AMD Responds to Intel's 9th Gen Benchmarks

Page 2 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Oct 22, 2018
1
0
10
Intel is playing games because they are falling behind and out of favour. I mean releasing false benchmarks and removing hyper threading from their lower tier chips are blatant moves.
 
There is a new video up on Hardware Unboxed, one of the sites that first broke the news of Intel's bogus paid for benchmark reviews, that shows why some reviewers had low wattage low temp processors and some had very high wattage high temperatures. It seems that Intel's new "most powerful gaming processor ever" will also cost you even more than previously thought. On motherboards that are rated for 95 watts and have inadequate VRM cooling the new i9 9900K throttles itself. The reason some were seeing low wattage and low temps is because their test rigs were either 95 watt or didn't have enough VRM cooling and the processors throttled to 4.2ghz all core instead of the 4.7ghz all core that most reviewers were getting with higher end boards. Only expensive boards like the Taichi Ultimate and Godlike motherboards are capable of handling the wattage and heat that the i9 9900k actually needs to run at true spec (4.7Ghz). In the case of the Godlike motherboard that will add an additional $600.00 US to your build cost, not to mention these space heaters also come with no out of the box factory heatsinks so your looking at the very least an additional $100 for a quality high end cooling solution. That means that your i9 9900k build for just processor, motherboard and cooler could easily run you $1300 US. If you don't spend the money on a quality motherboard with at least 150 watt capacity and high end VRM cooling then its not even going to run stock properly, has no capability of overclocking and at 4.2ghz all core is only very marginally "better" than the R7 2700X which is under $300.

I just don't see where Intel's 9th Generation falls into any useful category. Its not the best gaming processor. The best gaming processor today is the 8700k. It has 6 cores, 12 threads can overclock to 5ghz and not many games out there can make use of more cores than it has. It can also be gotten for over $200 cheaper than the i9 9900k. It also can't claim to be the best workstation as there are Threadripper processors in the same cost bracket that are better workstations. For streaming and game play it also makes no sense at its cost bracket seeing how well the R7 2700X does.

In short Intel needed the i9 9900k to totally crush the R7 2700X, which is why they paid for flawed benchmarks to make it seem that it actually did. They paid for results showing it 50% better in games which was total B.S. They needed that though because their "wonder processor" is deeply flawed for power consumption, heat generation and cost. For all the extra money for the processor (nearly double the cost of the R7 2700X) heat sink (at least $100), and high end motherboard you only get on average 12 -16% better performance in games which is usually something you cant even notice while playing and you pay twice price for it.
 


Considering that people have been using cheaper boards than the God Like or Taichi with the 8700K overclocked, and yes it easily pulls more than 95W, I would say this is probably not 100% accurate. I can see a super cheap board causing issues, yes. There is a reason the board is cheaper, it typically has cheaper components.

However I bet there are mid end boards that could easily power it without having throttling issues due to VRM heat and power. Of course it would be up to the reviewer who are claiming this to present the data. I would be interested to see this.
 


It's hard to say, the 9th gen processors haven't been out there long enough to get all the answers yet. The problem with saying the 8700k was ok with the mid range boards so the 9th gen will be as well is two additional cores and four additional threads pushing more watts and more heat. We really don't know if boards capable of handling the 8700k are capable of handling the 9900k. The 9900k is the highest clocked 8 core 16 thread processor ever and it has severe limitations due largely to its 14nm process. I have a feeling that boards that use to be made fun of for being "overkill" on VRM cooling are the boards that are going to be needed to get the most out of the i9 9900k. Even then it will require a quality PSU, heat sink solution, and plenty of case fans (something that is sometimes overlooked). If the motherboard isn't adequate for this processor then its going to throttle and go down to 4.2ghz all core, at which point its on par clock speed with an overclocked R7 2700X and is only very marginally better. For its cost that makes no sense.

The fact is after its release it makes no sense as a workstation vs Threadripper, it makes no sense as a gaming processor vs the 8700k (especially when cost and is taken into account) and as a streaming gaming rig it doesn't outperform the R7 2700X by enough to justify its nearly $580 price tag with no cooling solution which will run you at least another $100.
 


AMD is focused on IPC and improving on its arch, much more than Intel has been for years. First gen Ryzen over 50% better IPC than Excavator, second gen Ryzen (Ryzen+) 3% better IPC than first gen Ryzen, and next year Ryzen 2 on 7nm will be releasing with a claimed 13% gain over Ryzen+. I'd say that AMD is very focused on improving Ryzen. In fact right now the IPC difference between Ryzen and Intel is minor at best (somewhere around 3%). Where Intel enjoys its performance advantage now is mostly due to its very mature arch and high clock speeds. AMD isn't perfect, but they are doing alright and I hope they keep up the current trend. Its nice having real competition in the high end CPU market again.
 


+1 The FX 9590 wasn't a bad processor, but should have never had a $900 price tag. I don't know what the %$# AMD was thinking with that one. Pretty sure it has to be the #1 most overpriced processor in history. Really the FX 8350 and 8370s of that era were better all around processors. In the case of the FX 8370 I could usually get one to the same speed as the FX 9590 but drawing lower Vcore and creating less heat.
 


To be fair better IPC than Excavator was not hard. It was a poor design choice that resulted in lower IPC and performance. Ryzen was pretty much AMD catching up to Intel.

My general thoughts are that both Intel and AMD have hit a wall IPC wise and short of a miracle of a new uArch neither will see gains like Excavator to Ryzen or Pentium 4 to Conroe. Or even Westmere to Sandy Bridge TBH.

As for the power, yes two more cores add power draw but if its the power deliver system then a mid range Z390 board will probably do fine to power it for stock and most overclock settings. I have looked at a few mid range Z390s (180-250 range) and they all have very decent heatsinks for cooling the VRMs.

I personally don't care though as I have no interest in a new CPU and board right now. Everything is priced higher than I want it to (especially memory) and I don't see the value in either setup right now. I am personally trying to wait till NVDIMMs become a common factor or if they do.
 
To be really fair, I was part of the mega thread dealing with Ryzen, then called Zen, discussion and every expert on that thread (including myself) said AMD had 0% chance of catching up to Intel in one generation. Hopefuls were praying that Zen first generation could match Haswell level performance (IPC). Instead AMD came out and shocked the entire computer world with Ryzen taking it straight to Intel's top of the like Skylake processors. Better IPC than Excavator wasn't hard, but 50% better IPC and going toe to toe with the best Intel could offer in Ryzen's first generation was huge. I'm just saying that was really impressive.

I tend to agree though that there won't be another massive jump in IPC for either AMD or Intel, at least not for awhile. If Ryzen can actually pull off a 13% IPC improvement moving from 12nm to 7nm that would be a about as large as a jump as I can see either side making once Intel moves to 10nm.
 
Oct 25, 2018
1
0
10
I feel like AMD was within their right however this is common practice in literally all competitive market, however if i had to place a bet on who will win the race for the next 10 years i'd have to go with Intel, simply because AMD has only recently regained some notoriety due to huge marketing investment a very competitive pricing that cuts down on the profit margin and a crypto craze that probably benefited them more than it did to Intel just because of pricing but the crypto craze is coming to a end and Intel still has the lead while they didn't even spend one third of what AMD did on marketing in 2017 compared to Intel who's biggest chunk of the cash flow goes toward R&D. AMD shows signs of slowdown and it just means Intel are close to giving the last blow. AMD isn't going to disappear but it's going to go back dormant just like it did 15 years ago.
 
While I agree that Intel's R&D budget is much greater than AMD's that fact really hasn't helped them much as of late. They had to try slight of hand to showcase a 5Ghz 28 core processor that in reality could never reach 5Ghz 24/7 and needed major components and a large chiller to accomplish even a demo. Then they had to hire a biased company to do paid for benchmarks to hail their i9 9900K. Said company only came clean and reran the benchmarks with the AMD r7 2700X running all its cores and hopefully with full RAM speed which showed the initial results, i9 9900K being 50% better performance in games, was in reality only 12% overall faster while being vastly more expensive. Every "new" release for Intel is on 14nm or 14nm+ or ++ or maybe now +++. Intel is sticking with iCore and I really believe that they have squeezed just about everything they can out of that arch which is why they are having so many issues trying to transition to 10nm. In short Intel's vast budget for R&D really hasn't been an asset for them since Ryzen launched.

AMD on the other hand has a new arch that is already showing huge promise and has a lot of room for refinement. Compared to Intel AMD has gone from 14nm to 12nm and will be on 7nm next year all while Intel has been releasing +'s on their aged 14nm process. This advantage is going to lead to IPC and efficiency gains for AMD while Intel is going to have to answer by pushing 14nm harder and further and its got to be at near limit for what they can do with it.

Sometimes its not about overall budget, sometimes its about just having a newer, easier to work with arch. AMD has only just started to showcase Ryzen's potential, while Intel has reached a performance wall with iCore (the performance wall is further die shrinks).
 


I doubt AMD will get much more out of Ryzen. The performance wall has been hit for both.

As far as I have heard Intel does have a new uArch working in the back but until its close to launch we probably won't see much of it..

TBH what we really need as a new 64bit only uArch and ditch the x86 logic. Intel tried to do this but of course people are resistant to change.

Honestly I cannot even think of the last time I purposefully installed a 32bit version of Windows.
 


Totally agree with you on the overall need to ditch x86. I think AMD will be able to get a little more out of Ryzen with 7nm (early engineering samples have been leaked with ~13% IPC gain), but new 64bit uArch is what both companies are hopefully working towards. I've thought for a long time that the big delay with Intel's 10nm was iCore has hit a wall and Intel are working on an all new uArch. It woulnd't surprise me at all if Intel really is scrapping 10nm and instead is focusing on a new uArch. It also wouldn't surprise me if AMD has been using some of the extra money from large market share increases to do the same.