AMD Ryzen (Summit Ridge) Benchmarks Thread (use new thread)

[AtenRa]

Losing a few points of market share would hurt the bottom line because loss of revenue hurts profits.

If previously I sold $100,000 worth of goods at 60% gross profit margin and it cost me $30,000 to develop and market the goods, then my gross profit margin would be $60,000 and my net profit would be $30,000 (net profit margin of 30%)

If I then sell $90,000 worth of goods at 60% gross profit margin, and it cost me $30,000 to develop and market the goods (these costs are usually pretty independent of volume, but not entirely as there can be some profit-dependent operating expenses), then my net profit is just $24,000 (26.67%) on gross profit of $54,000 (60%).

My gross profit margin percentage in this case did not go down, but my total profit -- and thus net profit margin, the number that actually matters -- did. Investors would not be happy to see Intel just "give away" a "few points of consumer market share" because it would negatively impact their total profit. Intel is even more complex because its gross profit margin depends on factory utilization rates, so losing volumes not only would lose them revenue, but lead to higher cost per wafer, negatively impacting gross profit margin.

With that said, let's get back to the idea of Intel's potential response. Cutting prices does negatively impact revenue because even if you still "get the sale," you sold for lower than you otherwise could have. Your gross profit margin goes down too because gross profit margin is defined as (Revenue - Cost of Good Sold)/Revenue.

This is generally an option of "last resort" when you've got nothing better to respond with.

The better way to respond is to "disguise" the price cut by introducing better products at the previous price points and pushing the older/slower products (or newly branded equivalents) to lower price points. So if your product isn't selling well at $350, you reduce its price to a price point that the market will bear and then you bring out a better product at that $350 price point.

Think of what Intel did with the Kaby Lake Pentium chips. Instead of cutting prices on last year's Core i3, they just put hyperthreading on the Pentium chips to improve their value proposition. The i3 chips got speed boosts as well as an unlocked variant at the previous price points.

This is a price cut, people instead of buying $110 Core i3 6100 will buy $65 Pentium G4560, they will not buy Core i3 7100 at $110. The difference is that 2+2 Kabylake dies are cheaper to make than 2+2 2015 Skylake dies due to manufacturing process cost depreciation + higher yields. So at the end you sell higher volume at lower prices but due to lower COGs you end up having the same profits (If you sell higher volume).

Intel is currently pushing DIY consumers in to two segments, sub $100 with Pentium G4560 ($65) and second segment at $200 and higher (Core i5). There is absolutely no reason at all to even consider any Kabylake Core i3 at this time (except Core i3 7350K for Overclockers).

On the HEDT they could actually launch higher clocked 6+8+10 core SKUs at the same price or lower prices of current SKUs to increase sales. But its the HEDT platform I see Intel having the most problems from ZEN , if 8/16 ZEN can reach i7 6900K performance at lower prices. Because if Intel price cuts the HEDT SKUs it will start to cannibalize its mainstream Core i7s, Core i7 7700K.

 

[AtenRa]

I don't think your scenario is likely (why would AMD charge so little for such a part?), but let's suppose that it plays out. To respond to this, if I were in Intel's position, I'd do the following:

1. Take the Xeon E5 2687v4, a 12 core/24 thread part, unlock the multiplier, disable the QPI links, and sell that in the consumer market as a $999 extreme edition -- Core i7 6970X.
2. Reduce the price of 6950X to $799, possibly with base/boost frequency improvements (+200MHz or so).
3. Keep the 6900K around, but boost the clock frequencies (base/turbo), and rebrand it the 6930K. This becomes a $479 part.
4. Disable 12 PCIe lanes on the 6900K and rebrand it the 6870K. Sell it for $389.
5. Remove the 6850K from the lineup, and sell the 6800K for $329. 7700K would get a price cut to $319 in this case.

This would be a "stopgap" lineup in case of the (very unlikely) scenario that you described above. Then, when Skylake-X comes in, it can occupy those same price points but offer better IPC/clocks/features, helping to strengthen the value proposition of Intel's lineup. Skylake-X LCC die could also have 12 cores natively, so cost structure for Intel would improve on the top end part (which would in the hypothetical Broadwell scenario be handled by a big expensive MCC die).

As i have said above, If you have a HEDT 6-Core SKU at the same price of your mainstream Core i7 7700K then nobody will buy the Quad Core Core i7 7700K.
And im sure Intel doesnt want to sell more 240mm2 6-Core dies at $329 than Quad Core 120mm2 dies at $319.

Intel needs to have a nice $100 overhead at the entry HEDT SKUs like the Core i7 6800K over mainstream Core i7 7700K in order not to cannibalize the later.

 

[bjt2]

Why would 3.6/4.0 Part from AMD offer 7% lower performance than 6900K, if 3.45 GHz 8 core/16 thread offers 3% higher performance than 6900K?

If 8C/16T with 3.4 GHz/3.7 GHz costs 399$, there is no chance that 6C/12T will cost more than 300$. And that part will have higher core clocks than 3.6 GHz.

Guys, look at bigger picture, not only what Intel offers, and might offer, but what AMD might offer. And AMD wants to offer unmatched performance/dollar value.

The new horizon Ryzen was clocked at 3.4GHz (not 3.45) and managed to equal a stock 6900k (at about 3.5GHz) in blender and outperorm it in handbrake (probabily also at 3.5GHz) by 7%.
Obviously a 3.6GHz Ryzen with turbo and more importantly, XFR enabled should draw circle around the 6900K...

 

[bjt2]

Except that a 3.4 GHz Zen doesn't beat an i7 6900K, people are only thinking it does because of a couple of cherry picked benchmarks displayed at New Horizon.

Canard PC's leaks, most likely the closest representative of real world performance thus far, shows Zen IPC at roughly 5-6% behind Broadwell-E. Surely further optimization will help (as we are talking engineering samples) but I wouldn't realistically expect more than 5-10% over what Canard PC displayed considering they would have only several months to improve upon that.

I've always loved AMD but I swear the fanboy hype surrounding this release is just atrocious, far more than the hype surrounding Polaris or Fiji.

Maybe Zen does not beat BWE IPC, but 400Mhz more are 13%, that should more than compensate the 5-10% IPC deficit... And with also -45W TDP...

 

[Lovec1990]

Greetings,

i do not think many potential buyers will care about IPC of Ryzen most care about benchmarks and price as for XFR i think its not classical overclock, but it only rises boost as far your cooling and PSU can handle it so its on demand OC

 

[lolfail9001]

XFR enabled

For tech we have literally 0 factual information about, it is certainly the most hyped thing i have seen yet.

 

[JoeRambo]

For tech we have literally 0 factual information about, it is certainly the most hyped thing i have seen yet.

C'mon, in the past we had things like 'Hyperthreading done right", "reverse hyperthreading", they had way more epic hype attached, were light in details, and in retrospect made very little sense.

XFR is just fine in ratio of hype/info.

 

[KTE]

Obviously a 3.6GHz Ryzen with turbo and more importantly, XFR enabled should draw circle around the 6900K...

Maybe Zen does not beat BWE IPC, but 400Mhz more are 13%, that should more than compensate the 5-10% IPC deficit... And with also -45W TDP...[\b]

Lool


Sent from HTC 10
(Opinions are own)

 

[Lovec1990]

JoeRambo there is very little info abou XFR except for this image:

So based on this image unlike Intels turbo Ryzens Boost can go beyond his limits effectively overclocking itself as long its needed and returning too normal when its not needed

 

[lopri]

^ Yup. Kabylake, my a**. More like a new stepping Skylake. Or rather, Skylake with a new firmware.

That is a false characterization of KBL relative to SKL.

It looks to me that way. I do not quarrel with your disagreement, though, because I understand that there is a degree of subjectivity involved here. But see:

https://gfxbench.com/compare.jsp?benchmark=gfx40&did1=36460673&os1=Windows&api1=gl&hwtype1=iGPU&hwname1=Intel(R)+HD+Graphics+630&did2=26788949&os2=Windows&api2=gl&hwtype2=iGPU&hwname2=Intel(R)+HD+Graphics+530

A firmware trick to give more thermal headroom to the CPU at the cost of GPU performance? ^^

 

[TheELF]

JoeRambo there is very little info abou XFR except for this image:

So based on this image unlike Intels turbo Ryzens Boost can go beyond his limits effectively overclocking itself as long its needed and returning too normal when its not needed

No it can go beyond precision boost just like boost max 3 can go beyond the turbo boost 2 limits.
Also it seems to be the same thing boost max also boosts dependent on cooling (and a host of other factors)
http://www.intel.com/content/www/us...y/turbo-boost/turbo-boost-max-technology.html

Boost Max Technology 3.0
Availability and frequency upside of Intel® Turbo Boost Max Technology 3.0 state depends upon a number of factors including, but not limited to, the following:

• Type of workload
• Number of active cores
• Estimated current consumption
• Estimated power consumption
• Processor temperature
• Drive support

 

[lolfail9001]

C'mon, in the past we had things like 'Hyperthreading done right", "reverse hyperthreading", they had way more epic hype attached, were light in details, and in retrospect made very little sense.

XFR is just fine in ratio of hype/info.

I was not quite following tech back then. And not really, nothing with close to 0 information is "fine in ratio of hype/info" in my book.

Anyways, CPC, it is Monday already.

 

[bjt2]

For tech we have literally 0 factual information about, it is certainly the most hyped thing i have seen yet.

Not new for a new technology. No current CPU has it... I don't see how anyone can have factual informations about that. Maybe MB manufacturer, since they are developing BIOSes for Ryzen...

Hot chips presentation described very precisely the behaviour. If you don't believe it, your problem.

Lool


Sent from HTC 10
(Opinions are own)

What make you laugh? Admitting that the IPC disadvantage of Ryzen is 10% (and I don't believe it), a 3.6GHz base Ryzen more than compensate the IPC deficit. 3.2+10%=3.52GHz...
Mathematic is not an opinion.

Insulting other members is not allowed.
Markfw
Anandtech Moderator

 

[Glo.]

Except that a 3.4 GHz Zen doesn't beat an i7 6900K, people are only thinking it does because of a couple of cherry picked benchmarks displayed at New Horizon.

Canard PC's leaks, most likely the closest representative of real world performance thus far, shows Zen IPC at roughly 5-6% behind Broadwell-E. Surely further optimization will help (as we are talking engineering samples) but I wouldn't realistically expect more than 5-10% over what Canard PC displayed considering they would have only several months to improve upon that.

I've always loved AMD but I swear the fanboy hype surrounding this release is just atrocious, far more than the hype surrounding Polaris or Fiji.

Canard PC benchmarks show IPC at Haswell level. Precisely. Also, you have to bare in mind that the scores are with 3.3 GHz max boost clock for those benches.

Then we have New Horizon demo, which I believe more than any other "leak" we have seen to this day. Why there would be hype in factual data, we have seen, so far?

 

[inf64]

Canard PC benchmarks show IPC at Haswell level. Precisely. Also, you have to bare in mind that the scores are with 3.3 GHz max boost clock for those benches.

Then we have New Horizon demo, which I believe more than any other "leak" we have seen to this day. Why there would be hype in factual data, we have seen, so far?

The timing of CPC leak also points out to a certain stage of maturity of Zen platform they tested (test was done in late Nov if I remember correctly). Things have changed since then and Zen is now ready for release, meaning bugs have been ironed out and bios/motherboards are ready, with final (retail) chips being ready as well.

 

[Zucker2k]

C'mon, in the past we had things like 'Hyperthreading done right", "reverse hyperthreading", they had way more epic hype attached, were light in details, and in retrospect made very little sense.

QFT! Lol. Believe me, the animosity is quite toned down these days. You have to blame AMD for that though; they over-promised and under-delivered for a record stretch of time. This time though, things seem a bit different, even though it'll seem all Intel has to do is adjust prices across board and temporarily demote a couple of chips down from HEDT into the mainstream mix. Consumers win, right? That's what I'm hoping for. I want so badly to see octacore i7s and SR7s at the top of the mainstream segment for $350-$399.

 

[Erenhardt]

Seeing how 4c/8t is considered high-end/overkill gaming cpu, I just wish amd releases 6c/12t at $250...
ok, at $300...
...below $350? pretty please.

 

[lolfail9001]

Not new for a new technology. No current CPU has it... I don't see how anyone can have factual informations about that. Maybe MB manufacturer, since they are developing BIOSes for Ryzen...

We have had what, about 4 or 5 reliable or semi-reliable leaks about actual Zeppelin performance. Not a single one about XFR.

Hot chips presentation described very precisely the behaviour. If you don't believe it, your problem.

You mean that single slide? It's as vague as it gets in description beyond: "Sometimes CPU frequency can get higher than our preset boost frequency".

 

[Atari2600]

We have had what, about 4 or 5 reliable or semi-reliable leaks about actual Zeppelin performance. Not a single one about XFR.

You mean that single slide? It's as vague as it gets in description beyond: "Sometimes CPU frequency can get higher than our preset boost frequency".

Indeed. We've no real indication of what exactly XFR will typically do and how well it will do it.

For instance:

(1) If you are running an application using 3 threads, will it then assign the following affinities:
CPU0 = AllBackgroundThreads
CPU4 = App#1
CPU8 = App#2
CPU12 = App#3

With the rest being closed off, then clock CPU1 to the minimum required and clock up CPUs 4/8/12 to the max the cooling can live with?


(2) If you are running an application using 3 threads, will it then assign the following affinities:
CPU0 = AllBackgroundThreads
CPU2 = App#1
CPU4 = App#1
CPU6 = App#1

Shut down CCX2, then clock CCX1 to the max the cooling can live with?


(3) Allow the threads to float around the logical cores, but upclock whichever logical core the threads are on and downclock the rest?

(4) Do a test to find which logical cores on the CPU are most electrically efficient, then assign threads from most-efficient to least sequentially?

(5) Something else?


Then of course, the question is, how effective will this be? CPC have allegedly managed a 5GHz OC on a single thread on air on an early release (with limiter supposedly being motherboard). Will XFR on a 2 or 3 thread application allow the CPU to clock up 3 threads to the vicinity of 5GHz? Or will it be a more mundane 5% above max turbo (i.e. 4.0 -> 4.2 GHz) or similar?


I suppose a possible pointer to the efficiency of XFR could be CPCs allegation of Intel scrambling to release some higher bin KL. Have Intel got hold of an XFR representative sample (or heard through the grapevine how effective it is)? But that is a tenuous link at best!

 

[guskline]

It will be very interesting to see how far the highest Ryzen 8c/16t can be OC'd on custom water.

 

[blublub]

It will be very interesting to see how far the highest Ryzen 8c/16t can be OC's on custom water.

Easy 6Ghz on all cores - lol

 

[guskline]

Easy 6Ghz on all cores - lol

Funny but not realistic. I have a solid 4.4 Ghz OC on my 5960x (8c16t) so the highest Ryzen SR7 which appears to be 3.6 turboed to 4.0 should match if not beat that. Should be interesting!

 

[KTE]

What make you laugh? Admitting that the IPC disadvantage of Ryzen is 10% (and I don't believe it), a 3.6GHz base Ryzen more than compensate the IPC deficit. 3.2+10%=3.52GHz...
Mathematic is not an opinion. Have you attended third grade school? (I don't know precisely when percentages are thaught in american/english schools)...

Chill with the insults...

Your hype was delirious so I couldn't stop laughing.

200MHz more, equal perf at -45W and 50% price.

Then there's the pages about 4.5GHz 95W...

I don't know haha.

Sent from HTC 10
(Opinions are own)

 

[leoneazzurro]

He was speaking about quad cores and 45W, not 8 cores. And he was also saying 4,5 GHz turbo, not base. And he was the first and one to say Zen was an high frequency product when all others spoke doom&gloom with ES at 2,8GHz. Not that everything he (or you, for what's worth) says has or will meet the reality.
And I don't know what's more hilarious, his optimism or your continous AMD bashing.

 

[bjt2]

We have had what, about 4 or 5 reliable or semi-reliable leaks about actual Zeppelin performance. Not a single one about XFR.

You mean that single slide? It's as vague as it gets in description beyond: "Sometimes CPU frequency can get higher than our preset boost frequency".

Rumors says that the delay of Ryzen this time is MB manufacturer fault, for immature BIOSes. If the CPU seems to work and only XFR is missing, maybe it's this feature that was not implemented... I suppose that XFR must be enabled and configured in BIOSes...