AMD Ryzen SKU and Price Information/Speculation.

[CHADBOGA]

Ryzen single core at around 4ghz boost gets 162 cinebench single thread. i7 7700k@4.4ghz gets 165...If that's not a misread or a mistake, then AMD is actually back for real. The only thing Intel might have is a small advantage with clock speeds and OCability, possibly not amounting to much especially when faced with TWICE the cores and threads. Jesus. If the hype gets any heavier its going to collapse into a black hole.

With how well Zen appears to be doing, it makes one wonder if by the time Zen+ rolls around, if AMD will have an unambiguous lead over Intel in literally every metric, a desktop user would care about.

 

[Saylick]

Ryzen single core at around 4ghz boost gets 162 cinebench single thread. i7 7700k@4.4ghz gets 165...If that's not a misread or a mistake, then AMD is actually back for real. The only thing Intel might have is a small advantage with clock speeds and OCability, possibly not amounting to much especially when faced with TWICE the cores and threads. Jesus. If the hype gets any heavier its going to collapse into a black hole.

I believe the 7700K @ 4.4 GHz does somewhere in the ~200 range, which puts Sky Lake/Kaby Lake at around 7-8% better IPC than Zen. Relative to Haswell, Zen should have 1-2% higher IPC, which is still an astounding feat given AMD's budget and number of engineers.

With how well Zen appears to be doing, it makes one wonder if by the time Zen+ rolls around, if AMD will have an unambiguous lead over Intel in literally every metric, a desktop user would care about.

^ This. Zen+ should yield a 10% IPC gain over Zen, which puts it squarely into Sky Lake/Kaby Lake territory. If AMD can launch Zen+ with a 10% IPC gain AND on a refined 14nm LPP process (+5% max clocks) within 2018, I think we can say with confidence that AMD will have returned true to form. The good news is that Zen and its derivatives should all be compatible with an AM4 mobo, so that should make the upgrade process much easier to justify for those already on-board with AM4.

 

[moonbogg]

With how well Zen appears to be doing, it makes one wonder if by the time Zen+ rolls around, if AMD will have an unambiguous lead over Intel in literally every metric, a desktop user would care about.

I almost can't believe it, but that actually seems sort of likely. I genuinely never thought I'd see the day. AMD was in such terrible shape recently. I was counting the nails as they got hammered into the coffin, just waiting for the funeral to start. Suddenly they bust out like Clark Kent and becomes hero's overnight. CRAZY!

Also, I got the 7700K single thread score of 165 from Cinebench. The program listed it there, but maybe it was off....hang on...lemme google it...7700K shown at 187 stock...bah my bad.

 

[Saylick]

I almost can't believe it, but that actually seems sort of likely. I genuinely never thought I'd see the day. AMD was in such terrible shape recently. I was counting the nails as they got hammered into the coffin, just waiting for the funeral to start. Suddenly they bust out like Clark Kent and becomes hero's overnight. CRAZY!

Also, I got the 7700K single thread score of 165 from Cinebench. The program listed it there, but maybe it was off....hang on...lemme google it...7700K shown at 187 stock...bah my bad.

Haha, I think we'd all wait in line at the nearest Micro Center or take a day off to spam the refresh button for pre-orders if a 4.0 GHz Zen matched a 4.4 GHz 7700K in single thread. That would probably be a huge slap in the face for all consumers who bought Intel chips in the last 5 years as that would imply that 1) Jim Keller is a silicon god, and 2) Intel has been sandbagging like a mofo in the last 5 years. Give AMD one year for another Tock and I think we will have on our hands a chip we can probably feel comfortable keeping for the next 10 years.

 

[CHADBOGA]

2) Intel has been sandbagging like a mofo in the last 5 years.

I don't think Intel has been sandbagging on single core performance, but with all the resources they have, I suspect, as I have suspected for a long while now, that Intel have some real duds in key positions in R&D, which means they haven't been able to achieve the sort of technological performance metrics, one would assume they would be capable of.

 

[Saylick]

I don't think Intel has been sandbagging on single core performance, but with all the resources they have, I suspect, as I have suspected for a long while now, that Intel have some real duds in key positions in R&D, which means they haven't been able to achieve the sort of technological performance metrics, one would assume they would be capable of.

I think that's a fair point, given some of rumors about bad management and unhappiness within the Intel ranks.

If there is indeed a cloud of overhead and overall mismanagement within Intel's R&D department, they can say bye bye to marketshare in the next couple of years. AMD is a much smaller company that is far more agile than the giant Intel. It would be tough for Intel to pivot and respond well to AMD if there is a huge inefficiency in their bureaucracy. I mean, it's pretty self evident that AMD does not lack engineering talent; imagine what AMD could do if they had Intel's R&D budget...

 

[Arachnotronic]

I don't think Intel has been sandbagging on single core performance, but with all the resources they have, I suspect, as I have suspected for a long while now, that Intel have some real duds in key positions in R&D, which means they haven't been able to achieve the sort of technological performance metrics, one would assume they would be capable of.

CHADBOGA, you are spot on that Intel is not sandbagging -- believe it or not, this is the best they can do.

The problem is that their process technology R&D organization really blew it in terms of bringing new mfg technologies to market at good yields, and this completely blew up the product pipeline. While they continue to bloviate about their "3 year lead in logic density," the reality is that their product roadmap depended on 10nm being ready for mass production at good costs at the end of 2015 and then at the end of 2016, both of which did not happen.

Even 14nm to this day is not where it's supposed to be in terms of yields/cost.

The blame, IMHO, lies with the manufacturing organization.

Right now, Murthy is doing what he can to clean it up by working closely with the mfg side of the house and telling them what he needs and defining/executing towards reasonable, low-risk/high reward projects like more cores for mainstream (Coffee Lake) and annual process performance improvements (14nm+, 14nm++). But Murthy, capable as he is, can't fix this overnight.

Going forward, now that Intel has accepted the reality that it will be on 10nm, especially in PCs, for years and years, I think (hope?) we will see much better/interesting products.

 

[itsmydamnation]

I think that's a fair point, given some of rumors about bad management and unhappiness within the Intel ranks.

If there is indeed a cloud of overhead and overall mismanagement within Intel's R&D department, they can say bye bye to marketshare in the next couple of years. AMD is a much smaller company that is far more agile than the giant Intel. It would be tough for Intel to pivot and respond well to AMD if there is a huge inefficiency in their bureaucracy. I mean, it's pretty self evident that AMD does not lack engineering talent; imagine what AMD could do if they had Intel's R&D budget...

I Think its more intel has been focusing on trying to break into markets AVX512/FPGA/CELLular etc, pushing new tech that might not be there yet ( 3d xpoint) and haven't focused on their core market all in the name of growth. Now AMD is releasing something competitive all those other endeavors seem like a misstep in terms of focus.

 

[Saylick]

I Think its more intel has been focusing on trying to break into markets AVX512/FPGA/CELLular etc, pushing new tech that might not be there yet ( 3d xpoint) and haven't focused on their core market all in the name of growth. Now AMD is releasing something competitive all those other endeavors seem like a misstep in terms of focus.

Yeah, this is probably part of it, too. I remember reading somewhere how Intel went about buying up various companies for markets outside of processors.

I will say this though. Intel has been flourishing through fat profit margins on their core markets (consumer, server processors) and thus had plenty of cash to throw around in an attempt to diversify. At the end of the day, Intel's might comes from their core markets and it seems AMD is pose to strike Intel where it will hurt the most, i.e. those fat profit margins won't last indefinitely. Intel will need to refocus, like you said, to combat this threat. If Intel were a giant, AMD is out to bust them in the knee.

I suppose the same can be said for Nvidia as well. Nvidia has been trying to branch out into other revenue streams for the last 5 years. First it was mobile, which flopped, and now its into automotive and deep learning. Regardless, their bread and butter is still consumer GPUs. Depending on how well Vega and Navi turn out, AMD is setting itself up to knee-cap them as well within the next few years.

The whole CPU + GPU industry is going to be seeing a lot more competitive from here on out, and AMD is best prepared to leverage both their CPU and GPU knowledge to produce products that neither Intel or Nvidia could ever deem reasonably possible. For example, imagine an AMD APU with AVX2 that can handle small SIMD workloads on the CPU but switch over to the on-die GPU for larger SIMD workloads seamlessly. My understanding is that there is a cross-over point where it makes sense to use a GPU for large data set; AMD just needs to build an APU that can handle small and medium sized data sets, then have an off-die dedicated GPU for workloads larger than that. I look forward to reading more about their Infinity Fabric to see if this is a potential possibility in the near future.

 

[CHADBOGA]

CHADBOGA, you are spot on that Intel is not sandbagging -- believe it or not, this is the best they can do.

The problem is that their process technology R&D organization really blew it in terms of bringing new mfg technologies to market at good yields, and this completely blew up the product pipeline. While they continue to bloviate about their "3 year lead in logic density," the reality is that their product roadmap depended on 10nm being ready for mass production at good costs at the end of 2015 and then at the end of 2016, both of which did not happen.

Even 14nm to this day is not where it's supposed to be in terms of yields/cost.

The blame, IMHO, lies with the manufacturing organization.

Right now, Murthy is doing what he can to clean it up by working closely with the mfg side of the house and telling them what he needs and defining/executing towards reasonable, low-risk/high reward projects like more cores for mainstream (Coffee Lake) and annual process performance improvements (14nm+, 14nm++). But Murthy, capable as he is, can't fix this overnight.

Going forward, now that Intel has accepted the reality that it will be on 10nm, especially in PCs, for years and years, I think (hope?) we will see much better/interesting products.

Do you think that the missteps in manufacturing is why Intel will only be bringing a 6 core CPU to the desktop market that isn't part of their HEDT range?

I realise that in percentage terms, the number of people who buy 8 core processors will be quite small, but I am thinking about when Zen+ hits, if one can believe Intel's roadmaps, that they won't have an 8 core processor to go head to head with it and Zen+ may have equal or higher single core performance than Intel at that point.

 

[imported_jjj]

Do you think that the missteps in manufacturing is why Intel will only be bringing a 6 core CPU to the desktop market that isn't part of their HEDT range?

I realise that in percentage terms, the number of people who buy 8 core processors will be quite small, but I am thinking about when Zen+ hits, if one can believe Intel's roadmaps, that they won't have an 8 core processor to go head to head with it and Zen+ may have equal or higher single core performance than Intel at that point.

Intel is only about maximizing profits. The 6 cores would be at a premium if AMD doesn't force their hand. The main tactic for Intel has been and still is to upsell customers, push ASPs and margins up.They were able to do that because AMD was out of the picture. Intel could offer 8 cores easily at reasonable prices, they just don't want to. The plan wasn't to offer 6 cores at 350$, that is against their interests.

If they react next year remains to be seen, they might be too arrogant to and AMD improves it's position with Zen+ ,then in 2019 likely AMD would be going 12 cores on 7nm.

Intel is in a tough position and very different from AMD's. AMD needs a boost in revenue, in marketing. They got some 3% revenue share in PC so they are free to give us what we want and make a lot of $.
On the other hand, Intel has 97% revenue share in a declining market and with AMD back in the game it gets complicated. They are gonna get hit hard as the market is shrinking, AMD will gain some share and a competitive AMD will lead to a decline in ASPs. That's quite problematic for Intel as they have been trying hard to push ASPs up to offset the decline in units. Their PC revenue could decline by 30% by next year (vs last year) and the funny thing is that if they engage in price wars, it gets worse as ASPs and margins would be even lower.

 

[CentroX]

Kaby lake is at 190 cb single thread @ 4.4 ghz. Ryzen if it can reach 4.4 ghz will have 178. Still pretty good i guesd.

 

[unseenmorbidity]

Kaby lake is at 190 cb single thread @ 4.4 ghz. Ryzen if it can reach 4.4 ghz will have 178. Still pretty good i guesd.

More than enough for it to not matter.

 

[Arachnotronic]

Do you think that the missteps in manufacturing is why Intel will only be bringing a 6 core CPU to the desktop market that isn't part of their HEDT range?

No. Remember the basic design point for the 'S' CPUs is 45W notebooks, and with Coffee Lake they're going to optimize the circuit design and process enough so that they can fit a 6 core into that power envelope while maintaining/increasing single-core performance. 8 cores won't quite work in such a low TDP -- even 6 is going to be tough, but the gaming laptop and mobile workstation OEMs likely wanted something better to sell to customers than "hey, it's another quad with higher clocks."

I would actually say that it's pretty impressive that they're wringing enough out of the circuit/process design to actually allow for a six core SoC in that kind of power envelope and performance level -- though obviously, it would have been much better if they were prepping the actual Cannon Lake-H/S parts on 10nm (Gen 10 GPU + media, updated CPU cores, etc.).

I realise that in percentage terms, the number of people who buy 8 core processors will be quite small, but I am thinking about when Zen+ hits, if one can believe Intel's roadmaps, that they won't have an 8 core processor to go head to head with it and Zen+ may have equal or higher single core performance than Intel at that point.

The distinction between "mainstream" and "HEDT" is pretty darn arbitrary. Intel should be able to compete on core count/MT perf by appropriately adjusting the prices of its HEDT lineup and perhaps working with board vendors to perhaps widen the price of their board offerings a bit (basically getting them to build cheaper, less featured boards).

Honestly, mainstream desktop should really only be used by the typical PC OEMs; if Intel didn't totally screw it up, HEDT would be the "go to" platform for DIY enthusiasts using Intel CPUs. I think they fix this with X299/Skylake-X.

Having the Kaby Lake-X parts as "entry level" for the platform and also offering the higher core count (and higher IPC) Skylake-X parts for people who want more performance should make the X299 platform much more interesting than the X79/X99 crap.

 

[imported_jjj]

The HEDT platform is a server die, they just make an extra buck by selling it in consumer.
Skylake X isn't solving anything, still 10 core die and quad memory channel, large L3 plus server specific bits and a bunch of PCIe lanes. So still 2x the die size vs Skylake 4C+GPU. They use some 49mm2 for the quad core complex and maybe 5mm2 less for the GPU. They can't afford to sell 20 million HEDT at 100-400$ retail, the fix comes when they properly address the market with a die designed for consumer.

 

[Arachnotronic]

The HEDT platform is a server die, they just make an extra buck by selling it in consumer.

What do you think Summit Ridge is?

Skylake X isn't solving anything, still 10 core die and quad memory channel, large L3 plus server specific bits and a bunch of PCIe lanes. So still 2x the die size vs Skylake 4C+GPU. They use some 49mm2 for the quad core complex and maybe 5mm2 less for the GPU. They can't afford to sell 20 million HEDT at 100-400$ retail, the fix comes when they properly address the market with a die designed for consumer.

Skylake-X will take care of 6 core+, and Kaby Lake-X will handle 4 core/4 thread and 4 core/8 thread. What other options on a platform do gamers/enthuisiasts want?

 

[Ajay]

Intel is only about maximizing profits. The 6 cores would be at a premium if AMD doesn't force their hand. The main tactic for Intel has been and still is to upsell customers, push ASPs and margins up.They were able to do that because AMD was out of the picture. Intel could offer 8 cores easily at reasonable prices, they just don't want to. The plan wasn't to offer 6 cores at 350$, that is against their interests.

If they react next year remains to be seen, they might be too arrogant to and AMD improves it's position with Zen+ ,then in 2019 likely AMD would be going 12 cores on 7nm.

Intel is in a tough position and very different from AMD's. AMD needs a boost in revenue, in marketing. They got some 3% revenue share in PC so they are free to give us what we want and make a lot of $.
On the other hand, Intel has 97% revenue share in a declining market and with AMD back in the game it gets complicated. They are gonna get hit hard as the market is shrinking, AMD will gain some share and a competitive AMD will lead to a decline in ASPs. That's quite problematic for Intel as they have been trying hard to push ASPs up to offset the decline in units. Their PC revenue could decline by 30% by next year (vs last year) and the funny thing is that if they engage in price wars, it gets worse as ASPs and margins would be even lower.

Only 12 cores @ 7nm? That would be unfortunate for AMD vis-a-vis the server market.

Right now AMD is contained by capacity and TAM with Summit Ridge - this should ease the pressure on Intel.
Raven Ridge with address the TAM issue, assuming AMD has sufficient capacity (I don't know when the new capacity will be available at Fab 8).
The GFL foundry planned for Chongqing, China is supposed to be begin production this year - so I'm guessing that really means we'll see CPUs fabbed there in 2018.

 

[moonbogg]

Skylake-X will take care of 6 core+, and Kaby Lake-X will handle 4 core/4 thread and 4 core/8 thread. What other options on a platform do gamers/enthuisiasts want?

Honestly? We want the choice to break free from quad core peasantry without spending $500-$1,700 on a CPU from Intel. For Intel quad core prices, you get a 6 or even 8 core CPU from AMD. Skylake X, in my honest opinion, won't be a solution for hardly anyone because the prices will be straight through the roof. And Kaby-X is hilarious to me because its likely going to be an even more expensive quad core and only available on a very expensive, overly bloated platform. I really think Ryzen has Intel by the raisins for the next couple years.

 

[bjt2]

Ryzen single core at around 4ghz boost gets 162 cinebench single thread. i7 7700k@4.4ghz gets 165...If that's not a misread or a mistake, then AMD is actually back for real. The only thing Intel might have is a small advantage with clock speeds and OCability, possibly not amounting to much especially when faced with TWICE the cores and threads. Jesus. If the hype gets any heavier its going to collapse into a black hole.

Could be XFR into play here. More IPC in CB15 and not in blender? I suspect that XFR took Ryzen in the 4.3GHz range here...
EDIT: it seems a strange score... Anyway i would like to know if XFR was active...

 

[bjt2]

I believe the 7700K @ 4.4 GHz does somewhere in the ~200 range, which puts Sky Lake/Kaby Lake at around 7-8% better IPC than Zen. Relative to Haswell, Zen should have 1-2% higher IPC, which is still an astounding feat given AMD's budget and number of engineers.



^ This. Zen+ should yield a 10% IPC gain over Zen, which puts it squarely into Sky Lake/Kaby Lake territory. If AMD can launch Zen+ with a 10% IPC gain AND on a refined 14nm LPP process (+5% max clocks) within 2018, I think we can say with confidence that AMD will have returned true to form. The good news is that Zen and its derivatives should all be compatible with an AM4 mobo, so that should make the upgrade process much easier to justify for those already on-board with AM4.

Mmhhh... This score seems strange also to me... If Broadwell-E ST (6900K with 4.0GHz turbo core 3.0) scores 162, SKL at 4.4GHz, with +10% clock and +4% IPC should score more... But not 200. If I remember well, 200 were OCed skylake...

 

[Edgemeal]

I wasn't expecting Ryzen 8C/16T to priced so low out of the gate, AMD is really twisting my arm to upgrade!

Just for fun, Flashback to early P3-K7 CPU days.
These were the fastest CPUs (and lowest prices) on pricetag.com.
* First listings, give or take a few days.

First listing: Intel P3-733 - Nov 14 1999
P3 - 733.......Upgrade Source.............. $ 809
P3 - 700.......Upgrade Source.............. $ 749
P3 - 650.......Upgrade Source.............. $ 624
K7 - 700.......Tufshop.com................. $ 670
K7 - 650.......Shophereusa.com............. $ 505
K7 - 600.......Accubyte.................... $ 402

First listing: AMD K7-750 - Dec 09 1999
K7 - 750.......Componentsdirect.com........ $ 865
K7 - 700.......Volume Software............. $ 641
K7 - 650.......Volume Software............. $ 483
P3 - 733.......Upgrade Source.............. $ 789
P3 - 700.......PC Progress................. $ 772
P3 - 650.......Best Buy Micro.............. $ 582

First listing: AMD K7-800 - Jan 11 2000
K7 - 800.......Lynn Computer Products...... $ 839
K7 - 750.......Lynn Computer Products...... $ 639
K7 - 700.......United Distribution......... $ 499
K7 - 650.......United Distribution......... $ 299
P3 - 733.......ECONOPC, Inc................ $ 763
P3 - 700.......Paragon Technology Inc...... $ 719
P3 - 667.......All-Star/SMK Components..... $ 576
P3 - 650.......Accubyte.................... $ 557

First listing: AMD K7-1000 - Mar 09 2000
K7 - 1000......Componentsdirect.com........ $1399
K7 - 850.......Bay Area Micro.............. $ 735
K7 - 800.......Accubyte.................... $ 511
K7 - 750.......Accubyte.................... $ 348
K7 - 700.......All-Star/SMK Components..... $ 252
K7 - 650.......All-Star/SMK Components..... $ 206
K7 - 600.......All-Star/SMK Components..... $ 177
'
P3 - 800.......Accubyte.................... $ 761
P3 - 750.......Accubyte.................... $ 549
P3 - 733.......Upgrade Source.............. $ 470
P3 - 700.......Upgrade Source.............. $ 442
P3 - 650.......Atacom, Inc................. $ 332
P3 - 600.......Atacom, Inc................. $ 254

 

[CatMerc]

When was the last time AMD could release a 500$ CPU as a sane move?

 

[lolfail9001]

When was the last time AMD could release a 500$ as a sane move?

Has to be before Core 2. In fact, they had $1k CPUs back then, didn't they?

 

[majord]

When was the last time AMD could release a 500$ as a sane move?

FX-62 (For that matter, everything 4600+ at AM2's launch)


Athlon 64 FX-62 2.8 GHz 2 x 1024 KB 125 W $1,031
Athlon 64 X2 5000+ 2.6 GHz 2 x 512 KB 89 W $696
Athlon 64 X2 4800+ 2.4 GHz 2 x 1024 KB 89 W $645
Athlon 64 X2 4600+ 2.4 GHz 2 x 512 KB 89 W $558
Athlon 64 X2 4400+ 2.2 GHz 2 x 1024 KB 89 W $470
Athlon 64 X2 4200+ 2.2 GHz 2 x 512 KB 89 W $365
Athlon 64 X2 3800+ 2.0 GHz 2 x 512 KB 89 W $303

 

[Topweasel]

FX-62 (For that matter, everything 4600+ at AM2's launch)


Athlon 64 FX-62 2.8 GHz 2 x 1024 KB 125 W $1,031
Athlon 64 X2 5000+ 2.6 GHz 2 x 512 KB 89 W $696
Athlon 64 X2 4800+ 2.4 GHz 2 x 1024 KB 89 W $645
Athlon 64 X2 4600+ 2.4 GHz 2 x 512 KB 89 W $558
Athlon 64 X2 4400+ 2.2 GHz 2 x 1024 KB 89 W $470
Athlon 64 X2 4200+ 2.2 GHz 2 x 512 KB 89 W $365
Athlon 64 X2 3800+ 2.0 GHz 2 x 512 KB 89 W $303

I got a X2-4400+ around launch was the most expensive CPU purchase till my 3930k. Still probably my favorite CPU purchase ever. We needed that much more back then then we need a 8c16t CPU now. Those were the days.