 128
Skylake 129

Diff to Broadwell - 3-4%
Diff from sandy - 8-9%

I think surely even you can figure this one out

 

[KazeoHin]

He has played the wall of ignorance card, no logic can penetrate it. At this point i feel like i am being trolled



So lets do this again

Sandy 106
Ivy 110
Ryzen 120
Haswell 120
Devils Canyon 123
Broadwell 125
Kaby 128
Skylake 129

Diff to Broadwell - 3-4%
Diff from sandy - 8-9%

I think surely even you can figure this one out

I don't think many of you understand, these Intel fanboys start with a belief, then they seek out evidence to justify it.

They believe Zen has the same IPC as Sandy Bridge. They won't believe otherwise, and they'll ignore 10 benchmarks that say otherwise and link a single one that agrees with them.

 

[OrangeKhrush]

I don't think many of you understand, these Intel fanboys start with a belief, then they seek out evidence to justify it.

They believe Zen has the same IPC as Sandy Bridge. They won't believe otherwise, and they'll ignore 10 benchmarks that say otherwise and link a single one that agrees with them.

It is just so silly to be honest, how a person can have such blind devotion to a corporate. I use Intel because at the time there was nothing from AMD that could give me what I needed, that is not to say that right now how it stands a Ryzen 5 1600 doesn't scream out winning to my needs, I am also not bigot enough to blindly devout myself to team Intel. similarly I use a GTX 980ti, first because nothing from AMD's current lineup gets close and secondly it cost me the conversion equivilant of $250 in my currency and since Battlefield is my 99% percentile addiction it owns battlefield at 144hz. Does that mean I wouldn't consider Vega if it was compelling?

I like AMD IP's, I think in terms of innovation AMD are fantastic, the issue is implementation and money is normally the limiting factor.

 

[sirmonkey1985]

ah yes, but you see, If Ryzen is seen as "bad for gaming", less people will buy it.

Thus Nvidia wants people to perceive AMD's CPU as "Bad for gaming".

that's an idiotic way of looking at it.. bottom line nvidia only gives a crap about how much money they make, they don't care about anyone else. AMD even in their best year will never come close to making as much money as nvidia and locking themselves out of a platform because they want to make AMD look bad is a terrible way to run a business and to treat their customers. they already tried it with AM2+ when they tried to force AMD to pay an asinine amount of money to them to have SLI support on the AMD chipset because nvidia stopped producing chipsets and it screwed nvidia hardcore. they won't make that mistake again. i suspect the next WHQL drivers will fix any of the driver issues people are seeing.

 

[raghu78]

I don't think many of you understand, these Intel fanboys start with a belief, then they seek out evidence to justify it.

They believe Zen has the same IPC as Sandy Bridge. They won't believe otherwise, and they'll ignore 10 benchmarks that say otherwise and link a single one that agrees with them.

Very true.

 

[juanrga]

So lets do this again

Sandy 106 -----> 108
Ivy 110 -----> 111
Ryzen 120 or 116
Haswell 120 -----> 121
Devils Canyon 123 -----> 121
Broadwell 125 -----> 126
Kaby 128 -----> 131
Skylake 129 -----> 131

Fixed it for you. It is weird that you give different values for Haswell and Devils Canyon when both are the same muarch. or for KBL and SKL when both are the same muarch.

You have to have something constant to compare two different graphs. In the pcper bench 7700k at 3.5 Ghz shows itself to be 10% faster than the Ryzen at 3.5 Ghz. In the anandtech bench the 6700k which uses the same Skylake core and is pretty much the same chip is 22% faster than 2600k Sandy at 3 Ghz. A Ryzen at same clocks will effectively sit right in between at 120 (120 * 1.1 = 132). This proves Zen to be exactly mid-way between Sandy and Kabylake and pretty much on par with Haswell. When comparing cores for IPC we need to try and restrict the number of differing parameters like memory clocks, memory channels etc . There is room for a lot of creative manipulation when comparing different processors from different charts when trying to arrive at IPC conclusions for a workload.

Anyway here is another bench which locks Kabylake and Ryzen at 4 Ghz and the difference is even smaller. Kabylake is just 7.5% faster. The only caveat is this is a single CCX config arrived by disabling 1 CCX from Ryzen. Anyway we are looking at single thread performance here. Memory speeds were matched at DDR4 2400 CL15.

http://www.zolkorn.com/en/amd-ryzen...i7-7700k-mhz-by-mhz-core-by-core-en/view-all/

Its quite clear that Ryzen is very much a Haswell class core with few weak points like gaming performance (due to slow fabric speeds and higher L3 cache latency) and AVX performance (due to 128 bit load/store). You can expect Ryzen performance to improve with platform maturity as Ryzen loves high speed memory especially in games.

https://www.pcper.com/reviews/Proce...Core-i5/Memory-Speed-Scaling-and-Windows-10-P

You use Skylake as baseline and get 120 points. I am using Broadwell as baseline and get 116 points. The differences are due to measurement uncertainties and pretending that your viewpoint is more correct is just silly.

Sure you can find a review as zolkorn where the gap between between RyZen and KBL is reduced 3% on CB, and then I can mention ExtremeTech, which increases the gap again to 10%. Again those minor discrepancies are easily explained by measurement uncertainties. We can also consider another workload and see that KBL is about 30% ahead of RyZen clock for clock

Or we can consider an average of different applications and find that RyZen is 10--20% slower than Broadwell clock-for-clock

As you know well I have been for years claiming "IPC+SMT ~ Haswell".

RyZen is Haswell-like on total throughput. It is Sandy-like on general IPC and it is Sandy-like in SIMD code (16FLOP/core for Sandy and RyZen and 32FLOP/core for Haswell).

 

[juanrga]

I don't think many of you understand, these Intel fanboys start with a belief, then they seek out evidence to justify it.

They believe Zen has the same IPC as Sandy Bridge. They won't believe otherwise, and they'll ignore 10 benchmarks that say otherwise and link a single one that agrees with them.

No one said it has "the same IPC". And the claims are not based in a single bench, but on an average of several dozens of them plus analysis of the microachitectures such as for instance 2x128bit FMAC vs 2x256bit FMAC or 6-wide vs 8-wide.

Since this thread is about 1600 and 1400, you can find here the average gaming performance of the 1400. It is between SB i5 and i7

and here average compute performance with 1400 virtually at same level than SB i7

 

[KazeoHin]

No one said it has "the same IPC". And the claims are not based in a single bench, but on an average of several dozens of them plus analysis of the microachitectures such as for instance 2x128bit FMAC vs 2x256bit FMAC or 6-wide vs 8-wide.

Since this thread is about 1600 and 1400, you can find here the average gaming performance of the 1400. It is between SB i5 and i7

and here average compute performance with 1400 virtually at same level than SB i7

Wow, it beats a sandy with a ~10% clock DISSADVANTAGE.

you're not even providing evidence that backs up your claims...

 

[KazeoHin]

that's an idiotic way of looking at it.. bottom line nvidia only gives a crap about how much money they make, they don't care about anyone else. AMD even in their best year will never come close to making as much money as nvidia and locking themselves out of a platform because they want to make AMD look bad is a terrible way to run a business and to treat their customers. they already tried it with AM2+ when they tried to force AMD to pay an asinine amount of money to them to have SLI support on the AMD chipset because nvidia stopped producing chipsets and it screwed nvidia hardcore. they won't make that mistake again. i suspect the next WHQL drivers will fix any of the driver issues people are seeing.

The questionis simply: what does Nvidia have to gain by making their drivers work well with Ryzen? People already look at Ryzen as a "terrible CPU for games", so why would Nvidia want to put in ANY man-hours to change that?

Its an idiotic idea to think Nvidia DOESN'T plan ahead and plot out their best course of action. You think their near-monopoly status and mountain of cash was caused by them making short--term, self-defeating cash-grabbing decisions?

 

[FrgMstr]

The questionis simply: what does Nvidia have to gain by making their drivers work well with Ryzen? People already look at Ryzen as a "terrible CPU for games", so why would Nvidia want to put in ANY man-hours to change that?

Its an idiotic idea to think Nvidia DOESN'T plan ahead and plot out their best course of action. You think their near-monopoly status and mountain of cash was caused by them making short--term, self-defeating cash-grabbing decisions?

What is wrong with you people? Tinfoil hat fall off?

 

[legcramp]

gl selling AMD hardware with that attitude.

And who's that guy up there that keeps posting different random benchmarks every time he gets his previous post shutdown.. poor sap.

 

[OrangeKhrush]

No one said it has "the same IPC". And the claims are not based in a single bench, but on an average of several dozens of them plus analysis of the microachitectures such as for instance 2x128bit FMAC vs 2x256bit FMAC or 6-wide vs 8-wide.

Since this thread is about 1600 and 1400, you can find here the average gaming performance of the 1400. It is between SB i5 and i7

and here average compute performance with 1400 virtually at same level than SB i7

Fixed it for you. It is weird that you give different values for Haswell and Devils Canyon when both are the same muarch. or for KBL and SKL when both are the same muarch.



You use Skylake as baseline and get 120 points. I am using Broadwell as baseline and get 116 points. The differences are due to measurement uncertainties and pretending that your viewpoint is more correct is just silly.



As you know well I have been for years claiming "IPC+SMT ~ Haswell".

RyZen is Haswell-like on total throughput. It is Sandy-like on general IPC and it is Sandy-like in SIMD code (16FLOP/core for Sandy and RyZen and 32FLOP/core for Haswell).

6700K - 181 turbo is 4.2 181 * 3 / 4.2 = 129

4790 180 turbo 4.4 = 123

Ryzen 162 at 4ghz 162 * 3 / 4= 120

2600K 135 @ 3.8 135 * 3 / 3.8 = 106~

All your intel scores go up 1-2 while AMD goes down 4-5 marks, typical BS

Congratulations your predictions were wrong nothing really new the only person that actually believes them is yourself.

Your sandy nonsense is laughable, why do you need a 4.5-4.8ghz sandy to match a 3.6ghz Ryzen core in cinebench? If IPC was equal.

Can't wait for more cherry picking

 

[OrangeKhrush]

gl selling AMD hardware with that attitude.

And who's that guy up there that keeps posting different random benchmarks every time he gets his previous post shutdown.. poor sap.

Are you talking about Juanrga?

 

[DooKey]

SMH at all the fanboys in this thread. Fucking ridiculous of all of you on all sides.

 

[juanrga]

Wow, it beats a sandy with a ~10% clock DISSADVANTAGE.

you're not even providing evidence that backs up your claims...

At contrary, the new data just proves my point... again.

Here is no 4C/4T RyZen still; therefore we have to compare the R5 1400 with the i7 2600k. That is, we have to compare 4C/8T vs 4C/8T and be careful about SMT effects

The chips score 134.7 vs 131.9, respectively, on the compute graph. Correcting for clocks I obtain 151.5 vs 131.9, which means

RyZen IPC+SMT = 1.15 Sandy Bridge

this is Haswell territory, which implies the ancient prediction "IPC+SMT ~ Haswell" was accurate.

Now, one has to consider that RyZen has about 10% better SMT yields than Sandy because the Zen muarch is more distributed. Eliminating the extra yields gives

RyZen IPC = 1.04 Sandy Bridge

which implies the prediction "IPC ~ Sandy" was accurate.

And all this considering only the compute graph, where Ryzen performs better due to the microarchitecture being more optimized for throughuot than for latency. If we consider also the game graph where the R5 1400 is ~10% behind the i7 2600k and if we take an overall average of both compute and gaming the above numbers will be reduced a bit. Conclusion: RyZen performs just as predicted and it is very far from all those that fueled the hype during years.