Monday, September 30, 2019

AMD B550 Chipset coming soon


Reports are coming in that AMD and its partners are readying motherboards based on a new mainstream B550 chipset. Following the X570, B550 will be the second socket AM4 chipset designed specifically to go with the new Ryzen 3000 CPUs while still being backward compatible with 2nd gen chips. No word on pricing yet but it’s pretty safe to assume that it will be significantly cheaper than its enthusiast grade sibling.

B550 will succeed and replace the aging B450 which probably is the most popular midrange chipset right now thanks in no small parts to its exceptional value. With B550, AMD would be looking to carry the tradition on, added some incremental features. For starters, the new chipset will feature more SATA III ports and general purpose PCIe (Gen 2/Gen 3) lanes to handle the I/O duties. B550 will retain its predecessor’s overclocking ability – meaning it’ll just be as easy to OC a lower priced (and clocked) Ryzen processor as before. On a side note, it’ll be interesting to see if the passive chipset cooling (with a fan) trend of the X570 based boards will continue on the B550 as well.   

B550, however, will miss on some key highlights compared to the flagship X570. The implementation of the PCIe Gen 4 will be one of those. This means it’ll be a PCIe Gen 3 x4 interlink between the motherboard chipset and the CPU, unlike the PCIe Gen 4 x4 connection found in the X570 chipset. Fortunately, this will not affect real-world performance in any tangible way as we’ve already seen Ryzen processors performing identically on both X570 and B450 (also PCIe Gen 3) based motherboards. Plus, the downstream PCIe Gen 4 lanes from the processor itself should be there for the motherboard manufacturers to put those in good use. Reports also suggest that the population of the USB 3.2 Gen-2 (10 GB/s) ports may not see a rise – at two, it'll be identical to previous generation chipsets but much lower than the eight found in X570.

All in all, AMD’s new B550 chipset could be a perfect solution for budget builders among us, especially those looking for a reasonably future-proof Ryzen 3000 PC but still holding out due to the high price of X570 motherboards.

Tuesday, September 24, 2019

RX 5700 XT and RX 5700 Review: Navi makes a solid impression

The past two or three years have been a mixed bag for AMD's graphics wing, unlike its CPU division which has been on a role with Ryzen. The last time it had a major successful launch was three years ago, with Polaris GPUs. The pricey and power hungry Vega, while working as a stopgap solution, never quite took the fight to Nvdia's camp. And considering the 7nm Radeon VII was never meant for pure gaming, AMD had to bring something to table that could not only disrupt Nvidia's Turing line-up but pave the way for future Radeon products.

That something came in the form of 7nm Navi GPU, launched back in July along with 3rd generation Ryzen processors. Apart from bringing all the benefits of a process-shrink, Navi features a new and improved RDNA (Radeon DNA) graphic architecture and replaces GCN (Graphics Core Next) based Vega within AMD's product stack.

While RDNA isn't a complete architectural overhaul, many of its aspects have new elements introduced in almost every level. With RDNA, AMD's focus is on improved single-threaded performance and better utilization of fixed-function hardware. That means RDNA is better suited for gaming whereas GCN was better at handling complex HPC type workloads - part of the reason why AMD still keeps the GCN based Radeon 7 around.

Thursday, September 19, 2019

Update Round-up: AMD fixes Ryzen, Microsoft breaks Windows 10

Updates and patches have become big parts of our daily digital life these days and last few days both AMD and Microsoft have dropped some big ones; although, with very different results.

First up is AMD and its promised fix which we reported about in our in-depth coverage of Ryzen 3000 processors' boost-clock controversy. If you haven't yet read it then please do so for some much needed context and insight on the matter. Here is a TLDR version:.

With the recent debates over boost-frequency of  Ryzen 3000 series desktop processors, AMD promised an impending update that would fixate any oddities in boost behavior that bars the processor cores to hit the max-boost ceiling. The chip-maker claimed in a blog post that it has identified a bug which is responsible for sub-optimal performance of its boosting algorithm under certain workloads.

So, does the fix works? For the most part, yes it does. Some review-sites have already got their hands on the updated firmwire containing the fix and put their Ryzen CPUs through tests. Techpowerup's Ryzen 9 3900X clocked higher on the new BIOS with AGESA 1.0.0.3ABBA and even boosted beyond its 4.6GHz rateing which wasn't possible with the old BIOS.

Sunday, September 15, 2019

India adds two more supercomputers to strengthen researches: Aims to deploy over 60 in next three years


India's HPC (high Performance Computing) initiative has finally gathered some much needed momentum as the country adds two more supercomputers to its growing computational repertoire. According to press reports, these new supercomputers are being deployed at IIT (Indian Institute of Technology) Kharagpur and IISER (Indian Institute of Science Education and Research) Pune. 

This initiative comes as a part of  National Supercomputing Mission undertaken by India's Department of Science and Technology (DST) and Ministry of Electronics and Information Technology. The mission is currently in its first phase in which a series of six supercomputers are to be installed in education and research facilities across the nation. The first one, named "Param-Shivay", is already functional at IIT Varanasi and can churn up-to 837 Teraflops of data. By December this year, another three will be made available to IIT Kanpur, IIT Hyderabad, and Jawaharlal Nehru Institute of Advanced Studies (JNIAS) which will conclude the first phase.

Saturday, September 14, 2019

AMD roadmaps reveal Zen 3, RDNA 2 on-track for 2020 lunch: Zen 4 by 2022?

It's roadmap time again! Guru3D.com have published couple of new ones from AMD that talk about things to come in both CPU and GPU fronts. The slides, which seems to have been reserved for corporate briefing, make reference to Zen 3, RDNA 2 and even Gen 4. While nothing concrete or out of line of what we already know, this confirms that AMD is on track with future iterations of its current-gen designs.


Here, the CPU roadmap shows that AMD already has the design phase of  Zen 3 completed and currently is working on the design of its successor, Zen 4. Future CPUs (Ryzen 4000?) based on Zen 3 core-architecture will be built on a 7nm+, i.e. a more refined and mature 7nm, fabrication process at TSMC. Moreover, Zen 3 will utilize an EUV (Extreme UltraViolet) lithography compared to Zen 2's DUV ( Deep UltraViolet). With Zen 3, AMD's stated goal is enhanced power efficiency but the higher transistor density granted by the process refinements can be used towards raising the clock-speed as well.

The underlying timeline in the above roadmap spans five years and lack finer graduations which means no specific launch dateline. But it looks like AMD has a 2020 launch in mind for Zen 3.


It's not all quiet on the graphics front either! This year we saw AMD introducing its Navi GPUs based on company's latest graphics architecture dubbed RDNA. The next update to that will come in form of RDNA2 which is in design, also destined for 7nm+ process. At this point in time, we don't know much about RDNA2 and it could be at least a full year off from now considering AMD is yet to release its full RDNA product stack.

All in all, AMD sure seems poised to take its newfound momentum into the future and executed well enough, can usher a new era for the company!

Friday, September 13, 2019

AMD Ryzen 3000 Boost-Clock controversy and what you need to know



If you’re an enthusiast like me, you must be aware of the ongoing controversy over the boost-clock speeds of AMD’s recently launched Ryzen 3000 series processors. What started with some early adopters finding it hard to hit the rated boost frequencies on their respective Ryzen CPUs has now found its way into more serious contention, supported by claims from both reliable and unreliable sources. Things have specially heated-up after well-known overclocker “der8auer” conducted a survey which puts a big question mark on the ability (or the lack there-off) of most 3rd gen Ryzen CPUs to operate at their advertised max boost-clock! While this controversy doesn’t necessarily jeopardize Ryzen’s status as the current champ of desktop processors on the whole, it sure creates a cloud of confusion or two in the mind of average consumer and enthusiast alike. Here we’ll try to dispel any such confusion by going deep into the matter and seek the truth.

Wednesday, February 20, 2019

AMD Radeon VII Review: The highend contender

From the center stage of CES last month, AMD gave us the outlines of two major products. One was about Ryzen gen 2 and we expected as much; the other, however, caught many of us off-guard. Not to be out-done by peers from CPU division, RTG (Radeon Technologies Group) announced Radeon VII, world's first 7nm consumer graphics card with 16GB HBM2 video memory. Fast forward a month and the Radeon VII is here with a price tag of $700 (~60K). 

AMD Radeon VII and its packaging, click to enlarge

Considering AMD's focus on the mainstream of late, not many anticipated it would be gunning for high-end with its next GPU. But that's exactly what Radeon VII is supposed to do - to take the fight to the likes of Nvidia GeForce RTX 2080. And that's easier said than done as Nvidia has asserted its dominance over the high-end ever since the launch of its RTX 2000 series of graphics cards. Powered by Turing architecture, these GPUs not only routinely outperform AMD's Vega and Polaris based products but also features Real-time Ray Tracing. In our review of GeForce RTX 2060, we saw Nvidia's mid-range card to nibble at the heels of RX Vega64 indicating how desperately AMD needs to update its aging line-up.

Wednesday, February 6, 2019

Itanic nears its last voyage: Intel to phase out Itanium server processors

Intel is finally done with the sad saga that is Itanium! According to reports, the microchip giant has notified partners of its intent to cease development and manufacturing of its Itanium server processors beyond current generation which is the Itanium 9700 series. Intel will accept orders till January, 2020, while the last batches would be shipped by mid 2021. So begins the rather unceremonious end of life cycle for what was once perceived as the next frontier of computing.

Thursday, January 24, 2019

Coffee Lake CPUs without processor graphics: No IGP, same price

Intel’s CES campaign this year didn’t bring much joy to the average desktop users/gamers. AMD had Ryzen 3rd gen and Radeon VII up its sleeves, Nvidia released RTX 2060; meanwhile the only “new” processor Intel came up with during the show was a locked Core i5-9400. However, the silicon giant disclosed plans for releasing some Coffee Lake refresh chips with their integrated graphics chips disabled. Those versions will come with an “F” moniker suffixed with the original model name (i7 9700KF for example). Other than the disabled graphics, everything else i.e. frequency, no. Of cores, amount of cache etc. remain identical. Have a look.

Processor model
core/thread
Base/Boost clock
IGP
Memory
Cache
TDP
Core i9-9900K
8 / 16
3.6 / 5.0
UHD 630
DDR4-2666
16MB
95W
Core i9-9900KF
8 / 16
3.6 / 5.0
No
DDR4-2666
16MB
95W
Core i7-9700K
8 / 8
3.6 / 4.9
UHD 630
DDR4-2666
12MB
95W
Core i7-9700KF
8 / 8
3.6 / 4.9
No
DDR4-2666
12MB
95W
Core i5-9600K
6 / 6
3.7 / 4.6
UHD 630
DDR4-2666
9MB
95W
Core i5-9600KF
6 / 6
3.7 / 4.6
No
DDR4-2666
9MB
95W
Core i3-8100
4 / 4
3.6/ -
UHD 630
DDR4-2400
6MB
65W
Core i3-8100F
4 / 4
3.6/ -
No
DDR4-2400
6MB
65W

Monday, January 21, 2019

Nvidia GeForce RTX 2060 Review: Features, performance, value, verdict


Nvidia’s ‘60 series has traditionally produced some of the most sought after GPUs by value seeking gamers and enthusiasts. It probably started with the now legendary GTX 460, a GPU that created a market subspace of its own – the $200(~₹15K) sweet-spot. There is an unwritten rule of short within the world of desktop graphics. It goes like this: the first couple of GPUs in any generation, the x10s and x30s to more precise (ex. GTX 710, GTX 730 – the so-called “display cards”) are a big no-no as they are cut too thin for any serious gaming. The x70s and x80s are high-end flagships that come with a hefty price tag. It’s the x60s that represent the best of both worlds – the mainstream sweet-spot where value and performance converge.

That trend however takes a hit with the introduction of Nvidia’s latest graphics card, the GeForce RTX 2060 which comes with an asking price of $350 for the standard Founders Edition (FE). RTX 2060 is the third GPU based on Turing - Nvidia’s most recent graphics architecture. The company has already released a slew of cards based on high-end Turing GPUs like RTX 2070, RTX 2080 and RTX 2080 Ti which currently dominate the enthusiast segment. RTX 2060 on the other hand is supposed to be a mainstream part which makes the pricing a bit confusing and begs the question of whether it’s performance and feature-set justify such a high price. Let’s dive in to find an answer.

Tech and spec: TU106 gets a haircut

Similar to previously launched 2000 series models, RTX 2060 is built on a TSMC 12nm FinFet process designed specifically for Nvidia. Unlike previous generations however, Nvidia’s efforts with Turing aren’t totally focused on delivering raw gaming power.

The USP of Turing based GPUs is Realtime Ray Tracing which Nvidia recognizes to be a gamechanger. To reflect the sentiment and perhaps to instill potential buyers with the same, the company even changed it’s long standing GTX naming scheme to RTX.

Graphics card
RTX 2070
GTX 1070
RTX 2060
GTX 1060
RX Vega56
RX590
RX 580
GPU/Codename
TU106/Turing
GP104/Pascal
TU106/Turing
GP106/Pascal
Vega/Vega10
Ellesmere/
Polaris 30
Ellesmere/
Polaris 20
Shader cores
2304
1920
1920
1280
3584
2304
2304
Base clock
1410 MHz
1506 MHz
1365 MHz
1506 MHz
1156 MHz
1469 MHz
1257 MHz
Boost clock
1620 MHz
1683 MHz
1680 MHz
1708 MHz
1471 MHz
1545 MHz
1340 MHz
Memory clock
1750 MHz
2002 MHz
1750 MHz
2002 MHz
800 MHz
2000 MHz
2000 MHz
Memory Bus
256-bit
256-bit
192-bit
192-bit
2048-bit
256-bit
256-bit
V-RAM
8GB GDDR6
8 GB GDDR5
6GB GDDR6
6GB GDDR5
8GB HBM2
8 GB GDDR5
8 GB GDDR5
ROPs
64
64
48
48
64
32
32
Texture Units
144
120
120
80
224
144
144
Power(TDP)
185w
150w
160w
120w
210w
175w
185w
Price
55000
35000
34000
20000
40000
28000
22000

Unlike its faster and costlier siblings though, RTX 2060 faces some serious competition most of which come from Nvidia’s own camp. In terms of price and specs, RTX 2060 sits much closer to GTX 1070 than GTX 1060 -the card it’s supposed to succeed.

TU106 - logical diagram, click to enlarge

At the heart of RTX 2060 lies the TU106 GPU, the same piece of silicon that drives the enthusiast grade RTX 2070. With 2060 however Nvidia disabled six of TU106’s original 36 Streaming Multiprocessors (SM) resulting in a total of 1920 Shader cores along with 48 ROPs,120 texture units and an192-bit wide memory interface. While the founder’s edition’s base clock is slightly lower at 1365, the boost speed is up there with the rest of Nvidia’s recent offerings.


Introduction of GDDR6 memory is one of the key highlights of 2000 series. GDDR6 is capable of 14Gbps data transfer which is much faster than GTX 1060’s 8Gbps GDDR5. On a 192-bit bus, this generates a bandwidth around 336GB/s - plenty to feed the hungry CUDA cores.

The only negative here is Nvidia’s decision to stick with 6GB V-RAM with the mainstream card. This could be point of concern as most modern games have a good appetite for frame buffers and the rule of thumb is the more the merrier. Tittles with high-res texture pack and texture mods tend to gobble video memory like nobody’s business. Also, when you pay around 35K for a graphics card you’re not just thinking about today’s tittles, are you? Going forward, 6GB could fall a little short.

And that’s not the end of story for 2060’s memory capacity. Remember how Nvidia released GTX 1060 in two variations – the 6GB and the 3GB models? Turns out, it’s at it again, only this time there shall be no less than 6 different memory configurations to deal with! Have a look -
  • NVIDIA GeForce RTX 2060 6GB GDDR6 (Founder’s Edition)
  • NVIDIA GeForce RTX 2060 6GB GDDR5
  • NVIDIA GeForce RTX 2060 4GB GDDR6
  • NVIDIA GeForce RTX 2060 4GB GDDR5
  • NVIDIA GeForce RTX 2060 3GB GDDR6
  • NVIDIA GeForce RTX 2060 3GB GDDR5
At this point we don’t know if lower memory variants will also see a reduction in other resources like CUDA cores like GTX 1060 3GB did last time around. Everything in this article is based on the 6GB GDDR6 Founder’s Edition. We’ll update this space with more informations about pricing and performance of other variants once they become available.

Features: Ray Tracing and DLSS explored

Ray Tracing is an advanced rendering process that brings a huge difference in how light is perceived within a 3d-world. Often used by movie studios to create life-like 3d imagery, it traces the path of light to create a more accurate and realistic interaction of light-rays with other 3d objects.

Reflections with and without RTX, click to enlarge

Computer or console games however use a method called Rasterization where light is mapped in a 2d plane and stored as Raster files to be used by complex shader algorithms on different shapes (polygons).

Ray Tracing on the other hand treats light like more like “real world” light resulting in more realistic reflections, diffusions and shadows. But those come at a cost of compute resource. Despite of it’s inferior visuals, Rasterization is a mature and well entrenched tech. It’s well optimized to take leverage of modern GPU’s massive parallel processing power. Where as Realtime Ray Tracing calls for dedicated hardware and as such Nvidia has dedicated a big chunk of its Turing die space to RT cores which exclusively handles tracing duties. As for Ray tracing and computing RTX 2060 gets 30 RT cores and 240 Tensor cores.

Another piece of Turing puzzle is DLSS or Deep Learning Super-Sampling. Super-Sampling is a technique that involves rendering frames in higher than native resolution and downscaling it. With DLSS Nvidia utilizes Turing’s Tensor cores to create a deep learning network, one that will sample as many frames as possible to identify the best pattern that suits a game. The process gets better overtime. The idea is to infer a final output image to the monitor which is smoother and free of aliasing artefacts. Nvidia claims DLSS should increase performance compared to conventional TAA (Temporal Anti Aliasing).

RTX 2060 also fully supports Nvidia’s variable refresh rate technology – G-sync; which makes it ideal for any G-Sync compatible monitors and even some AMD Free-Sync compatible ones too, if the reports are true.

Performance: The little card that can do

This is where every mainstream product earns its bread and butter. Thanks to Turing’s advanced and efficient graphics architecture and the fact that Nvidia didn’t trim TU106 too thin, the newest GeForce on the block performs very well. Here are key performance highlights -
  • On average RTX 2060 is faster than both RX Vega 56 and GTX 1070 while it treads blow with GTX 1070 Ti. All of these are costlier cards but expect some price drops under changed circumstances.
  • Depending on the game, RTX 2060 is 10-15% faster than GTX 1070 in FHD and QHD. The gap shrinks in 4K but that’s not a target resolution for this card.
  • On a generation to generation basis RTX 2060 is around 50% faster than GTX 1060 6GB. The later however was around 85% faster vs. GTX 960. We’d still have taken 50% increment gladly except for the price hike!
  • RTX 2060 effortlessly plays most modern games at 1920×1080 (FHD) and does surprisingly well at 2560×1440 (QHD). This makes it a good choice for those who already have 1440p monitor or considering one down the future.
  • Faster GDDR6 combined with Nvidia’s very efficient tile based color compression algorithm seems to have offset the V-RAM deficiency, for now.
  • Compared to previous gen the power consumption has gone higher thanks to a larger die and more computing hardware in form of RT and Tensor cores. But it’s still more efficient than competing AMD GPUs.
  • The FE design looks and works very well with good heat dissipation and low noise. It features the same PCB design used in RTX 2070 FE. It draws power with one 8-pin PCIE connector which is well within the grasp of a good quality 500w PSU.
  • Overclocking is capped by power limit. Reviewers are able to get ~10% on core/engine clock which is not that much. Memory overclocking seems to work better.
  • As for Ray Tracing, RTX 2060 FE is capable of ~60 FPS gameplay in Battlefield 5 with RTX set on “High/Medium” @1080p. Framerates fall below 60 in QHD.
GeForce RTX 2060 FE, click to enlarge

We have gathered and analyzed lots of benchmark data from reliable tech sites around the web. Here is how RTX 2060 performs on average @1440p:
  • Tom’s Hardware found RTX 2060 to be 15% faster vs. GTX 1070, 11% faster vs. RX Vega and 2% faster vs. GTX 1070 Ti.
  • Anandtech benched their sample to be 14-15% faster vs. GTX 1070, 11% faster vs. RX Vega, 2% faster vs. GTX 1070 Ti and a whooping 59% faster vs. GTX 1060.
  • HardwareCanucks found RTX 2060 is 12% faster vs. GTX 1070, 6% faster vs. RX Vega, 48% faster vs. GTX 1060 and slower by 2% vs. GTX 1070 Ti
  • TechPowerup found RTX 2060 to be 15% faster vs. GTX 1070, 10% faster vs. RX Vega, 40% faster vs. GTX 1060.
Value and verdict: The vanishing sweet-spot

It’s not easy to find value within a market that lacks competition. Pascal based GPUs were already more than competitive against AMD’s senescent Vega and Polaris parts. With Turing Nvidia has not only broaden the rift in performance but also pricing. An average customer might mistake RTX 2060 for a mainstream part curtsey of its x60 branding while the $350 price tag is there to remind them of that reality. The following chart will make clear of that.

Price of Nvidia 60-series cards over the years, click to enlarge
We expect to see the RTX 2060 Founders Edition at around ₹34K on various e-trailers. Street price should be lower by a couple of thousands but it may take a while.

Clearly we can’t associate RTX 2060 with the value that x60 series is known for. It performs like an updated GTX 1070 and the price reflects that. On the plus side it delivers better performance to price ratio than any other Turing based graphics card. RTX 2070 particularly looks overpriced at more than ₹50K, delivering just about 15% faster frame-rates at a 30% price premium, when compared to its cheaper sibling.

With only 30 RT-cores, RTX 2060 isn’t quite the poster child of Ray Tracing but it’s still the cheapest gateway to the scene. While very promising, Both Ray Tracing and DLSS, lack game adoption at this point. Battlefield 5 is the only game that currently supports DirectX Raytracing (DXR) and DLSS is yet to be adopted by a AAA tittle. Going forward there might come a day bearing fruition of Nvidia’s formidable developer collaboration, when these techs see a high degree of adoption and optimization but it’s not today.

Also, We know mainstream Indian gamers are more about FPS/₹ than they’re about eye-candy! And we can’t blame them. The last couple of years have been quite hard on average Indian PC gamers with a trident of GST, falling rate of Rupee and Bit-coin mining skyrocketing the price of gaming hardware. It’s only recently that things have started to revert to normal with the bursting of crypto-mining bubble and a somewhat steady Rupee.

So, with all said and done, here goes the final verdict:
  • RTX 2060 is well beyond the budget of most mainstream gamers. Those looking for a sweet-spot GPU for 1080p gaming, look elsewhere. AMD’s RX 580 or RX 570 cards represent superb value for money at around ₹18, 000 and ₹13, 000 respectively. Green team’s own GTX 1060 6GB is another viable option. Don’t buy a higher resolution monitor and you’ll be fine.
  • Those who already have a GTX 1070 or RX Vega 56, won’t gain much buying into RTX 2060. The slim performance gain isn’t enough to justify an upgrade.
  • If you already have a 1440p display or planning to get one and have a budget of ₹31K-34K, RTX 2060 fits the bill perfectly. Go ahead and buy it! If you’re upgrading from something like a GTX 970, you’ll see a huge uplift in your FPS.
  • For the deal-hunters among you, keep your eyes open for some heavy discounts on Vega 56 and outgoing GTX 1070 (Ti).
  • For the rest of you, sit tight and watch how Radeon VII pans out.