AMD's Future Chips & SoC's: News, Info & Rumours.

[juanrga]

I expect Zen2 to bring 5--10% increase over Zen IPC.

 

[jdwii]

https://www.forbes.com/sites/antonyleather/2017/09/05/amd-ryzen-threadripper-the-fascinating-story-behind-the-processor-that-beat-intel/#1b899682230c

Explains how we got thread ripper was a side project never was officially planned

 

[jdwii]

I can think of ways Amd can improve IPC by quite a bit more then 10% but they have to make the individual core bigger not to sure they want to do that as they brag about increasing IPC without increasing power consumption.

Jumping to 7nm would allow them to make the core wider(4ALU+4AGU) while also allowing the FP unit to handle full 512Bit AVX2

Improvements to other areas could be done as well such as improving the speed of the infinity fabric, lowering cache latency which for L3 cache would be nice and finally the memory controller which is Amd's ryzen weak point in some cases barely handling 3200mhz on all models.

The amount of improvements Amd could do is endless but the amount they will actually do isn't.

Previously someone from Amd can't remember who bragged about how 1% of IPC costs 1% more in power consumption at least so if we go by that alone according to Global Foundry 7nm should decrease power consumption by 60% so obviously that is probably BS and that number is probably closer to half that.

30% more IPC from Zen to Zen 3 could be seen in my opinion at least. BUT i'm not saying Amd will even do this they might pull a good old FX and just barely make the core bigger while just lowering the cost of manufacturing by simply shrinking Zen to 7nm and lowering the price so they can compete in the lower end market which was my biggest fear about Ryzen and Amd.


Basically the way i look at it is Amd will either fight Intel head on and try to improve IPC as much as they can or they will do small adjustments of 5% IPC keeping the core around the same size and admit defeat and lower prices to I3-I5 levels on mainstream parts.

 

[juanrga]

https://www.forbes.com/sites/antonyleather/2017/09/05/amd-ryzen-threadripper-the-fascinating-story-behind-the-processor-that-beat-intel/#1b899682230c

Explains how we got thread ripper was a side project never was officially planned

I already said here in the forums that TR didn't appear in the roadmaps a pair of years ago. The only planned sockets were: AM4, SP4, and SP3.

Then the SP4 socket failed to get traction from the industry, which raised doubts about what do with those 16-core CPUs and, in the other hand, Skylake-X was much more powerful than AMD expected, and in a last minute change AMD merged the SP4 and SP3 platforms in the new SP3r2 socket, releasing the 8, 12 and 16 core ThreadRipper.

TR4 is the marketing term for the SP3r2 socket.

 

[juanrga]

The amount of improvements Amd could do is endless but the amount they will actually do isn't.

Law of dismissing returns. Improving from X to 1.4X is much more easy than improving from 1.4X to 1.5X.

Previously someone from Amd can't remember who bragged about how 1% of IPC costs 1% more in power consumption at least so if we go by that alone according to Global Foundry 7nm should decrease power consumption by 60% so obviously that is probably BS and that number is probably closer to half that.

For x86, the relation between core IPC and its power consumption is nonlinear. 10% more IPC has a cost of about 20% more power consumption. The core with the higher IPC is always the less efficient (everything else the same of course), because the power consumption increases at a faster rate than performance.

 

[-Fran-]

I can think of ways Amd can improve IPC by quite a bit more then 10% but they have to make the individual core bigger not to sure they want to do that as they brag about increasing IPC without increasing power consumption.

Jumping to 7nm would allow them to make the core wider(4ALU+4AGU) while also allowing the FP unit to handle full 512Bit AVX2

Improvements to other areas could be done as well such as improving the speed of the infinity fabric, lowering cache latency which for L3 cache would be nice and finally the memory controller which is Amd's ryzen weak point in some cases barely handling 3200mhz on all models.

The amount of improvements Amd could do is endless but the amount they will actually do isn't.

Previously someone from Amd can't remember who bragged about how 1% of IPC costs 1% more in power consumption at least so if we go by that alone according to Global Foundry 7nm should decrease power consumption by 60% so obviously that is probably BS and that number is probably closer to half that.

30% more IPC from Zen to Zen 3 could be seen in my opinion at least. BUT i'm not saying Amd will even do this they might pull a good old FX and just barely make the core bigger while just lowering the cost of manufacturing by simply shrinking Zen to 7nm and lowering the price so they can compete in the lower end market which was my biggest fear about Ryzen and Amd.


Basically the way i look at it is Amd will either fight Intel head on and try to improve IPC as much as they can or they will do small adjustments of 5% IPC keeping the core around the same size and admit defeat and lower prices to I3-I5 levels on mainstream parts.

By doing a simple "percentage" game, AMD can't improve ~7% of performance (IPC+speed for the terminology-generous fellas) per gen alone to stay on par with Intel, even if they "only" improve in ~5% chunks given the current ~7% disparity. It would take them way too long to catch up in the metric that matters in the server sector: efficiency.

I hope they do better, but I'm not really sure there is such a clear cut method of improving Zen's design that much. They could bring ancillary improvements that will indeed increase overall performance of the platform that will mask lesser gain in pure CPU metrics. The most interesting and obvious one for me is IF tunning. This is just their first iteration and hands-on play with IF in the MCM and integrated die for the Zen siblings. They can also tune cache a bit more and IMC for performance and that mesh control for the speed tuning mechanism they had for... Richland was it that had it first?

Adding AVX512 smells like a bad avenue to follow if they're not making something for HPC. Improving AVX and AVX2 throughput would yield better performance improvements all around I'd say. Oh, and NUMA capabilities. They need to jump on the software subsiding bandwagon for that one and bring more software from the server market to the consumer market.

So, in short, I don't think they need to revolutionize the Zen core design to get some decent gains here and there. Without getting too inside the architecture itself, I don't think we could be digging more and that is knowledge I still haven't captured completely. I need to visit Agner's site more often, lol.

Cheers!

EDIT: Bedtime reading!
http://www.agner.org/optimize/microarchitecture.pdf

 

[juanrga]

Adding AVX512 smells like a bad avenue to follow if they're not making something for HPC.

AVX512 is used outside HPC. Google servers use AVX512.

 

[-Fran-]

Adding AVX512 smells like a bad avenue to follow if they're not making something for HPC.

AVX512 is used outside HPC. Google servers use AVX512.

I'm pretty sure Google does very serial work in their servers that cannot be made parallel at all. Yup.

Cheers!

 

[juanrga]

Adding AVX512 smells like a bad avenue to follow if they're not making something for HPC.

AVX512 is used outside HPC. Google servers use AVX512.

I'm pretty sure Google does very serial work in their servers that cannot be made parallel at all. Yup.

The parallel part is run on the AVX512 units, the serial part is run on the rest of units.

 

[aldaia]

Someone posted that in annother forum regarding 8180 ($10000) and Epyc 7601 ($4200) with spec 2006

Heise.de publised their SPEC2006 results for Platinum 8180(28C/56T), Epyc 7601(32C/64T), Gold 6148(20C/40T) and E5-2699 v4(22C/44T), all of them in 2P configuration.
https://www.heise.de/ct/ausgabe/201...6148-E5-2699v4-und-AMD-Epyc-7601-3787494.html
All measurements with 64-bit code, without auto-parallelization, without SmartHeap with Intel System Studio 2017Up4 (opt: AVX512 / AVX2)Or x86-open64 4.5.2.1 (AMD, opt for bulldozer / piledriver).

It's paywalled(1.5 Euro). I've had a look at it and the most important findings are:

2X Epyc 7601 and 2X Xeon Platinum 8180 are extremely close. Xeon was 12% ahead in int_rate_base and 4.5% ahead in fp_rate_base.

Without NUMA optimiztion, Epyc is ahead in STREAM-Triad for very low thread counts(<8) sometimes by almost 2X over Skylake-SP Xeons. The 8180 is bandwidth-starved compared to the 6148 by 35% for 16-32 threads. When fully saturated with the maximum permissible number of threads, bandwidth on Skylake-SP dips while Epyc remains constant.

The huge performance disparity in libquantum is obliterated when compiled for 64-bit. Compile times were reduced to two seconds from 10 minutes on 32-bit mode. Yes, minutes. Epyc and Xeon 8180 give the same performance even with AVX512 enabled on the latter in 64-bit mode.

tamz_msc, Aug 12, 2017

And with the newer spec 2017:

SPEC2017 results by heise.de are out:

Something is wrong with the integer performance on EPYC on the 'speed' benchmark(which measures time to completion for a single job) - 2x Xeon Platinum 8180 is over 60% faster than 2x Epyc 7601 using intel17 -O3. Same issue with gcc, so might be a compiler issue. AMD says their own AOCC does fine, but not tested due to time constraints.

This discrepancy is not there when running the 'rate' benchmark(measuring throughput running multiple copies of each job) - Xeon Platinum is only ~5% ahead of Epyc with intel17 -O3.

There is no issue with fp performance however. In the 'speed' benchmark Xeon Platinum is just ~4% ahead of Epyc. However, in the throughput-oriented 'rate' benchmark Epyc pulls ahead of the Xeon Platinum by ~16%, again with intel17 -O3. Due to the huge memory bandwidth it offers, in some of the benchmarks, like fluid-dynamics using the lattice-Boltzmann method, Epyc is over 55% faster than Xeon! Epyc does very well in general with throughput-oriented numerical PDE solvers from what I've seen, which is expected because these kinds of HPC workloads absolutely love memory bandwidth and generally run without issues when made NUMA-aware.

AVX512 does nothing over AVX2 with the Intel compiler; AVX2 is still slightly faster for the 8180, however with the 20C Gold 6148, AVX512 is marginally faster than AVX2 in the 'speed' benchmarks, which suggests that there is some kind of bottleneck with the higher core count chip; which is ironic given Intel's claims with the mesh topology and cache restructuring, but not entirely unexpected. On gcc 7.1 however, skylake-avx512 opt. flag is behind broadwell opt. flag with the 8180, sometimes by a long shot with the rate benchmark.

Bottom line(my interpretation): it's down to application preferences when it comes to which processor you're going to buy. There are still issues with int performance on Epyc, so Skylake-SP is the winner integer workloads as of now. However, Epyc is obviously much faster in fp performance, and since AVX512 does nothing at the moment for Intel, it is obviously the better choice.

tamz_msc, Saturday at 9:16 AM

 

[goldstone77]

Intel Core i7-8700K, Core i7-8600K, Core i3-8350K and More 8th Gen Coffee Lake Desktop CPUs Listed For Pre-Order – Launch Imminent in Early October
Author Photo
By Hassan Mujtaba
1 min ago

Intel’s upcoming 8th generation desktop platforms have taken the internet by storm and there’s nothing that can stop them. In reaction to the leaks, online retailers have started listing the CPUs (Coffee Lake) up for pre-order along with engineering samples that are available on several China based retail sites.

Intel Coffee Lake Pre-Orders Pop Up at Online Retail Stores – Core i7-8700K, Core i5-8600K and Core i3-8350K Included In The Listings
There are a plethora of online retail sites that have listed the Intel 8th gen desktop processors codenamed Coffee Lake. Of course, I won’t be mentioning prices as they vary from retailer to retailer and are not indicative of the final MSRPs which would be much lower. Regardless, you might want to check out the following package names to find if your nearest retailers has the Coffee Lake desktop chips listed for pre-order.

Note – Final clocks are still not determined, the clock frequencies mentioned below are directly from Intel slides however pre-orders reveal different clock speeds so take them with a pinch of salt.

Intel Core i7-8700K – The Coffee Lake Flagship With 6 Cores and 4.3 GHz Boost Across All Cores
The Intel Core i7-8700K will be the flagship processor of the lineup. It will feature 6 cores and 12 threads. The chip will be compatible with the LGA 1151 socket and rumors are that board makers will extend support of Coffee Lake onto 200-series and even 100-series platforms. The chip will be Intel’s first hexa core product and will be based on the 14nm process node.
In terms of clock speeds, we are looking at a 3.7 GHz base frequency which boosts up to 4.3 GHz (6 core), 4.4 GHz (4 core), 4.5 GHz (2 core) and 4.7 GHz (1 core). These clocks are really impressive so we can expect much faster gaming performance on Coffee Lake parts than the current generation of CPUs. The chip is fully unlocked allowing for overclocking and comes with a TDP of 95W. There’s 12 MB of L3 cache on board along with a GT2 tier iGPU. The chip supports DDR4-2666 MHz memory (native) and up to 4400 MHz+ (OC). We can expect the Core i7-8700K to cost around $349 US when it launches in Q3 2017.

Intel Core i5-8600K – The Gamer Aimed Hexa Core With Overclocking Capability and 4.4 GHz Boost
The Intel Core i5-8600K is the second unlocked chip in the lineup and features a hexa core design. We have learned that Intel Core i5 series will stick with 6 cores but they will not feature a multi-threaded design. The chip probably features 9 MB of L3 cache while being supported on the LGA 1151 socket.

In terms of clock speeds, we are looking at 3.5 GHz base, 4.2 GHz (6 core), 4.2 GHz (4 core), 4.2 GHz (2 core) and 4.4 GHz (1 core) boost clocks. The chip will feature a TDP of 95W and we can expect a price range of around $249 US which will make it a pretty good option for gamers that want some extra cores in their hands.

Intel Core i3-8350K – A Quad Core i3 Processor With 4 GHz Clock Speed
The Intel Core i3-8350K might be the only Core i3 chip in the 8th Gen family that enables overclocking support. Intel’s 8th Gen Core i3 family is also the first to feature 4 cores which will be a great deal for budget builders. The Intel Core i3 8350K comes with 4 cores and 4 threads that are clocked at 4.0 GHz. The chip features 8 MB of cache and will feature a TDP of 95W.

There’s also the Intel Core i3-8100 which is a 4 core, non-multi threaded SKU. This chip has a lower clock speed of 65W and features a clock speed of 3.6 GHz. There’s total of 6 MB of L3 cache on the processor itself which confirms the exclusion of multi-threading. The Intel Core i3 chips will not feature boost clocks but they will have a more stable base clock speed and we can expect prices of the chips to be in the sub-$200 US range.

Intel Pentium G4660, Pentium G4700 and Pentium G4720 With 2 Cores, 4 Threads and Upto 3.9 GHz Clock Speeds Also Detailed – Under $100 US
The Intel Coffee Lake family will also feature at least three Pentium series SKUs that include the Pentium G4660, Pentium G4700 and the Pentium G4720. These chips will feature 2 cores and 4 threads with clock speeds rated up to 3.90 GHz. There will also be 4 MB of L3 cache on board the chips and prices will be adjusted around the $100 US bracket. Rest of the processors that will be featured in the family are listed below.

I put this in the wrong forum, could a mod delete this I'll repost it.

 

[goldstone77]

AMD is gaining ground on Intel faster than analysts had thought
Published: Sept 6, 2017 10:49 a.m. ET
By
RYAN
SHROUT
Ryan Shrout is the founder and lead analyst at Shrout Research, and the owner of PC Perspective. Follow him on Twitter @ryanshrout.

Aug. 10 saw the release of the Ryzen Threadripper processor, a high-priced and high-margin part that competes against the flagship consumer-product line from Intel. These AMD processors account for a large portion of this retailer’s revenue, considering unit sales volume and the timing of the release. Intel also launched new high-performance desktop processors last quarter, but the pent-up excitement and interest in Threadripper appears to dwarf the flagship Intel options.

Diving into the breakdown of the sales data shows that AMD has a diverse processor lineup that has many popular options. The highest-selling unit is $220. This is a much higher ASP (average selling price) than AMD could claim before Ryzen’s release (because of performance deficiencies) and should lead to better margins as the sales extend into the second half of the year.

This data, of course, are sourced from only a single hardware vendor in Germany, but sales reports from North American stores like Newegg and Amazon AMZN, +0.04% are showing a similar, if not as dramatic, trend. All indications tell us that Threadripper is selling better than the Core i9 family in North America, giving AMD the edge in the high-margin segment.

 

[jdwii]

http://www.tomshardware.com/news/amd-bristol-ridge-35w-apu-available-now,35398.html

Meh now i guess the OEM market will start to buy AM4 now with these crappy APUs out.

Now lets get raven ridge out.

 

[jdwii]

AMD is gaining ground on Intel faster than analysts had thought
Published: Sept 6, 2017 10:49 a.m. ET
By
RYAN
SHROUT
Ryan Shrout is the founder and lead analyst at Shrout Research, and the owner of PC Perspective. Follow him on Twitter @ryanshrout.

Aug. 10 saw the release of the Ryzen Threadripper processor, a high-priced and high-margin part that competes against the flagship consumer-product line from Intel. These AMD processors account for a large portion of this retailer’s revenue, considering unit sales volume and the timing of the release. Intel also launched new high-performance desktop processors last quarter, but the pent-up excitement and interest in Threadripper appears to dwarf the flagship Intel options.

Diving into the breakdown of the sales data shows that AMD has a diverse processor lineup that has many popular options. The highest-selling unit is $220. This is a much higher ASP (average selling price) than AMD could claim before Ryzen’s release (because of performance deficiencies) and should lead to better margins as the sales extend into the second half of the year.

This data, of course, are sourced from only a single hardware vendor in Germany, but sales reports from North American stores like Newegg and Amazon AMZN, +0.04% are showing a similar, if not as dramatic, trend. All indications tell us that Threadripper is selling better than the Core i9 family in North America, giving AMD the edge in the high-margin segment.

I was reading something on this and it makes sense that people are not buying Intel parts right now with coffee lake coming soon who the heck would buy a I5 4 core 4 thread CPU when in 2 months they will get 6 cores.

Expected price drops on Ryzen to stay competitive
1800X $329.99
1700X $279.99
1700 $249.99
1600X $209.99
1600 $179.99
1500X $129.99
1400 $109.99
1300X $99.99
1200 $79.99

For Amd Ryzen to stay relevant prices need to drop to this level to compete with Coffee-lake i actually thought about these prices for awhile this should be good enough to fight off coffee-lake and i doubt Intel will lower prices of coffee-lake after release. Zen on 14nm+ will probably also be around this price if its just the same thing with 200-300mhz improvement.

Edit the good news is threadripper should continue to offer a better value in most cases vs X299 and not sure if you guys know but X399 will support NVMe RAID support soon which is freaking cool think about 2 960's in RAID that is some massive bandwidth which can be used nicely on these workstations

ECC memory support+64 PCI-E this platform really does have enough to live for a few years for future CPU revisions. Lets see if Intel keeps X299 around just as long?

Edit just for fun my 3.9ghz OC has almost the same single score performance as my older 4790K at stock which features a 4.4 turbo
https://s26.postimg.org/i6qa3tkx3/R15.png

Been increasing performance by tweaking memory as much as i can like lowering TRC which shows a 2-3% boost as the bios likes to set this to a crazy high 75T.

https://www.overclock3d.net/reviews/cpu_mainboard/amd_ryzen_memory_optimisation_-_the_effect_of_trc_timings/1

 

[goldstone77]

http://www.tomshardware.com/news/amd-bristol-ridge-35w-apu-available-now,35398.html

Meh now i guess the OEM market will start to buy AM4 now with these crappy APUs out.

Now lets get raven ridge out.

I know hardware unboxed just unboxed one of these for testing not too long ago. Let's see if I can't give him a nudge to get it going! hahaha

 

[juanrga]

AMD is gaining ground on Intel faster than analysts had thought
Published: Sept 6, 2017 10:49 a.m. ET
By
RYAN
SHROUT
Ryan Shrout is the founder and lead analyst at Shrout Research, and the owner of PC Perspective. Follow him on Twitter @ryanshrout.

Aug. 10 saw the release of the Ryzen Threadripper processor, a high-priced and high-margin part that competes against the flagship consumer-product line from Intel. These AMD processors account for a large portion of this retailer’s revenue, considering unit sales volume and the timing of the release. Intel also launched new high-performance desktop processors last quarter, but the pent-up excitement and interest in Threadripper appears to dwarf the flagship Intel options.

Diving into the breakdown of the sales data shows that AMD has a diverse processor lineup that has many popular options. The highest-selling unit is $220. This is a much higher ASP (average selling price) than AMD could claim before Ryzen’s release (because of performance deficiencies) and should lead to better margins as the sales extend into the second half of the year.

This data, of course, are sourced from only a single hardware vendor in Germany, but sales reports from North American stores like Newegg and Amazon AMZN, +0.04% are showing a similar, if not as dramatic, trend. All indications tell us that Threadripper is selling better than the Core i9 family in North America, giving AMD the edge in the high-margin segment.

Newegg USA sales posted in the X299 thread say otherwise .

 

[Shaq-34]

I know hardware unboxed just unboxed one of these for testing not too long ago. Let's see if I can't give him a nudge to get it going! hahaha

A12-9800 vs A10-7860K compared (DDR4 2400 vs DDR3 2400)

Rise of the Tomb Raider - https://www.youtube.com/watch?v=Bp45DQo8mUM
F1 2017 - https://www.youtube.com/watch?v=OAZylXNgEto
DiRT 3 - https://www.youtube.com/watch?v=-9mrcZUZd48

 

[YoAndy]

I know hardware unboxed just unboxed one of these for testing not too long ago. Let's see if I can't give him a nudge to get it going! hahaha

A12-9800 vs A10-7860K compared (DDR4 2400 vs DDR3 2400)

Rise of the Tomb Raider - https://www.youtube.com/watch?v=Bp45DQo8mUM
F1 2017 - https://www.youtube.com/watch?v=OAZylXNgEto
DiRT 3 - https://www.youtube.com/watch?v=-9mrcZUZd48

So (pretty much) they are garbage.

 

[Shaq-34]

So (pretty much) they are garbage.

I guess AMD don't really have time for BR right now, it is not even promoted or marketed. Plus it is bad PR compared to Ryzen - it's hard to reason why one should invest in Bristol Ridge - but this is strictly my opinion.

Anyway, hopefully, bios updates will iron out the frame spikes.

 

[goldstone77]

AMD's 35W "Bristol Ridge" A-Series APUs Now Available at Retail
by Anton Shilov on September 6, 2017 10:00 AM EST

We have some good news for low-power AMD builders this morning: AMD has (finally) begun to sell the 35W versions of their "Bristol Ridge" desktop APUs. Overall the company has released 3 35W retail Bristol Ridge SKUs, the A12-9800E, A10-9700E, and A6-9500E, with these chips fleshing out the low-power segment of AMD's AM4 platform through the end of the year.

AMD originally released its Bristol Ridge A9000-series APUs to OEMs in mid-2016, targeting desktops and laptops. The SoCs integrate one or two Excavator v2 modules (two or four x86 cores in AMD’s nomenclature), a Radeon R5/R7 iGPU featuring AMD’s GCN 1.2 (3rd generation) architecture and up to 512 stream processors, a dual-channel DDR4 memory controller and so on. Earlier this year AMD finally decided to release a rather broad lineup of its 7th-generation A9000-series APUs on the retail market, enabling end-users to build their own inexpensive AM4 systems, essentially popularizing the AM4 ecosystem compatible with the company’s latest Ryzen processors in general.

The originally planned retail lineup for Bristol Ridge included 11 SKUs with a mix of 35 W and 65 W TDPs, but until now, only select 65 W SKUs were available. In the meantime, the interest towards inexpensive Mini-ITX systems is high these days. AMD spent a little over a month after the formal introduction of the family to bring the 35 W A12-9800E and A10-9700E APUs to store shelves. In the last days of August, the new chips hit store shelves across the world: they can be purchased in the U.S., Europe and in Japan, a clear indicator that they are indeed available broadly.

The AMD A10-9700E costs $85 – $91 in the U.S., whereas the more advanced AMD A12-9800E is available for $105 – $113, depending on the retailer. Meanwhile, the A6-9500E, the most basic 35 W Bristol Ridge APU, can be purchased for around $55 – $58. In Europe and Japan, prices are traditionally higher, see the table below for details. Overall, the 35 W APUs from AMD do not look very affordable, but it is typical for chipmakers to charge extra for energy-efficient parts.

From AMD’s standpoint, the main purpose of Bristol Ridge CPUs and APUs is to fill the gap in the AM4 lineup before the company releases Raven Ridge APUs and cheap CPUs based on the Ryzen microarchitecture later on.

Let's see if they can iron out the issues with stuttering. A10-7860K showed noticeably smoother game play. For $113 it's inline with what I expected. It does make me hopeful for the next generation based on Ryzen and Vega where we can expect 50% more shader cores.

The Raven Ridge APU has 4 ZEN CPU Cores and the IGP gets 704 Shader Processors, which is substantial I must say. The shader procs should be similar to the Vega generation. The engineering sample APU is tagged under SKU 2M3001C3T4MF2_33/30_N. And from that we can extract a 3.0 GHz base clock frequency with a 3.3 GHz boost (again this is an engineering sample). The APU would SMT to 8 treads. The 704 shader processors run at 800 MHz. It has 2MB (4x512kB) of L2 and 4MB of L3 cache. The GPU thus has 11 Compute Units and with 64 Stream Processors per CUthat forms 704 Stream Processors.

There is a rumor that there will be an even more high-end APU based on Raven Ridge, with 1024 Shader processors, which is the equivalant to a RAdeon RX 560.

Click here for a link to the article

 

[-Fran-]

The 1800X has:
- Total L1 Cache -> 768KB
- Total L2 Cache -> 4MB
- Total L3 Cache -> 16MB
- Cores -> 8 ("4+4")

The top of the line BR APU would be:
- Total L1 Cache -> ~256KB ?
- Total L2 Cache -> 2MB
- Total L3 Cache -> 4MB <--- This is the interesting bit I like. They managed to squeeze L3 into the APU!
- Cores -> 4 ("4+0" I would assume)

Cheers!

 

[goldstone77]

The 1800X has:
- Total L1 Cache -> 768KB
- Total L2 Cache -> 4MB
- Total L3 Cache -> 16MB
- Cores -> 8 ("4+4")

The top of the line BR APU would be:
- Total L1 Cache -> ~256KB ?
- Total L2 Cache -> 2MB
- Total L3 Cache -> 4MB <--- This is the interesting bit I like. They managed to squeeze L3 into the APU!
- Cores -> 4 ("4+0" I would assume)

Cheers!

Click here for link

 

[8350rocks]

I expect Zen2 to bring 5--10% increase over Zen IPC.

I expect 10-15% IPC +5-10% clockspeed.

 

[juanrga]

The 1800X has:
- Total L1 Cache -> 768KB
- Total L2 Cache -> 4MB
- Total L3 Cache -> 16MB
- Cores -> 8 ("4+4")

The top of the line BR APU would be:
- Total L1 Cache -> ~256KB ?
- Total L2 Cache -> 2MB
- Total L3 Cache -> 4MB <--- This is the interesting bit I like. They managed to squeeze L3 into the APU!
- Cores -> 4 ("4+0" I would assume)

You are mixing Bristol Ridge with Raven Ridge.

 

[juanrga]

I expect Zen2 to bring 5--10% increase over Zen IPC.

I expect 10-15% IPC +5-10% clockspeed.

Don't say that to people that expect 5.7GHz clockspeed at stock settings!