AMD's Ryzen CPUs (Ryzen/TR/Epyc) & Vega/Polaris/Navi GPUs

Any Raven Ridge laptops announced?
Would love to pickup something thin and lite in the 13”-15” size AMD powered, really great to see them back in the game.

 

It's been quiet for about a month, I posted some earlier, here's a list that was updated on 3/7/2018, IDK how complete it is:

List Of Ravenridge Laptops (Ryzen 5 2500U, Ryzen 7 2700U) (Updated 03/07/2018)
http://apusilicon.com/list-of-ravenridge-laptops-ryzen-5-2500u-ryzen-7-2700u/

I found mostly a lot of repeats of laptops on that list...google search sorted by date over the last month:
https://www.google.com/search?q=Raven+Ridge+laptop

 

MSI B350M Gaming Pro review
by Hilbert Hagedoorn on: 03/15/2018 08:56 AM
http://www.guru3d.com/articles-pages/msi-b350m-gaming-pro-review,1.html

" MSI B350M Gaming Pro motherboard
A good value proposition at 77 bucks.

Spoiler: Details...

In this article we'll review the MSI B350M Gaming Pro, with the release of Vega based desktop APUs (Raven Ridge) from AMD, it is interesting to peek at the low-cost B350 chipset based motherboard, in this review, we'll pair the €77,- MSI motherboard with a Ryzen 5 2400G and take it out for a spin.

Yes the sub 79 USD/EURO MSI B350M Gaming Pro is a lovely looking Micro ATX motherboard that offers all the basic features you need to build a nice mainstream PC. Especially with Vega based desktop APUs (Raven Ridge) from AMD, the Ryzen 3 2200G and Ryzen 5 2400G this could be a lovely offering.

AMD announced two APU based on vega initially, both compatible with existing B350 and X370 motherboard with merely a firmware update. But X370 might be a notch too pricey and perhaps too much if you can suffice with an integrated graphics solution. AMD released Ryzen 3 2200G and Ryzen 5 2400G that AMD are the first CPUs with embedded Radeon Vega architecture graphics. There are other notable features to mention as well, The APUs now have the official default support of DDR4-2933 in dual-channel operation. And for AMD faster memory is important for two reasons, the first being it reacts better to CPU bound gaming, and secondly, system memory is an elementary link for AMD's APU to get the highest performance possible. For integrated graphics, the memory cannot be fast enough as that system memory is used.

• AMD Ryzen 3 2200G with Radeon Vega 8 graphics
• AMD Ryzen 5 2400G with Radeon Vega 11 graphics.

The APUs are developed under codename Raven Ridge, it has a 209.78mm2 and roughly 4.94 billion transistors. These are both quad-core processors, Ryzen 5 gets SMT (hyper-threading). Thus the 2400G gets SMT enabled, 8-threads. The Radeon 2200G however just has four cores enabled and SMT (threading) disabled as AMD does not want to cannibalize their own Ryzen 5 series.

The Ryzen 3 2200G and Ryzen 5 2400G APUs will be configured as a single CCX configuration (4+0), a 'normal' Ryzen 5 1400 processor typically is confirmed as a 2+2 CCX configuration. With the G series you'll notice some other subtle changes, the base clock frequencies compared towards regular Ryzen 3 and 5 is going upwards, the Ryzen 3 2200G, for example, will get a 3.5 GHz base clock with a boost towards 3.7 GHz. The Ryzen 5 2400G, will take that up a notch and starts at a 3.6 GHz base-clock with a boost to 3.9 GHz. Keep in mind, and in comparison, the fastest clocked Ryzen processors are all clocking in at a maximum 4.0 GHz, plus XFR range. So the APUs are pretty close to that as a theoretical threshold as well. Both parts are rated at a 65W TDP. There are more changes compared to regular Ryzen though, 4MB opposed to 8MB L3 cache, but the same 512K L2 cache per core.

But back to the MSI B350M Gaming Pro, a mainstream mobo that seems to be a good match for the 200G series. it's not an overclockers board, you get just two DIMM slots, but other than that it's fitted with all the basics you need, including Gigabit Ethernet as well as a full speed M2 slot and four SATA3 connectors.

If you pair it with a Ryzen 2000G APU slash processor, please do be aware of the fact that PCIe lanes move from x16 to x8 lanes in the “Raven Ridge” design. Let's head on over to the next page where we'll talk a little more about the architecture and then slowly but steadily dive into the benchmarks.

It's a really long article with lots of images of the motherboard... so if you are interested in the whole thing, the middle part, please go to the site for this one.
....

Spoiler: Final Words, Conclusion

Final Words
If you are on the lookout for a really affordable motherboard to be paired with the Ryzen 2000G series APUs, hey this one ticks all the right boxes. Even paired with a regular Ryzen processor this offering actually would still be a terrific deal, if you are savvy with the one PCI-express slot and M2 slot really. Other than that, performance wise you're not going to see much of a difference compared to X370 motherboards. The one footnote I need to make is that IF you opt regular Ryzen with this motherboards, overclocking will be more limited considering the relatively simple phase design. But even then you'd probably achieve very reasonable results. Yes, that is the word I was looking for, everything about the MSI B350M Gaming Pro is very reasonable, the looks, the features, and sure as heck the price level. See, there is a certain kind of people (not our reader-base), that do not want to fork out cash for a dedicated graphics card. These guys and gals hardly play games, and when they do .. they really do not care too much about resolution and quality. Then there's the kind of person that wants to build a PC merely for the Windows PC experience, browsing websites, work a little, and multimedia. It's for these two types of people where the Ryzen 3 2200G and 5 2400G make a bunch of sense in combo with this 77 Euros costing motherboard. BTw please keep in mind that only the Ryzen 2000G series APUs have that integrated GPU, for regular Ryzen you'll need to purchase a dedicated graphics card.

While you will not reach tremendous high-end levels of CPU performance with an APU, there definitely is plenty to be found to drive a mainstream PC with a mid-range dedicated graphics card. So that's not an issue either. Pricing wise you need to be comparing these new APUs to the Core i3 series, and not the Core i5 series. So why are there no Core i3 products in our charts you might wonder? Well, when Intel was still providing review samples we asked many times, but they always refused irrevocably to send them out. To date that has not changed. Ryzen 3 2200G, for example, rips open a HUGE segment in the market for a starting price of 99 USD, you get four cores and an integrated Radeon Vega IGP that will beat down any intel IGP to date. You have seen the results, you can clock the 2400G to 3.9~4.0 GHz, but that 4 GHz domain has been the same for all Ryzen procs to date. If you take a cheapo B350 motherboard, select your memory wisely and fire off a Ryzen 2000G processor on there with say, a Radeon RX 570 or GeForce GTX 1060, well you are there, rock-solid gaming performance. Make no mistake, I feel that we all need to move to 6 and 8-core processors as these simply offer a snappier and faster PC ecosystem with all the things we do simultaneously on a PC these days. That doesn't count as much for gaming though, a 4-core CPU in many scenarios offers the better value.

Small tip: if you pair the B350 motherboard with the Ryzen 2000 G series then AMD is going to offer you flexibility here as you may increase that framebuffer size.

Basically in your BIOS look for UMA Size, the BIOS will default to 1GB Frame buffer. For best performance, this should be set to 2GB from the BIOS.

1. MSI BIOS: go to Settings / Advanced / Integrated Graphics Config and set “UMA Frame Buffer Size” to 2GB
2. Gigabyte BIOS: Go to Chipset and set “Integrated Graphics” to “Forces”, UMA Mode to UMA Specified and UMA Frame buffer size to 2GB.

Boom all of the sudden you have a 2 GB framebuffer - Upcoming BIOS versions will support even larger options such as 4GB and 8GB.

Intel HD 630 IGP for example used in the 8700K we've shown, is over half slower, utilizes just 256 MB and hardly runs any game if it works at all. The four-core Ryzen 3 & 5 G series offers solid value for money. I am also happy to report that the memory compatibility and stability has been enhanced greatly. If you pick your memory right, you will not have any issues. In the BIOS enable the XMP, and you'll boot straight into Windows with your new timings. But again, I find four-core processors to be less attractive as playing with all the six and eight core processors recently simply feel faster and better in your desktop environment and applications, but it is a very fine and very sufficient processor series though.

Prices and value
As mentioned, at a selling price of 77 Euros and $79 in the USA really, this is just terrific value. Obviously, you are more limited with the B350 chipset in terms of a max 32GB memory, just one M2 slot and four SATA3 ports, but for 90% of the people, that's plenty enough. So the balance between features, design, quality, and versatility compared to price, hey it's just really good.
Performance
Whether you purchase a 300 USD motherboard or one that is 79 USD, the baseline APU or processor performance simply is the same on all motherboards and brands There will be tiny marginal differences. Here again, the difference between B350 and X370 would be the same. Your differences are to be found in build-quality (phase design/components and added 3rd party stuff). If you want to heavily tweak a processor, hey you'll still get there albeit you'd be a little more limited. So the performance overall is the same (pretty much). For this motherboard specifically, it has been around ever since the original Ryzen launch, so it's tweaked and refined. That did show in our benchmarks, but yeah it's merely a percent here and there., these are the differences we're talking about, and in reality, you'll have a hard time (if at all) noticing such a marginal difference. So performance really is spot on.

The Memory
We mentioned DDR4 memory a couple of times already; the new AGESA based BIOSes have improved a lot. 3200 MHz is a non-issue with the right memory, we recommend 2933 MHz or better when you build a gaming rig. And if you will be playing games over the APU, faster memory is always better as your Radeon Vega IGP will be actively utilizing that system memory. This motherboard has two DIMM slots, thus 32 GB is your maximum. We used FalreX modules from G.Skill, you simply activate the XMP profile in the BIOS, and after a reboot, you're all good to go and rocking at 3200 MHz.

The Power
The tested setup (IGP no Dedicated graphics card) at idle will sit in the 27 Watt range, with a dedicated graphics card installed (GeForceGTX 1080 / 16 GB memory / SSD and the motherboard) that rises towards roughly 40 Watts. When we stressed the processors with a Prime 1024M run we reached roughly 80~90 Watts, that's low enough for what it needs to be, but a bit higher than I anticipated it to be. Overall though, this is certainly decent enough to see. So overall, these are acceptable values with a four-core product.

The Tweak
Ryzen 3 2200G and Ryzen 5 2400G have been able to reach 3900 MHz on all four cores with merely a crappy stock cooler. The 2400G, however, needs a lot of voltage, I halted at 1.475V at the close to 4 GHz domain. But it would have liked a bit more. Perhaps liquid cooling could have solved a thing or two, as again I have been testing with the stock cooler. The Ryzen 3 2200G I left at default voltage, as it reached 3900 MHz on all cores perfectly fine. That's a great tweak, offering good value for this 99 USD part.

The conclusion
You just cannot complain about what you have seen tested here. B350 should be entry-level stuff, but really MSI made something mainstream out of it. The looks are pretty decent if you ask me, the performance is there, you have some tweaking options. And while you are more limited with one M2 slot and four SATA ports, the reality for the most of us is that that is enough for a fully fetched PC. You can argue the single PCIe x16 slot, but here again, it's pretty much the one slot most people need. Two extra x1 PCIe slots will get you a little-added flexibility for the 000.1% that get goosebumps of the words 'embedded audio' and still purchases a dedicated soundcard or say an Ethernet jack or WIFI. You can use it for regular Ryzen, but B350 is pretty sweet for the new Ryzen 2000G series which by the way has fewer PCIe links available. Your graphics card will get just eight Gen 3 PCIe links, no biggy in the performance segment, but worth a note I'd say. People that have a need for just an Internet or HTPC will love this gear. When you build a gaming PC with a dedicated graphics solution, the sky is the limit at value prices. you do need to factor in though that for Ryzen gaming you need reasonably fast frequency memory, and that is more expensive on the total bill. An APU or Ryzen proc with this MSI B350M Gaming Pro and an SSD will result in a proper PC for your everyday workload and gaming experiences. The MSI B350M Gaming Pro sells at just under the 79 USD/Euro marker, it's simply great value for your money.

ATH +++

- Hilbert out

 

ASUS X470 Motherboard manual leaks, and reveals a lot of info
by Hilbert Hagedoorn on: 03/18/2018 08:37 AM
http://www.guru3d.com/news-story/asus-x470-motherboard-manual-leaksand-reveals-a-lot-of-info.html

"At German forums, a user has discovered ASUS X470 manuals, including specification summaries for the upcoming Ryzen 2000 motherboards. The boards include the PRIME X470 PRO, but many more boards are listed.

In their forums, hardwareluxx user emissary42 posts building block diagrams of several motherboards, including the TUF X470 PLUS Gaming, Prime X470-PRO, and the ASUS Strix X470-F Gaming.

The block design diagrams pretty much reveal most of the features of the upcoming motherboards, all mobos will get two M.2 slots, whereas X370 had a single M.2 mostly slot. It, however, looks like it that only the primary M2 slot run at x4 PCIe lanes. Ryzen has four lanes dedicated in the processor for one M2 unit, any other would be linked through the chipset at x2 lanes. The designs also show a standard ATX form factor with four DDR4 compatible DIMM slots.

The specification sheets specifically show a difference in 2nd Generation Ryzen and 1st Generation Ryzen, when looking at DDR4 memory support, Zen+ seem to bring support for faster memory, the ASUS feature sheet shows Ryzen 2000 (Pinnacle Ridge CPUs) to support memory up to 3466 MHz. So, there is some gain to be found there as well. Overall, exciting times alright, we can't wait what AMD brings to the table next month.


Theme: 2nd Gen AM4 Motherboards - News, Specs, Reviews & Images (2018)
https://www.hardwareluxx.de/communi...ews-bilder-2018-a-1188041-5.html#post26215749

ASUS X470 2nd Gen Ryzen Motherboard Layouts And Specs Break Cover In Leak
by Shane McGlaun — Sunday, March 18, 2018
https://hothardware.com/news/leaks-asus-x470-mainboards

 

AMD Ryzen 7 2700X and 5 2600 Benchmarks got posted in a SiSoftware Review
by Hilbert Hagedoorn on: 03/17/2018 11:34 AM | Source
http://www.guru3d.com/news-story/am...reviews-was-shortly-posted-at-sisoftware.html

"SiSoftware, who you know from SANDRA software had an article online for the AMD Ryzen 7 2700X & Ryzen 5 2600. In the content, benchmarks have been posted based on Sandra results, that means no gaming results.

"SiSoft has tested a Ryzen 7 2700X and 5 2600, you, however, should focus on the performance metrics between a Ryzen 7 1700X and the new Ryzen 7 2700X. You'll notice comparative variances running into 10 to even 20% depending on the target workload, which is not bad really. According to the SiSoftware reviewer Ryzen 2000 brings noticeable improvements in bandwidth and latency, perhaps even further finetuned with future firmware releases. Ryzen 2000 also bumped up the default memory frequency, with 2933 MHz DDR4 memory:

SiSoft 'Ryzen2 is a great update that will not disappoint upgraders and is likely to increase AMD’s market share. AMD is here to stay!' Ironically all All Ryzen 7 2700X SiSoft benchmarks are still available here. For Ryzen 5 2600X values go here. Check the benchmarks below, all based on synthetic Sandra runs. Props to videocardz for this spot.
"
It looks like SiSoftware pulled the review...

" Not Found
Apologies, but the page you requested could not be found. Perhaps searching will help."

Here's the link in case it comes back to life

• AMD Ryzen2 2700X CPU Performance, AMD Ryzen2 2700X Cache & Memory Performance.

 

12 Months With RYZEN, Thoughts From An Early Adopter

 

That's great news on the 2700x, the cpu that I will be looking to upgrade to. With the x470 chipset having better memory performance, that may be in the cards as well. Just give me 400mhz+ overclockibility, and I will be super happy!

Sent from my SM-G935T using Tapatalk

 

Still no clues about laptops with desktop Ryzen... I think that Clevo won't release AM4 version of P870/P775 etc...

 

That is what I am waiting to see, the overclock ability.

 

well jets just go crazy here for a minute: the new node is on 12 nm, so that would be a "linear" theoretical performance increase of +16.7% vs. the current 14 nm at identical power usage. interestingly, this is exactly the range that was shown in the leaked but now pulled sisoft sandra review of the upcoming 2700X vs. its 1700X predecessor, ranging anywhere between 10 and 20% boost (granted, this is also based on a 14-15% clock boost difference between the two cpus )

Link: https://www.notebookcheck.net/AMD-R...-review-available-from-April-19.289396.0.html

The 2700X has a turbo boost of 4.35 Ghz. If we assume that AMD designs its upcoming CPUs with the same safety margins in place as Gen 1 Ryzen, coupled with the Turbo of the 1700X at 3.8 Ghz and the fact that max. overclocking on Ryzen 1 is approximately 4.0 - 4.1 Ghz on air and/or water, we can calculate the following possibilities:

1) 1700X max Turbo 3.8 Ghz, max. stable OC 4.1 Ghz, that equals a plus of +7.8 % OR 3 standard multis of 100 Mhz each
2) 2700X with a max Turbo of 4.35 Ghz could potentially mean +3 standard multis of max. OC headroom (= 4.65 Ghz) or +7.8% OC ability (= 4.69 Ghz), thus resulting anywhere in the range of 4.6 - 4.7 Ghz
3) If we assume an ideal linear proportion of OC ability and improved process node, thus +16.7%, we would have: 4.1 Ghz + 16.7% = 4.78 Ghz, so roughly 4.8 Ghz

So...am I crazy to be hoping for max. stable OC in the range of 4.6 to 4.8 Ghz for good sample CPUs of Ryzen 2? Man that would be awesome....

ESPECIALLY since that would then automatically also apply for the next round of Threadripper CPUs

 

The problem is AMD is already hitting the power dissipation limits for the ThreadRipper, and higher core count Epyc is already underclocking - more cores require more power at the same speed, so underclocking is needed to meet power target limits.

Any power / frequency / performance increase, even with the same core count is going to also have the same ceiling of power limits. Something has to give somewhere to accommodate that additional power and thermal requirement.

So, IDK what to expect in actual use.

 

lets hope thats where the new node at 12nm steps in to allow clocks increase inside the same power envelope i dont think we will see higher core count yet, thats probably gonna be in 2019/20...

Sent from my Xiaomi Mi Max 2 (Oxygen) using Tapatalk

 

I agree with you here, which due to the 8 core TR being out, the $500 2800X would make less sense, especially with the performance comparison you pointed out between the 1700 and 1800X chips. This also frees up a bit of room in binning for the TR line this round, allowing for them to hit the ground hard with the top bins instead of scraping the top while trying to put out two high performance lines. It may also allow for a non-X variant on the 1950X successor, which the bios seemed to suggest may happen with the 1920 and 1900 chips last time that never materialized. Either way, they now know they have a place in HEDT, which also means better board designs are likely for the 499 chipset boards, or whatever they have to call them (there is a server x499 from the past, meaning there may be platform naming issues, so just wanted to put that out there).

Either way, I think you and I are waiting for the 7nm variant, whether first or second gen 7nm. Considering the wait to get PCIe 4.0 out and PCIe 5.0 being published in early 2019, along with DDR5 published this summer and finalized, there is a chance that waiting for the 7nm+ refined chip could give the real large jump we want, specifically on the use of EUV lithography, which will increase yields among other benefits. I'm still reading to catch up on this thread though.

 

See the Gamers Nexus video on this topic. It is FUD and nothing of real concern ATM.

I think your numbers are high. You are assuming the single core boost is where the majority of the power will come from, are not looking at the refinements in IMC (there was talk of changing things with suppliers, etc., but I have not seen HARD info on this, just rumors and refinements on latency), and potentially infinity fabric, along with large refinements in the senseMI and power management which will allow for all cores to easily hit higher speeds at stock before the throttle down. I'm still saying expect 4.4-4.5. It is a safe expectation for all core (the only true OC that counts) and is a 10% jump. 4.3 should be attainable, as @hmscott mentioned there are other issues to consider.

We will see OC numbers soon enough. But, so far, everything people are making a big stink about, like the 5ghz Zen2, etc. have actually been something I have said since August or so last year. It is based on the same source, GloFo, as well. But no one has mentioned the rumors of the Epyc chips at 7nm going to TSMC and the graphics 7nm at GloFo, flipping the script, or anything new and groundbreaking, other than the new x470 chipsets seem like the new chips may support extra PCIe from the CPU compared to the x370 chipsets, but those also could just be from the chipset, and yank to that...

 

And to play devils advocate and present the worst case scenario... the 1800X was the top (mainstream) bin of Ryzen 1, 1800X as top bin XFR'd to 4ghz, with most lesser chips OCing to around that, and unicorns reaching beyond 4.2ghz

Top bin 2700X is XFR to 4.35ghz, similar headroom suggests most refresh chips falling within 4.4-4.5

Which would be nice, but really I think the 2700X needs to be compared to 1800X, not 1700X

I wonder if AMD are trying this time to engineer a wider gap between 8 core X and push the better bins across to the TR4 socket to make the 8 core TR less pointless.

 

I would love to see 4.8 GHz. If 4.8 is achievable on the TR that would be a worth while upgrade to me. 4.4 or 4.5 is not quite enough. My fear is with XFR2 that they have used up the headroom for overclocking.

 

GDC 2018 - March 19th - 23rd Moscone Center, San Francisco, CA
AMD Presentations at GDC 2018

GDC 2018 - March 19th - 23rd Moscone Center, San Francisco, CA
AMD Announces Real-time Ray Tracing Support for ProRender and Radeon GPU Profiler 1.2 by Nate Oh on March 19, 2018 8:15 PM EST...

GDC 2018 - March 19th - 23rd Moscone Center, San Francisco, CA
Expanding DirectX 12: Microsoft Announces DirectX Raytracing by Ryan Smith on March 19, 2018 1:02 PM EST "To many out there it may seem like...

 

dont get me wrong, i just took the single core boost as a vantage point from which to base oc headroom on. the 1800x had boost clocks of 4-4.1 ghz (with xfr) and thats exactly where most ryzen 1 cpus would max out at. so assuming that there wont be a 2800x i used the 2700x boost clocks of 4.35 ghz as a reference point

and yes, worst case could indeed be that exact boost clock, namely roughly 4.4 ghz! so im guessing anything in the range of 4.4 to 4.8 ghz could be in the realm of possibility.

Sent from my Xiaomi Mi Max 2 (Oxygen) using Tapatalk

 

Nvidia Thread
Nvidia GPP's first victim(s) https://www.reddit.com/r/Amd/comments/85n378/nvidia_gpps_first_victim/ "...

 

I just read part of this article, albeit a slightly old one ( https://www.gamersnexus.net/guides/...-guidelines-how-to-kill-cpu-with-safe-voltage), and it talks about AMD CPU voltages above 1.3V causing CPU degredation - I'm sure I've seen users on here running in the region of 1.4V on their AMD CPUs so just wanted to bring this to your attention so you don't damage your CPUs, assuming you're not aware of this article or the potential dangers.

 

Yup same goes for killing Intel 8700 / 7700 series CPU's, that's why Intel recommended *not* overclocking their unlocked CPU's

Intel tells Core i7-7700K owners to stop overclocking to avoid high temps
By Paul Lilly May 04, 2017
https://www.pcgamer.com/intels-tells-core-i7-7700k-owners-to-stop-overclocking-to-avoid-high-temps/

https://www.google.com/search?q=amd+ryzen+threadripper+safe+voltage+bad+for+cpu

https://www.google.com/search?q=amd+ryzen+threadripper+maximum+safe+voltage+bad+for+cpu

https://www.google.com/search?q=ryzen+safe+soc+voltage

https://www.google.com/search?q=amd+ryzen+threadripper+high+voltage+bad+for+cpu

 

That Intel related stuff is just phooey, it's safe to overclock Intel & AMD CPUs. Sure if your cooling system is bad or if you were unlucky and got a bad thermal interface job between the internal heat spreader and the CPU chip then you're gonna see spiking of temperatures close to max allowable temperatures for the processor. Intel just covering their ass with their 'recommendations' for not overclocking K series CPUs.

 

Yeah, but the same CPU killing effect over time has already been reported here, not able to hold OC after so many months at high OC, and Mr. Fox's first 8700K randomly died, who knows why but it seems like a bad omen for what might be coming long term high OC on Intel CPU's.

Always run under the maximum "safe voltage", reduce daily OC to something safe and keep it cool for long life.

That's always been the way, and my CPU's last for decades. Never had one fail due to OC, and I've had plenty.

 

Well, yeah, you gotta do that if you want your CPU to last. When did his CPU die, he didn't mention it in the desktop overclock lounge? I'd be curious to ask him about it to find out how he was running it, I think he generally runs less than 1.45V most of the time.

 

I was watching his new 8700K performance - better than the first one - and saw him mention the 1st one died.

IDK if this was the post I saw... there may be more posts:

http://forum.notebookreview.com/threads/clevo-overclockers-lounge.788975/page-1535#post-10696183

 

I understand, but let me put it another way: for the original Ryzen, the all core max OC was relative to the same as the single core top clock. I see no reason to expect more even with the 12nm die shrink. The silicon wafers, which are part of the equation, may have changed slightly changing silicon limitations, and the process at 12nm may have changed the equation, same with the architecture refinements. Even with that, and unlike with Intel, I think a fair amount of the boost came from other refinements helping with memory latency and other latencies in the CPU. Yes, some came from increased speed, which isn't really a true IPC gain, but rather increasing the speed so that you perform more cycles in the same amount of time, whereas IPC is the number of instructions processed per cycle. This is why testing true IPC requires locking the CPUs to a specific speed, but then you need to control for other factors (which is impossible) to try to get close to the real IPC of the chips.

Now, XFR2 needs put into context. First, we all know that AMD did not leave room on the table for overclocking, at least in regards to Ryzen. There is no reason to expect different with this round. They worked hard on increasing headroom, though, which means they got a descent boost in speeds due to that and their refinements to SenseMI and XFR2. I cannot find the awesome video that explained it, but it showed that the all core, especially out to 6 or 7 cores, then can run at the higher speed compared to like a 2 or 3 core boost on the original XFR. It also was examining Intel's changes to their boost and voltages and I believe it may have been an adoredTV video, but am unsure (but here is an Epyc yield video relevant to what may be expected from the upcoming process: ).

In any case, this means they are giving stock users more bang for their buck. It also needs remembered that multi-core on AMD (SMT vs HT), AMD's solution works better than HT. The better IPC and speed abilities on Intel, though, make up for this deficit, in fact keeping the performance crown. This needs remembered just like AMD's improvement in IPC being compared the Kaby, which is NOT the current Kaby Refresh or Coffee. It still falls short, but closes the gap and suggests hitting at or above Skylake, which once you compare that to Intel's refinement of Broadwell, which was a **** stain, to skylake and the time frame for that, it makes sense and suggests AMD is doing well with their refinement team, even if the speed doesn't match Intel's (you have to remember, they are refining 8C/12T chips, while Intel was doing 4C/8T chips, and some have shown quad chips clocking to mid to high 4GHz ranges, although I have not personally examined the claims because I cannot care about quad core chips anymore (have a 16C). But, the improvement in IPC, if proven true, would show they are catching up, whereas the 7nm process is in a different field, and architecture changes will likely then bring it in parity with logic density (in fact, maybe greater density if the articles about 10nm leaks are true) while also having 5ghz speeds, something that Intel achieved on 14nm with ease AFTER doing a third and fourth refinement. We don't know speeds on 10nm yet. And then there is whiskey lake, which suggests a potential 5th gen or a refresh only months after coffee (I say they called it whiskey lake due to you needing lots of whiskey to deal with the internal issues in Intel right now, but that is just a joke in bad taste).

But, there is something different with AMD and how Intel started to do binning and qualification for chips. AMD, at least with first gen Ryzen, was still seeming to use a similar default voltage for the various chips. If you remember, around the time of Haswell, Intel started varying the stock voltage so that they fell in at least three different voltages. This is a way to categorize and still qualify chips into different bins while being able to kick any that cannot do that to a lower stack, and considering how varied Intel stacks are, it makes sense. It also allows for a higher yield as the voltage rates, so long as in range and chip having proper functionality, it could pass when it would otherwise fail. This is also why Intel leaves so much more performance on the table and has a larger variance in clock rates achievable by overclockers. AMD is really quite consistent, but it does leave one to wonder what benefits could be achieved, if any, from that variable stock voltage if applied to AMD chips. This would be less than Intel, but only because AMD has the consistency and leaves less performance on the table (which admittedly effected growing pains with some chips needing a set voltage and not able to use auto at release, IIRC in seeing reports on such things). It may have also contributed to the finicky nature of the platform as well. But you get the point.

I previously did an analysis on the first leaks showing 6% attributed to the clock speed and the other 9% of the 15% attributed to other factors (it is back in analyses in January and February, mostly talking with Jeks). So, since I have not fully analyzed the new leaks, I have not compared my original analyses to the current numbers, but seeing 15% recur and a 10-20% range, with more benches included, it suggests that it isn't that far off, which is why some are complaining on clock speeds being less than what they expected, even when the performance jump beats what Intel's actual jump is, not their claimed BS each year. If those numbers bear out, or are better once put in the right hands, then the speed isn't as big of an issue due to other improvements, although it hurts with the people that think higher speed is ALWAYS faster, which was proven false long ago during the Athlon and P4 time.

Because of all this, I am sticking with 4.3-4.5 on the 8-core chips, not really analyzing the others. Too many factors, from wafers to refinements, to say anywhere above that. But, that doesn't mean the performance jump won't be sizeable. We must remember, Skylake-X uses Kaby 14nm+ process, so the IPC will be closer to the current 8-core IPC, but with slower speeds, better SMT, actual memory latency and other latency improvements which may or may not match Intel's, etc. It should be applauded, as what is learned here will also be thought about with the 7nm chips, although 7nm is more of the larger architecture tweaks. But, if AMD's refinement did in one refinement what Intel did with two, minus speeds (which we need to look to BW-E 8-core speeds for better comparison on speed on all core rather than Skylake-X, which is Kaby), it shows good progression. But we must manage expectations and hype. This is why I give this book of analysis to you!

You are looking at the wrong voltage. This is referring to the SOC voltage that controls the IMC, which is separate from the voltages that are used on the core, which 1.35 is fine and AMD gave guidance that above 1.35-1.4 will cause degradation. SOC is the IMC, which to see it go so far over the recommended voltage is scary, although the consensus has been 1.2V max for awhile, with 1.15 or less being the preferred voltage. Very important to know which voltage is being referenced here. But it is good info!

 

This is the good thing about the Taichi x399, it does not use the voltages from the CPU, but a much stricter set from the board.

 

The Intel OC problem was related to voltage spikes. Intel also publishes, or used to publish, the max voltage for each category and many OCers would publish guides saying safe voltages for air and water. Now, once you get to a point, matching between heat and voltages just gets easier. But a voltage spike can mean you need to back off the OC you used to use slightly to make sure the spikes don't kill the chip. For novices, that is why Intel advised not to OC. If you can watch what the spike is (meaning top voltage), heat spikes, effects on the VRM, etc., you can come up with something reasonable that is a good overclock and relatively safe for the chip. If you don't know what you are doing and are already pushing voltage limits, then the spike can flash over the amount and vaporize transistors, causing degradation or killing the chip. It is just knowing what you are doing and being aware of the capabilities of your cooling solution.

 

Here is the adoredtv talking about the leaked slide deck and precision boost 2 and XFR2.

 

I think this is the GN video that addresses the problem with CTS.

 

My thoughts on no 2800x is that in time the process of 12nm will improve. Maybe around the time frame of the TR we will also see a 2800x with slightly improved clocks. Not committing to a roadmap allows them to see what happens in the future. As it is now we probably will see the 2700x specs in the TR CCX's, having a committed 2800x would probably make consumers expect it for the TR.

 

If that is true, it would suggest holding back a sku to go against the 8-core Coffee lake released this summer, while dropping around the time of TR. It is possible, but I'm not convinced of such a thing.

 

AMD's response on CTS security report:
https://www.amd.com/en/corporate/security-updates

"On March 12, 2018, AMD received a communication from CTS Labs regarding research into security vulnerabilities involving some AMD products. Less than 24 hours later, the research firm went public with its findings. Security and protecting users’ data is of the utmost importance to us at AMD and we have worked rapidly to assess this security research and develop mitigation plans where needed. This is our first public update on this research, and will cover both our technical assessment of the issues as well as planned mitigation actions.

The security issues identified by the third-party researchers are not related to the AMD “Zen” CPU architecture or the Google Project Zero exploits made public Jan. 3, 2018. Instead, these issues are associated with the firmware managing the embedded security control processor in some of our products (AMD Secure Processor) and the chipset used in some socket AM4 and socket TR4 desktop platforms supporting AMD processors.

As described in more detail below, AMD has rapidly completed its assessment and is in the process of developing and staging the deployment of mitigations. It’s important to note that all the issues raised in the research require administrative access to the system, a type of access that effectively grants the user unrestricted access to the system and the right to delete, create or modify any of the folders or files on the computer, as well as change any settings."

Basically, CTS is full of it... and it was quite likely a (failed) attempt to lower AMD's stock price by making some overblown noise.

 

The second paragraph actually has information that is important. This is saying that they figured out attacks on AMD's equivalent to Intel's ME. But, the fact that you can only use them if having unrestricted admin access means ability to actually pull off the attack is low. But, overall, we were all right. It was a pure attack meant to lower the value of the stock. Unlike Spectre and Meltdown, which gave elevated access to steal information, this requires having the admin access to even try, which anyone that has that access can do lots of malicious things.

 

@Raiderman - just saw you have more points than me on HWBot. Want to lend me that 1080 Ti? Just need to borrow it for a minute and I'll send it right back!

 

If anyone can just go out and "buy points", or in this case "borrow points", by purchasing "loot boxes of high end hardware", what's the point?

Roll the dice on getting a "loot box" with "golden samples" of Winning units of hardware, and get more points.

It all sounds like a crude yet immediately direct form of "pay to win"

 

It isn't due to his card being uber good, which I don't know how good it is (GPU OC is not my expertise, I do better with CPU side). It is more that he beat me, we are both in team premamod, and I want to see how it does stepping from a Ryzen chip to a TR chip. But, to an extent, all modern OCing is pay to win, unfortunately.

 

lol dont you think that basically most if not all users do that kinda thing on hwbot to boost their points? that and more

Sent from my Xiaomi Mi Max 2 (Oxygen) using Tapatalk

 

Exactly my point, and as long as you know how the game is rigged, what it takes to win, in this game you only need to learn the rules to spend the $ to buy the points, and win.

 

Not exactly true. If you don't understand what the settings are that allow you to do it safely and not fry your components, doesn't matter if you drop $800+ on a quad core CPU, you are not going to get the score a person that is really good with LN2 will get. Considering the people here are doing cooled air or liquid cooled, this is the best of enthusiast, not elite or extreme. Even in enthusiast circles, there are levels involved.

But, even if you are the best in the world with LN2, you cannot win without dropping the money for that $800 binned quad core. That is where the lottery no longer exists and it has become pay to play, because if you don't pay, you are guaranteed to lose. So, at this level, I don't consider borrowing hardware for points to be an issue or cheating. It is just getting more experience by playing with more hardware. Now, if speaking to the need to have the top hardware in each category for each test to become the top, and the increasing costs of chips, rather than the promise of decreasing costs, even when we have a mature node and the margins of the corp. go WAY up, then, yes, I agree. It is pay to play. But, think of it like a hobby. Let's say you don't have a rally car, but your friend does. You both have a love of the sport and you both are competent at driving. Would you ask your friend if you could take their car out for a spin? Now, I wish we all lived closer together because we could meet up and do projects and overclock and chill. But, unfortunately, we live all over the world (but thank god for the internet and forums like this one to connect us).

But, also, the ask was more in jest, asking for his hardware to beat him with, which is arrogance and just being that guy.

 

To be fair hackers could use these exploits with in combination with other hacking tools for elevating privileges.
But still this way of making public statements with 'RyzenFall' and other marketing crap names would nothing more but actually force me into buying future Ryzen processors.

 

VideoCardZ had a nice 2700x / 2600x data summary chart and links to the SiSoftware benchmarking results from the prematurely posted review article later pulled from the SiSoftware website, in addition to the info already posted:

AMD Ryzen 7 2700X/Ryzen 5 2600 review posted ahead of launch
A review posted at SiSoftware website was quickly removed, but here’s a copy of all important bits. Sorry, no gaming benchmark.
Published: 16th Mar 2018, 19:06 GMT
https://videocardz.com/75381/amd-ryzen-7-2700x-ryzen-5-2600-review-posted-ahead-of-launch

" Conclusion
According to SiSoftware reviewer Ryzen 2000 brings welcome improvements in bandwidth and latency, which should further improve with new BIOSes.

Ryzen 2000 supports 2933 MHz DDR4 memory out of the box, which should provide an overall improvement and better experience (in gaming).

SiSoft:

"Ryzen2 is a great update that will not disappoint upgraders and is likely to increase AMD’s market share. AMD is here to stay!"

All Ryzen 7 2700X SiSoft benchmarks are available here. For Ryzen 5 2600X values go here."

 

PowerColor Radeon RX Vega Red Dragon: triple-fan custom card
PowerColor preparing Radeon RX Vega Red Dragon, in RX Vega 56 form with a fully custom design and triple-fan cooler
By: Anthony Garreffa, Posted: 2 hours, 32 mins ago
https://www.tweaktown.com/news/6129...-red-dragon-triple-fan-custom-card/index.html

"PowerColor is starting to tease their new Radeon RX Vega 56 Red Dragon graphics card, a fully custom RX Vega 56 with a custom PCB and triple-fan cooler.

VideoCardz broke the story, with PowerColor soon after listing the Red Dragon series on their own website with the RX Vega 56 variant.

There could be a Vega 64 version released as well, but the company hasn't made anything official - at least at the time of writing, I'll edit this story if they do.

The company has used what looks to be a custom PCB, while rocking a huge triple-fan cooler that should keep the temperatures as low as they can go, given how hot the Vega GPU and HBM2 run."

PowerColor Radeon RX Vega Red Dragon
Published: 21st Mar 2018, 18:52
https://videocardz.com/newz/powercolor-radeon-rx-vega-red-dragon-pictured

 

2700X TDP is a tad underwhelming for the given clocks.
105W?
What kind of manuf. process did they use?
First the technical specs stated that the manuf. process (12nm) is suitable for high performance parts and high clocks, has lower voltage, and touts better efficiency, and now we see 11% increase in TDP which also rose with a similar/same increase in boost clocks?

Shouldn't the TDP remain at least the same with this kind of clock boosting on the new manuf. process?
Unless 'better efficiency' in this case meant proportional increase in TDP with clock increases (might indicate that power consumption wouldn't suddenly jump too high when clocks are raised)... still, if these numbers are accurate, wouldn't that contradict the technical specs of the process?

Also, why was Skylake included as a comparative intel CPU and not intels 8700k?

 

When I saw that increase for TDP I took that increase as another dimension AMD expanded to improve OC potential.

We won't know what the actual stock power draw, and power draw vs OC range will be until it starts shipping and experienced OC reviewers / users report.

AMD looks like it took advantage of improving every aspect possible in the process change to improve performance, the TDP headroom bump is just another aspect.

Given AMD designed a higher 105w TDP CPU sku, we should see increased power delivery from x470 motherboards, and higher performance cooling options.

 

AMD Ryzen 7 2700X pictured, full review on Monday
Published: 23rd Mar 2018, 12:32 GMT
https://videocardz.com/75489/amd-ryzen-7-2700x-pictured-full-review-on-monday

" AMD Ryzen 7 2700X review before launch
Heads up to those in France. On Monday everywhere in France (and on Saturday in Paris) a new issue of CanardPC Hardware magazine will be available. This magazine will feature two Ryzen 2000 reviews, the Ryzen 2200G/2400G and also Ryzen 7 2700X, which is launching on April 19th.

While this is all interesting, and we can’t wait to see the results, this is in fact, the first picture of retail Ryzen 7 2700X to be made public. No other details were revealed, let’s hope it was tested on X470 motherboard as well.

CPC:

We have found significant improvements on some points … but also an unpleasant surprise that we let you discover!

The full cover:

Source: CPCHardware "

CPC Hardware # 36 arrives on newsstands
POSTED BY: DOC TB 23 MAR 2018
https://www.cpchardware.com/cpc-hardware-36-debarque-en-kiosque/

Spoiler: Details...

" First of all, you do not have to rush immediately to your newsagent: this new number will only be available from Monday everywhere in France (and tomorrow in Paris).

PC Hardware Duck celebrates its ninth birthday! To commemorate the event with dignity, we have put together a summary of choices to make this issue truly collector. On the menu first, the new Ryzen series 2000. We have sifted the Raven Ridge models (Ryzen 3 2200G and Ryzen 5 2400G), which are distinguished by the presence of an integrated Vega graphics chip , and also the declensions Pinnacle Ridgebased on Zen + architecture. These should be announced by the end of April by AMD, with four CPUs planned initially. The Ryzen 7 2700 (X) and the Ryzen 5 2600 (X) are distinguished by the exploitation of the new 12 nm GlobalFoundries process and modifications of the initial Zen architecture. We have found significant improvements on some points ... but also an unpleasant surprise that we let you discover!

Another topic of this 36 th issue: the famous Linky meter which we follow the adventures of meadows. By studying its financing plan, we can see that nearly two billion euros (more than a third of the total budget) are expected thanks to the fight against fraud. It incorporates for this an impressive power relay that allows the operator to cut the power at a distance, to prevent an unscrupulous "user" does not benefit from free electricity.Problem: we were able to easily "hack" this device and restore power to a disabled Linky counter. Even if other protection mechanisms exist, it is still a surprising vulnerability.We also tackle another problem that Enedis will have to face: the imminent update of the IEC 61000-2-2 standard, ultimately much less protective than hoped by the distributor. Finally, we analyze the critical report of the Linky Program and we give you our opinion on the question.

We also offer two background files: the first addresses the technologies of the future. We will be talking about quantum processors, tomorrow's memories (DDR5, GDDR6, etc.), storage, next-generation interface buses, 128K HDR 4D display and super mega Wi-Fi 12G (among others). The second issue is preservation - and recovery - of long-term data. In addition to a first part on the best means of resuscitating old media (diskettes, ZIP, forgotten optical disks, audio cassettes, etc.), we were able to visit the collections of the National Library of France (BnF) and see the means used to archive and preserve the digital heritage. An amazing report made by Dandu.

This new opus also contains various and varied hardware mess tests as well as a power comparison from 35 to 100 €. Believe it or not, but we now find decent models for less than 50 €, and blocks that border on perfection from every point of view for 100 €. We list them. We also offer a list of the most unusual graphic cards, a history of major trials that have shaped the history of high-tech, an exclusive test of 486 DX2 / DX4 and Pentium in a video game, and well other exciting things!"

 

CPC Hardware AMD Ryzen 7 2700X review goes online
Published: 26th Mar 2018, 12:37 GMT
https://videocardz.com/75504/cpc-hardware-amd-ryzen-7-2700x-review-goes-online

"A member of Reddit community has published the graphs and summary from CPC Hardware review, which is available in the latest printed issue starting from today.

Ryzen 2000: lower latency, but high power consumption
Gaming benchmarks, which include GTA5, GRID Autosport, Battlefield 4, ARMA 3, X3:TC, Witcher 3 and COH2 indicate a 4% gain compared to last-gen flagship Ryzen 7 1800X.

The Ryzen 7 2700X tops the synthetic benchmarks, which are based on HEVC to H264 encoding, Mathematica, PovRay, Blender, 3DSMax and Corona tests. There’s a 14% difference between 2700X and 1800X.

The most disappointing issue with Ryzen 2700X is high power consumption, which goes 13W above 1800X. The new flagship, however, has much smoother frequency scaling in multi-core use.

The review was conducted with an A320 motherboard, which does not support XFR2. Although, according to CPC Hardware, changes from X370 to X470 chipset are cosmetic and should not provide any substantial performance differences.





Source: Reddit

CanardPC Hardware summary
https://www.reddit.com/r/Amd/comments/8789bs/canardpc_hardware_summary/

Please buy CPC if you are French (online if possible because of their distributor problems... ) but:

• Improved latency, slowly closing the gap with Intel
  
  
• Much smoother turbo
  
  
• Fantastic "brute force" performance, still worse than a 8400 in games.
  
  
• Power consumption goes wild.

According the CPC, the X470/B450 and X370/B350 chipsets are ENTIRELY IDENTICAL, they are literally renames. XFR2 Enhanced and Precision Boost only depend on the board but because not all old gen boards were up to spec AMD wanted to avoid a situation were some boards had XFR2/PrecBoost and some board didn't and decided to rebrand the chipset and impose more stringent standards."

 

If these numbers are accurate, they would have been more attractive if the TDP remained the same.
At this junction, it looks like power consumption INCREASED... which is in contradiction to the manuf. process specs.
Oh well... let's wait until the CPU is actually released so we can see independent benchmarks.

 

TBH, this is about what I expected. 12nm on a + or ++ might be more competitive but it is essentially a brand new process. This is why I think down the line, as the process matures, we will see better in the TR's and maybe even a 2800x offering.

This is also why 7nm will level the playing field for AMD. Either it competes against Intel with the long in the tooth 14nm or if Intel offers a new process it will be fairly new as well. So it looks like AMD will be back in the game and hopefully it will stay there.