Rumor: benchmarks of upcoming Ryzen 7 CPU surface

We've got some more juicy gossip from South Korea for you today, gerbils. Hardware news and benchmark comparison site Hardware Battle posted some benchmarks of what appears to be an upcoming Ryzen 7 CPU based on the "Zen+" architecture. HWBattle's now put the info behind a password, so that link won't really get you anywhere. Fortunately, VideoCardz re-hosted the relevant information for the peanut gallery to pore over.

Image: HWBattle, via Videocardz

HWBattle posted a partially-redacted SiSoft Sandra screenshot that seems to show a CPU with a base clock speed of 3.7 GHz and a maximum clock speed of 4.35 GHz. No shipping Ryzen CPU clocks that high out of the box today. It's possible that 4.35 GHz number represents this mystery CPU's maximum XFR clock.

Image: HWBattle, via Videocardz

There's also a noteworthy memory latency comparison between the new chip and a Ryzen 7 1700X running at the same core clock and memory speed. AIDA64's cache and memory benchmark shows slightly reduced latency at every level of the memory hierarchy when compared to the the 1700X.

Finally, Hardware Battle secured results from a few synthetic benchmarks. The site's numbers from the physics test in 3DMark Fire Strike show the new chip with a solid lead on Intel's Core i7-8700K, although Ryzen CPUs were always competitive in this test.

Image: HWBattle, via Videocardz

Ryzen is also known to do well in Maxon's Cinebench, where the new chip performs a lot like a 4.35 GHz Ryzen 7. The extra clock speed seems to put it neck-and-neck with Intel's Core i5-7600K in the single-threaded test, too.

SiSoft Sandra's single-threaded arithmetic tests tell a similar story. In the Dhrystone integer math test, the new chip is the fastest of the bunch, narrowly edging out even the mighty i7-8700K. Meanwhile, HWBattle's FPU tests with Sandra's Whetstone module once again put the mystery chip in a photo finish with Intel's Core i5-7600K. It's unclear what instruction sets HWBattle chose to test with Whetstone and Dhrystone, however, as Sandra certainly supports the most recent of Intel's SIMD extensions for these benchmarks. We'd expect higher performance from Intel's various AVX2-ready cores in floating-point-heavy tests than is on display here.

AMD's refreshed Ryzen lineup should launch next month if things go according to plan. Thanks to Videocardz for the heads-up.

Comments closed
    • Unknown-Error
    • 2 years ago

    IIF this is not fake, then 2700X at 4.35 GHz (CB 178) and 8700K at 4.70 GHz (CB 206) means if were adjust the clocks and CB scores, the Intel IPC advantage is around 7pc (maybe 8pc). My, my………..things have come a loooong way.

    • Leader952
    • 2 years ago

    AMD Ryzen 2000 โ€˜Pinnacle Ridgeโ€™ CPU Official Specifications, Pricing and Performance Numbers Leak Out โ€“ Ryzen 7 2700X Leads The Pack With 4.35 GHz Boost Clock at $370 US

    [url<]https://wccftech.com/amd-ryzen-2000-desktop-cpu-official-specs-performance-prices-leaked[/url<]

    • Shobai
    • 2 years ago

    [quote<]AIDA64's cache and memory benchmark shows slightly reduced latency at every level of the memory hierarchy when compared to the the 1700X.[/quote<] Is that a reasonable call to make? L1 latency, as pictured, did not improve - was there more detailed information elsewhere? If it can be assumed that AIDA64 is reporting accurately then it seems the L2 and L3 caches have been tweaked minorly since the 1700X.

      • auxy
      • 2 years ago

      L1 latency did not appear to improve because the value is too small for AIDA to display.

        • Anonymous Coward
        • 2 years ago

        So there is evidence that AMD has shaven (making up a number here) a cycle of L1 latency, or something? Last I heard there isn’t too many cycles of latency on L1.

          • Zizy
          • 2 years ago

          Nope, no cycles shaved here, but slightly higher clocks = slightly less ns.

          There aren’t many cycles to shave and a single one would be a huge percentage jump. Intel and AMD both have 4 cycle L1 for simple access and 5 for complex for their current stuff. The least I have seen is 3 cycles for both.

            • Anonymous Coward
            • 2 years ago

            Ah, reduced latency due to higher clocks? Thats pretty well not news.

            However in particular the L2 measurements look better than clock speed alone would explain.

    • Zizy
    • 2 years ago

    Not bad, about 10% performance increase. Seems this is the 2700X, but I wonder if 2800X will even exist or will all of those dies be used for the TR?

    • ronch
    • 2 years ago

    Not very impressed by a 100MHz speed bump over last year’s top base clock. Seems to me AMD is having a bit of trouble pushing clocks. Maybe they went too aggressively with the high density libraries with Zen? If you look at die shots while sipping tea like I do since I was a todd, you will notice that Zen’s structure from 10,000 feet don’t feature many squarish structures apart from the memory structures. I noticed these ‘mushy’ structures when AMD started doing high density with Steamroller. And it’s noteworthy that Skylake and Kaby Lake still feature squarish structures throughout the core. Not sure about ARM etc. because they don’t really give us die shots.

      • willmore
      • 2 years ago

      You might be getting that backwards. When humans route logic, they tend to make them into rectangles. When computers do it, they just put the gates wherever.

      The reason that caches are rectangular is that they’re always laid out by humans–and it’s logical to make them rectangular.

        • just brew it!
        • 2 years ago

        Err… caches are rectangular because they are in fact a regular array of millions of identical logic cells, by definition. Form follows function; it has nothing to do with whether it was hand-routed or not. Laying it out any other way would be stupid.

          • DancinJack
          • 2 years ago

          This.

          • willmore
          • 2 years ago

          Again, that’s exactly what I was saying. Humans see it as a regular array of memory because that’s how our brains work. A computer would not feel compelled to implement it that way. For a computer, it’s a set of timing constrainsts that it needs to solve. That’s why computer generated logic tends to be more fluid in structure and human generated is more regular geometrically.

            • just brew it!
            • 2 years ago

            You’re thinking more in terms of the glue logic (where I generally agree with you). For something regular like a cache, the best approach is almost certainly going to be a hand-optimized canned cache block. The routing to get the signals to/from that cache block may be done by computer, but having the computer place and route each individual memory cell would be silly.

      • Anonymous Coward
      • 2 years ago

      I thought the boost speed of 4ghz to 4.35ghz was a worthy improvement. The small base clock increase likely illustrates thermal limits in high-wattage code.

      • xeridea
      • 2 years ago

      This is 2700X, it is possible there is a 2800X…

    • the
    • 2 years ago

    Nice improvements all around given that this is more of a tweak of Zen than a serious re-engineering. Higher clocks and lower latencies are all that AMD needs for a rapid follow up.

    The other improvements could come in interconnects which we won’t know of any changes (if any) until Thread Ripper and Epyc get updated later in the year. Outside of AVX heavy code, the improvements in clock, latency and hopefully interconnect could close the gap in the server world.* Things are getting very, very interesting in this market now.

    *What helps AMD here is that many of the Spectre/Meltdown fixes have a negative performance impact in the server space.

      • Srsly_Bro
      • 2 years ago

      You made a hypothetical argument then made the claim that things are “getting very, very interesting in this market now.”

      Are you basing that claim off your improvement assumptions?

      • Klimax
      • 2 years ago

      Also AMD is helped by total lack of Spectre patches they have still to release…

      • Anonymous Coward
      • 2 years ago

      I’ll be interested to see what the power cost of gaining 10% clockspeed is. Very pleased to see that AMD is (apparently) going to be able to pull of a meaningful tweak after such a short time.

    • Welch
    • 2 years ago

    Well this would be in line with what AMD said about their Ryzen+ and reduction in latency coupled with moderate boosts in clocks. I’m curious if any about the 400 series chipset will further improve on the memory speed/latency as this mystery chip is being tested in an x370.

    • ronch
    • 2 years ago

    The new AMD Future Processor is here!

    • PrincipalSkinner
    • 2 years ago

    Been waiting for this for my next upgrade. And then I decided I’m getting a laptop. Sigh…

    • JosiahBradley
    • 2 years ago

    It’s going to be so weird having an AMD CPU with an nVidia GPU…

      • PrincipalSkinner
      • 2 years ago

      An electric arc will spark between the two frying them both.

        • Sahrin
        • 2 years ago

        Or nVidia will roll out a driver update that toasts the GPU first. Could go either way, really.

      • kvndoom
      • 2 years ago

      You found a video card?!? Where? Please tell us!

        • DoomGuy64
        • 2 years ago

        Probably a used mining card from ebay, explaining the sparks and magic smoke.

          • Waco
          • 2 years ago

          Mining cards are going to be a great used card – except for fan wear, I see no reason to even worry.

            • Klimax
            • 2 years ago

            Thermal paste might be problem and I remember reading about some problems with stuck electrons in silicon causing various failures in logic gates.

      • jarder
      • 2 years ago

      Your weirdness threshold is super low. Now an Intel GPU paired with a Nvidia CPU on the other hand….

        • K-L-Waster
        • 2 years ago

        Sounds like a Tegra and a Xeon Phi.

        Yeah, weird…

      • K-L-Waster
      • 2 years ago

      You mean I was crazy to have an Athlon 64 and a GTX 8800? Why didn’t you warn me!?!

      • Rapster
      • 2 years ago

      You could get an Intel CPU with a built-in AMD GPU if you prefer. ๐Ÿ˜‰

      • Voldenuit
      • 2 years ago

      Back in the day I had an AMD Opteron CPU on an nvidia nforce motherboard and a Radeon (then independent) 9700Pro GPU.

    • Chrispy_
    • 2 years ago

    I was hoping for a 4GHz base clock, 4.4GHz boost clock for the R7 1800X and R5 1600X successors but I guess the gap to Intel is still closing.

    For all we know, this *isn’t* the top-end chip, but I guess we’ll find out soon enough.

      • Goty
      • 2 years ago

      If it’s the same chip that was has been making the rounds in the recent rumors, it seems like it might be the 2700X, or whatever they end up calling it.

        • Chrispy_
        • 2 years ago

        So a 2800X could conceivably be a 4.0GHz base and 4.5+ on boost?

          • Anonymous Coward
          • 2 years ago

          Pulling numbers out of my butt, I suspect that 4ghz base clock is too much to ask.

            • Chrispy_
            • 2 years ago

            Really? The 1800X is 200MHz faster than the 1700X, so that would make the 2800X base clock 3.9GHz, most likely.

            I figure that if they make it to 3.9GHz, they’re just going to bump the voltage very slightly to make the 4GHz mark for psychological marketing reasons. 4 GHz is better than 3.x GHz

    • jensend
    • 2 years ago

    People also said the Zen+ launch this month would be accompanied by 400-series chipsets; I’m surprised we haven’t seen any more information released or leaked about those.

      • HERETIC
      • 2 years ago

      APRIL 19th
      [url<]https://www.fudzilla.com/news/processors/45768-full-set-of-amd-ryzen-2000-pinnacle-ridge-slides-leak-online[/url<]

    • odizzido
    • 2 years ago

    I am looking forward to seeing this along with new intel testing.

    Also I am not sure how many of us there are still, but I am personally running an i5-750. I imagine there are still quite a few 2500K users as well. I would like to see how a processor like that stacks up against the newest ones so I(we?) can decide if it’s finally worth upgrading.

      • Austin
      • 2 years ago

      ๐Ÿ˜€ I’m still running 2500K; a little worried how impacted it will be by Meltdown/Spectre but to build a new system that’s only LESS compromised and impacted would cost ยฃ300 for CPU (ยฃ100), mobo (ยฃ100) and DDR4 (ยฃ100) for roughly equivalent performance, extra ยฃ50 for a superior CPU but that’s a shed load of money when it’s Intel who f’ed up in the first place.

      ๐Ÿ™ Worse I still run a lovely though aged hexa core 1366 Xeon; stock voltage o/c’ed @3.7ghz that runs triple channel DDR3-1600 (equal bandwidth to the newest mainstream dual channel DDR4-2400 setups).

      • fredsnotdead
      • 2 years ago

      i7-860 here, same question…

      • dragontamer5788
      • 2 years ago

      i7-2600K User here.

      I would have upgraded last year, but DDR4 prices were (and still are) ridiculous. I’m going to wait until 2H of 2018 before an upgrade. That’s plenty of time for these “Zen+” CPUs to come out, as well as possibly IceLake.

      > I would like to see how a processor like that stacks up against the newest ones so I(we?) can decide if it’s finally worth upgrading.

      It seems like the Ryzen Processors are single-threaded close to a 3rd maybe 4th generation Intel. But multithreaded performance is huge with 8-cores. As usual: Intel still rules single-threaded, but AMD has caught up dramatically.

      ———

      Between the DDR4 upgrade (much higher bandwidth, better parallelism, lower costs), the core-count doubling (8 core Ryzen vs 4-core Sandy Bridge), and the improved single-core performance, I’d say a Ryzen is definitely worth the upgrade over Sandy Bridge.

      Let alone Nehelem or older architectures. Coffee Lake’s 6-core looks exciting as well, but Ice Lake looks like it will mainstream AVX512. So I’d wait from the Intel side if you’re curious about SIMD performance.

        • Krogoth
        • 2 years ago

        Ryzen+ is literately on the heels of Skylake family in terms of IPC performance for non-AVX loads. Skylake family just has a higher clockspeed ceiling. This is what makes up the difference in single-threaded performance.

        Skylake family still wins in the power efficiency game for portable platforms though.

    • chuckula
    • 2 years ago

    Some people foolishly believe that Ice Lake is Intel’s panicked response to RyZen+.

    That’s not true. Because to be a panicked response, a product has to actually ship.
    Intel just has panic, NO RESPONSE!

      • Kretschmer
      • 2 years ago

      Per this article, Intel’s response was released in 2016 as the i5 7600K. ๐Ÿ˜‰

        • Waco
        • 2 years ago

        Except it’s about 1/3 the overall processing power. ๐Ÿ˜›

          • MOSFET
          • 2 years ago

          Except in the real world ๐Ÿ˜‰

          Ok, adding at least some substance, it’s just an analogy. 7600K is light, 350hp, twin-turbo, low emissions, with a 3.9sec 0-60. Maybe the Ryzen is the Toyota Tacoma pulling the Space Shuttle.

            • Waco
            • 2 years ago

            Eh, you’re making it sound like the single-threaded gap is huge. It’s not, and the extra cores make a huge difference in well-threaded apps.

          • Kretschmer
          • 2 years ago

          I was tongue-in-cheek, but it IS silly to act like these chips have no competition when TR’s own article points at a 2.5-year-old midrange chip as slugging it out with the latest top-end Ryzen for lightly threaded workloads.

          Ryzen has strengths and weaknesses. It’s a better balance than Bulldozer, but you need to utilize a lot of threads to make the platform truly perform. I heard for years how Bulldozer would be great when people used all the threads. RyZen is actually quite good (unlike Bulldozer), but it also trades per-thread dominance for a lot of cheap cores.

            • Waco
            • 2 years ago

            10% difference in lightly threaded workloads isn’t something that I call super significant. :shrug:

            • MOSFET
            • 2 years ago

            [quote<]Ryzen has strengths and weaknesses. It's a better balance than Bulldozer, but you need to utilize a lot of threads to make the platform truly perform.[/quote<] Exactly. I knew my comment wouldn't be popular, but it's the truth about my perceptions of Ryzen vs Kaby Lake vs FX. I use each of those platforms plenty for qualified commenting..

            • Waco
            • 2 years ago

            I think the point is, without benchmarking single-threaded performance, you’ll rarely know there’s a Ryzen or Kaby in a particular machine. You will, however, notice 16 threads when you start hammering the machine with big workloads.

      • tipoo
      • 2 years ago

      Maybe it was the caps at the end, but I read this like a Trump tweet, lol. Just needs a SAD or VERY BAD.

        • chuckula
        • 2 years ago

        Well he did promise MAGA: Make AMD Great Again didn’t he?

        • Anonymous Coward
        • 2 years ago

        We need more comments that sound like Trump tweets.

          • K-L-Waster
          • 2 years ago

          Clock speed wars are good and easy to win!

          (Ok, I think most of us would actually agree with the first part of that…)

            • tipoo
            • 2 years ago

            Did we not live through the Pentium 4 Heatburst? ๐Ÿ˜›

          • tipoo
          • 2 years ago

          We’re a LAUGHING STOCK folks, we make these big beautiful APUs and other countries slap single channel RAM and hard drives on them! NO GOOD. NO LONGER.

            • K-L-Waster
            • 2 years ago

            We’re going to make the best dual channel RAM, and Intel will pay for it. They’ll be happy to pay for it.

      • ronch
      • 2 years ago

      I’m amused at how you sometimes pretend to be on AMD’s side, but of course those of us who’ve been here long enough know you’re just kidding around. Or maybe your Future is Fusion?

        • Anonymous Coward
        • 2 years ago

        Inconsistency is an important part of tweet-style comments.

Pin It on Pinterest

Share This