News Hyper-Threading Coming to Intel's Comet Lake Core i5 and i3 Lineup: Benchmark Submission (Upd)

InvalidError

Titan
Moderator
Also I don't know what ultra-competitive is supposed to mean but the third refinement of ryzen and intel can still counter them with ancient skylake tech...
Another problem with calling the market "competitive" is that manufacturers have no incentive to actually compete against each other when they're selling everything about as fast as they can manufacture it regardless of price. Lenovo blames shortages of Intel chips for its sales being 3-4% lower than they should have been.
 
AMD has delayed the 16-core 32-thread Ryzen 9 3900X until November, so company has a smaller window of opportunity to roam entirely uncontested as the performance leader on the mainstream desktop.
I think you mean the 3950X? The 3900X is the 12-core, 24-thread chip that has been out for a few months already.

Also, as far as the leaks I've seen go, Intel's upcoming i9s for their mainstream platform will apparently only be going up to 10-cores, so while that processor might compete with AMD's existing 12-core 3900X in terms of multithreaded performance, the 16-core 3950X will likely remain well ahead at the heavily multithreaded tasks these processors are most suited for. Intel can push clock rates higher, but I can't see that making up for the 3950X having 60% more cores.

Also I don't know what ultra-competitive is supposed to mean but the third refinement of ryzen and intel can still counter them with ancient skylake tech
"Ultra-competitive" in terms of pricing for a given level of performance. Amd's $200 Ryzen 3600 performs similar to last years $300+ i7s, and isn't that far behind the current models, and their $330 3700X isn't far behind the similarly more expensive i9s. And since they're more efficient, you don't need a $100 cooling solution to keep their temperatures in check. And at the kind of heavily-multithreaded tasks you would want a high-core count processor for, the 3900X, and soon the 3950X are completely uncontested in terms of "mainstream" desktop performance.

Sure, Intel can compete with their lineup, but without AMD being "ultra-competitive" they would have little reason to. Without that kind of competition, I really don't think Intel would be turning what had been their premium i7 processors into budget i3s in the span of a little more than a couple years.
 
  • Like
Reactions: TJ Hooker

InvalidError

Titan
Moderator
Sure, Intel can compete with their lineup, but without AMD being "ultra-competitive" they would have little reason to.
If the market was truly "ultra-competitive", AMD and Intel would be rushing cores and features down the product stack as they used to 20 years ago. They aren't competitive at all, both sides are merely delivering the minimum incremental gains at the highest prices they think they can get away with while demand is outstripping supply.
 
If the market was truly "ultra-competitive", AMD and Intel would be rushing cores and features down the product stack as they used to 20 years ago.
I would still say the pricing is "ultra-competitive" relative to the competition, and definitely relative to the market a few years back. And "rushing cores" is kind of what's happening.

Compared to Intel's Kaby Lake chips that were the current generation until just 2 years ago, Intel will have effectively doubled core and thread counts across their lineup with their next generation of processors. I would hardly say that doubling the amount of processor you get for your money over the course of a few years amounts to "merely delivering minimum incremental gains."

And AMD has nearly doubled per-core performance over their pre-Zen architecture, and added SMT for even more performance at heavily multithreaded tasks, all while significantly improving efficiency.

When AMD remained stuck on their Bulldozer-derived architectures for way too long, things became uncompetitive, so Intel had little reason to substantially improve their processors from one year to the next. The current processor market is far more competitive than what it had been for a number of years.

As for wanting more, there's only so much that's practical for these companies to do. 7nm production is limited, so it's not like AMD can churn out a massive supply of processors and sell them for substantially less than they are. And if they don't turn enough of a profit on them, they might lack the funds to stay competitive with Intel in the years to come. It also sounds like Intel's 10nm architecture didn't turn out quite as planned, making it a less-than-ideal successor to their existing desktop processors. Unlike 20 years ago, it's a lot harder and far more expensive to move to smaller manufacturing nodes, so these companies can't just rely on the easy performance gains that came from being able to ramp up a processor's frequency each year. And as far as increasing core counts goes, most current software doesn't even fully utilize the cores available at the mid-range now, so additional cores are arguably only needed for niche use-cases at this time.
 
  • Like
Reactions: TJ Hooker

InvalidError

Titan
Moderator
I would still say the pricing is "ultra-competitive" relative to the competition, and definitely relative to the market a few years back. And "rushing cores" is kind of what's happening.
If the market was truly "ultra-competitive", AMD wouldn't have stuck with 6c12t at $200 for three years, it would have pushed forward. Since Intel got held up by 10nm delays, AMD had no reason to rush and it is now AMD's turn to wait and see Intel's response. A lack of response is not competition.

Back in the days of ultra-competitive market, AMD and Intel were leapfrogging each others on features and price every year, even twice a year and the cost of mainstream parts was coming down twice a year. Now, it takes years for either one to respond to the other and it is only the minimum response required to maintain relative relevance at a given price point.

If the market was truly ultra-competitive, gross profit margins would be down to 15% instead of 40-60% and an 10700/3700-class CPU would cost $150-200 instead of $300+.
 
If the market was truly "ultra-competitive", AMD wouldn't have stuck with 6c12t at $200 for three years, it would have pushed forward.
This comes back to my last point. What exactly does the average user currently "need" more than 6-cores with 12-threads for? The vast majority of people are not going to see any noticeable difference in performance in most common software from having more cores and threads than that, and even that's probably a bit more than what's needed by most. 8-core, 16-thread processors are currently kind of a niche product, as most software doesn't scale well to numerous threads, and that definitely applies to core counts beyond that. Maybe we'll see some good use-cases crop up in the coming years now that higher core counts are becoming more common, but at the moment the benefits are a bit questionable.

It sounds like you're just disappointed that the unrealistic specs that were rumored for the 3000-series late last year didn't turn out to be true. Honestly, they seemed rather unlikely to me from the start. If AMD had made their latest 6-core, 12-thread processors available for $100-$130, what exactly would most people have to gain by buying anything more? Again, AMD needs to be profitable to remain competitive, and if they are making very little on each processor sold, combined with the limited production on the 7nm process, they're going to have a hard time staying competitive, and could end up stuck in another Bulldozer era. Research and development isn't free, and that's something that should arguably be figured into a processor's cost.

Also, it's worth pointing out that we kind of actually got much of that rumored hardware at those price points, in the sense that first and second-generation Ryzen processors are now available at those prices. You can get a Ryzen 2600 for about $120 now, or a Ryzen 1600 for about $100 from some retailers. The 1600 was a $220 part when it launched just two-and-a-half years ago, and now it costs less than half of what it did then, and is actually priced lower than what the 4-core, 4-thread Ryzen 1200 originally launched for. Again, that's approximately a doubling of multithreaded performance at those price levels compared to what AMD made available when they launched first-gen Ryzen. Likewise, you can get an 8-core, 16-thread Ryzen 2700 for $170, or a 2700X for $195, if you are willing to trade the better IPC of a Ryzen 3600 for more cores.

Or on the Intel side, an i5-9400F sells for around $150, while offering better performance than their i7-7700 did for around $300 just two years ago. At that time, you would be looking at dual-core i3s with less than half the multithreaded performance and lower performance per-core within that price range. So yeah, I would say competition has been significantly improving performance at a given price level, at least as far as multithreaded performance is concerned. Maybe you don't see massive gains in terms of lightly-threaded performance, but again, a lot of that comes down to being limited to what the current process nodes are capable of.
 

InvalidError

Titan
Moderator
So yeah, I would say competition has been significantly improving performance at a given price level, at least as far as multithreaded performance is concerned.
Better than before, sure. Ultra-competitive, I still say no. Ultra-competitive is when there is on-going pressure all the way to current-gen parts, not only on older-gen and non-competitive parts. Right now, the "competition" amounts to nothing more than forcing Intel to sanity-check its pricing a bit.

As for limiting the number of available cores at any given price point to "what people need", there was practically no software capable of leveraging multi-cores back when the first mainstream multi-socket, multi-threaded and multi-core systems hit the market 15-25 years ago. Meaningfully threaded software didn't become common in the mainstream until about ten years ago. Software developers aren't going to waste effort on writing and optimizing software for hardware almost nobody can reasonably be expected to have at the time. If you want more massively threaded software, you have to put the CPUs out there then wait 5+ years for enough people to have 'em before developers will bother, hence multi-threaded software and games not really catching on until CPUs with 4+ threads became a well-established baseline.
 
And at the kind of heavily-multithreaded tasks you would want a high-core count processor for, the 3900X, and soon the 3950X are completely uncontested in terms of "mainstream" desktop performance.
And as far as increasing core counts goes, most current software doesn't even fully utilize the cores available at the mid-range now, so additional cores are arguably only needed for niche use-cases at this time.
You basically defeated yourself here.
AMD is competitive in niche use-cases.

At every price point ryzen is great for these niche use-cases but loses to intel's CPUs in gaming and other things that mainstream users would actually care about.
 
You basically defeated yourself here.
AMD is competitive in niche use-cases.

At every price point ryzen is great for these niche use-cases but loses to intel's CPUs in gaming and other things that mainstream users would actually care about.
Nope, you seem to be misinterpreting things here and selectively cutting out the most relevant part of that paragraph...

"Ultra-competitive" in terms of pricing for a given level of performance. Amd's $200 Ryzen 3600 performs similar to last years $300+ i7s, and isn't that far behind the current models, and their $330 3700X isn't far behind the similarly more expensive i9s. And since they're more efficient, you don't need a $100 cooling solution to keep their temperatures in check. And at the kind of heavily-multithreaded tasks you would want a high-core count processor for, the 3900X, and soon the 3950X are completely uncontested in terms of "mainstream" desktop performance.
The point is that for $200, Ryzen gets you close performance to the current i7s, and enough threads to stay relevant for a while, most likely longer than the current i5s, while offering similar performance today. Their increased thread-counts are absolutely relevant as far as gaming processors around that price level are concerned. And the 3700X's additional cores become relevant for those streaming or otherwise multitasking while gaming, while Intel currently restricts SMT to their i9s.

The vast majority of people building gaming systems are likely to be mostly graphics limited either way, meaning that money would typically be much better put toward graphics hardware rather than the CPU, aside from perhaps very high-end builds running 2080s and 2080 Tis, but that's another one of those "niche" audiences not representative of most systems. Unless you have a card around that level and want to eke out a little more performance, putting that money more toward a better graphics card is probably the better option for gaming.

As for my suggestion about "mid-range" core counts not getting fully utilized, I'm talking about Ryzen's mid-range core counts, as in 6-cores with 12-threads, and presumably Intel's next generation of mid-range processors that may finally offer the same. The 3900X and 3950X are absolutely niche products, which is why I put "mainstream" in quotes, but again the point comes back to price, where even within this niche, they are offering the better option around their price points. And due to them and the upcoming Threadrippers being "ultra-competitive", we see Intel slashing prices of their HEDT processors in half next generation. And its a similar case across the lineup, where we are likely to see current i9 performance at i7 prices, current i7 performance at i5 prices, and current i5 performance at i3 prices. Intel wouldn't be doing that unless they felt the competition was being "ultra-competitive".
 
  • Like
Reactions: TJ Hooker

InvalidError

Titan
Moderator
And due to them and the upcoming Threadrippers being "ultra-competitive", we see Intel slashing prices of their HEDT processors in half next generation.
I wouldn't call a one-time price drop to partially correct a decade of stagnation being ultra-competitive, competition is supposed to be an on-going condition. I'll agree to call the market competitive when we get significant price cuts on current-gen parts at least once before next-gen parts launches becomes imminent. Once new parts are imminent, price cuts on old parts are traditionally more about clearing the channel for next-gen parts than competition.