[Techspot] AMD Ryzen 5 1600 vs. Intel Core i7-7800X: 30 Game Battle! [Links Fixed - Updated]

[formulav8]

the head of INTC Engineering group left the company today. I have a feeling he must have had a good talking to about the current designs.

Yeah a 20 yr vet leaving like that doesn't happen very often. Wonder what happened...

 

[Thunder 57]

That's an extremely poor showing for Skylake-X. In Excel the i7-7800X (a six core CPU) uses more power than the i7 6950X (a ten core CPU) and is slower. In CB15 It also uses more power than the i7 6900K (an eight core CPU) and is also slower.

Intel has destroyed efficiency by using thermal paste, adding AVX-512, using a mesh, and changing the cache hierarchy.

One would think the mesh should save power. Perhaps it will in another generation or two as the ring bus was a very mature design. I also thought the new cache layout would bring better performance. Then again, this is the first time I ever recall Intel using a non-inclusive cache and there may be a learning curve there as well.

 

[.vodka]

Looked from a high level, Intel and AMD are basically using the same cache structure now, a big L2 with a victim L3 divided in chunks. 1.375MB chunks in each core connected by the mesh, vs 8MB chunks in each CCX connected by IF.

They agreed in the past with a small L2 and a shared L3 (Phenom I / Nehalem) obviously with their own differences in caching policies, and they now agree this is the way for scaling to higher core counts. Interesting.

 

[teejee]

Yeah a 20 yr vet leaving like that doesn't happen very often. Wonder what happened...

Well Per Hammarlund left two years ago for apple. He was an Intel fellow and chief architect for Haswell among others. And he was main responsible for Hyper Threading ( P4 ).

And just before he left he was chief architect for "future cores". If you quit an assignment like that then something is wrong. Sounds like a dream task for an CPU architect.

Sent from my LG-D855 using Tapatalk

 

[Zucker2k]

Francois seems frustrated with the current decision-making at Intel.

 

[moinmoin]

I dont know if this is related to the turn of events in the CPU landscape with AMD's MCM vs Intels BIG chip but the head of INTC Engineering group left the company today. I have a feeling he must have had a good talking to about the current designs.

https://www.techpowerup.com/235385/francois-piednoel-quits-intel

Hope for Intel that this isn't indicative of engineers playing second fiddle to the marketing people.

One would think the mesh should save power.

What makes you think that? Altogether the paths will be more and longer (2-4 paths from any given point depending on location instead always 2 in the case of a ring bus) so more power consumption is to be expected.

 

[formulav8]

And just before he left he was chief architect for "future cores". If you quit an assignment like that then something is wrong. Sounds like a dream task for an CPU architect.

Francois who just quit Intel posted on linkedin and it looks like Intel (Management/Environment) is the reason.

Being a Tech-leader does not mean being a strong minded not listening feared old man, it means enabling the young engineers to be successful, and lead them by example, accept when you are wrong, and enable learning out of it, and always seek the less senior point of views. Being innovating all the time for long period of time is only possible in those conditions. And If your environment or management does not enable that kind of ways, move on!

The info was posted by someone at the TR comments page.

 

[french toast]

Remember there's also 14, 16, and 18 cores coming.

We'll see once people try to overclock these...

Arrives in time for bonfire night?
Il prep the marchmellos

 

[formulav8]

Any date on Intel having higher than 10 cores and especially the 18 core model?

 

[moinmoin]

Any date on Intel having higher than 10 cores and especially the 18 core model?

Nothing beyond the months so far afaik, 12 cores in August and the rest in October.

 

[maddie]

Yeah a 20 yr vet leaving like that doesn't happen very often. Wonder what happened...

At the most fundamental level? The Ryzen die and Intel's reaction to it.

Francois seems frustrated with the current decision-making at Intel.

Quite a few of us have believed and argued that Intel's woes are only beginning and that there is no quick and painless solution. They have pushed their product line to extracting as much margin as possible and have little to no room to maneuver except reducing prices. We all know what happens when margins fall in a stagnant to declining market.

Intel's situation
Process: advantage lost
Architecture: advantage lost (Zen lies between the ring-bus and mesh designs)
production cost: advantage lost (small dies and glue, per Intel)

I'm actually very surprised at this early resignation, unless the CEO, etc are going to cut key personnel to save on costs as a short term solution. If this is in any way correct, then Intel is committing Hara-Kiri. We'll see over the next few months who else leaves and why.

 

[moinmoin]

I'm actually very surprised at this early resignation, unless the CEO, etc are going to cut key personnel to save on costs as a short term solution.

Not sure if that's really early, the lead time for bigger changes are like two years so internally they already know what the can achieve and release in one to two years time. So something significant may have happened at Intel within the last year that's exacerbated now with the clear increased competition they have to face.

Somebody somewhere on this forum posted a French computer magazine cover from late last year that titled the downfall of Intel or some such (I wish I knew what to search for). If that was based on leaked internal info I wouldn't be surprised if there was some power struggle between management/marketing and engineers that the latter lost, and now that the former are proven to be unprepared the latter are not as eager to save the day again. In any case for AMD the timing so far appears to be excellent (kind of ironic considering they were late with their lineup themselves and still appeared to rush the release at the beginning).

 

[french toast]

^^ canard pc^^ the author of CPUZ.

 

[maddie]

Not sure if that's really early, the lead time for bigger changes are like two years so internally they already know what the can achieve and release in one to two years time. So something significant may have happened at Intel within the last year that's exacerbated now with the clear increased competition they have to face.

Somebody somewhere on this forum posted a French computer magazine cover from late last year that titled the downfall of Intel or some such (I wish I knew what to search for). If that was based on leaked internal info I wouldn't be surprised if there was some power struggle between management/marketing and engineers that the latter lost, and now that the former are proven to be unprepared the latter are not as eager to save the day again. In any case for AMD the timing so far appears to be excellent (kind of ironic considering they were late with their lineup themselves and still appeared to rush the release at the beginning).

Ryzen performance & cost appears to have been a shock to Intel. Witness the apparently rushed late Skylake-X high core count releases and that recent insane stack of marketing slides, a lot belittling AMD, one of which even quoted a Wccftech article as a source. Imagine that for a second. I'm certain no one here would have ever come even close to predicting that.

Internally, I'm sure there's a lot of turmoil with many little fiefdoms being defended.

I agree with the point of a previous malaise, as these things never develop overnight but slowly over time. The problem is that dominance with no competition can hide underlying problems. Intel is no longer close to competitive as it once was.

Some here are also so conditioned by the past few years that they can't imagine anything different. They argue that AMD is selling CPUs too cheaply without realizing that they're basing their thinking on a sample size of one, namely Intel pricing. What exactly is a reasonable price for a CPU? I don't know, but we're going to find out.

The fight is on folks.

 

[VirtualLarry]

They argue that AMD is selling CPUs too cheaply without realizing that they're basing their thinking on a sample size of one, namely Intel pricing. What exactly is a reasonable price for a CPU? I don't know, but we're going to find out.

Well, as a consumer, I haven't really liked "Intel pricing" (except at the low-end).

But it remains to be seen how AMD's "Disruptive Pricing" of their Ryzen 7 and Ryzen 5 CPU will turn out, and whether they are being sold at sufficient pricing and gross margin, to successfully fuel their R&D pipeline, in order to effectively iterate on their Zen micro-architecture design.

If AMD is pricing them sufficiently low enough, that they are effectively constricting Intel's R&D pipeline efforts, then are they constricting their own? Or is AMD such a "lean, mean, fighting machine", and Intel a "bloated pig of R&D", that Intel will suffer from AMD's low pricing, while AMD will be unaffected.

 

[maddie]

Well, as a consumer, I haven't really liked "Intel pricing" (except at the low-end).

But it remains to be seen how AMD's "Disruptive Pricing" of their Ryzen 7 and Ryzen 5 CPU will turn out, and whether they are being sold at sufficient pricing and gross margin, to successfully fuel their R&D pipeline, in order to effectively iterate on their Zen micro-architecture design.

If AMD is pricing them sufficiently low enough, that they are effectively constricting Intel's R&D pipeline efforts, then are they constricting their own? Or is AMD such a "lean, mean, fighting machine", and Intel a "bloated pig of R&D", that Intel will suffer from AMD's low pricing, while AMD will be unaffected.

All good points, but I live in a country that suffers from the Dutch Disease condition, and know 1st hand that easy money can be a double edged sword. Easy money in Intel's case being no competition as a barrier to raising prices. Witness the high cost acquisitions that were disposed of relatively quickly, the brash jumps into new market segments using cash as an enabler.

I believe that AMD is very lean and well guided.
I believe that Intel is bloated. How quick to change? I don't know

 

[moonbogg]

With the worse than expected gaming performance of Skylake-X, I think it would've made more sense to only offer 8 cores and up for X299 since coffee lake will make X299 obsolete for most people looking for a 6 core Intel solution, and I bet most people who buy Intel HEDT go for the 6 core models. AMD is doing it right here by offering 8 cores on mainstream and saving their HEDT for the big chips. I could say a lot about this right now, but I'll be happy to summarize my feelings by saying Intel's game plan got tossed in a blender and we are witnessing the resulting mess.

 

[VirtualLarry]

but I'll be happy to summarize my feelings by saying Intel's game plan got tossed in a blender and we are witnessing the resulting mess.

Nice analogy! LOL!

 

[Shmee]

Amazing show for the 1600. I almost should have waited for that instead of getting the 1700, but either one is a very good value with plenty of thread power.

 

[NTMBK]

One would think the mesh should save power. Perhaps it will in another generation or two as the ring bus was a very mature design. I also thought the new cache layout would bring better performance. Then again, this is the first time I ever recall Intel using a non-inclusive cache and there may be a learning curve there as well.

The mesh is meant to scale to very high core counts (28 cores)- it isn't the optimal solution for low core counts, but it will kick the ringbus' arse at high core counts. It's a trade off.

 

[SPBHM]

The mesh is meant to scale to very high core counts (28 cores)- it isn't the optimal solution for low core counts, but it will kick the ringbus' arse at high core counts. It's a trade off.

and with 6 core coffe lake, I think x299 for 6-8 core is going to look pointless in a few months.

 

[SAAA]

Came across an excellent review video from Hardware Unboxed featuring the brand new Intel i7 7800X versus the AMD R5 1600. Both 6-core models are equipped with identical RAM and put through the paces in 30 games with both stock and overclocked, complete with power consumption testing. Video Link -- Text Link

At stock frequency:
i7 7800X is using 22.6% more power than Ryzen 1600X

Overclocked:
i7 7800X is using 17.9% more power than Ryzen 1600X

Note: incredibly.. Ryzen 1600X is using a good bit less power than a stock clocked, quad core 7700k.

Gaming composite:

Performance is identical with both platforms overclocked. Almost spooky how they are completely equal. 7700k is about 10% ahead but if you take out just 1 game from the stack, Gears of War 4, the delta is equal to 6%. Not a bad showing for the Ryzen despite a 20% clockspeed advantage.

My biggest takeaway here would be the value you get from each as a gamer.

In fairness you could delete the AOI but OC would suffer according to most reviews on Skylake-X, and you could use less RAM but as the author states, it would be odd to get such a platform and use 4x4gb to maintain quad channel memory, or conversely equivalent RAM but only dual channel operation. A closer comparison for some may be more like $514 vs $800 if you use a cheap air cooler and less RAM. That ~$300 difference may very well mean the difference between a 1070 and a 1080Ti.

Gaming Value:
Ryzen 1600: 0.245 FPS/$
Core 7800X: 0.1575 FPS/$

note: I used the more Intel friendly $800 figure with air cooler and dual channel memory.

Bottom line: AMD Ryzen 1600 a whopping 55% more cost effective for gaming compared to the Intel i7 7800X all the while being 20% more power efficient.

Honestly, cut the crap.

All I see in these particular numbers is game performance not gaining a thing trough oc, just look at stock vs oc for every cpu and tell me how 2-5% difference matters when the % of OC should give much more. Maybe the limit is more on the gpu/game engines? You know the same reason a quad still beats both OC six cores, at stock too. Though productivity benches may paint another picture there, or percentile frames or whatever other measure that's missing.

Then as for efficiency, don't say 20% because at stock that 7800X is definitely running all core 4GHz and it's within a few watts of Ryzen when OC to 4GHz too. Surprise it's less efficient at 4.7GHz, or 17.5% higher clocks (that 1600X will never ever reach under any practical cooling); yeah games don't show that performance increase at all, but any other program should.

Twice the cost for same game performance (as of today when clearly they both aren't used well enough to be beaten by quad cores) but possibly 20% better productivity.
Intel's price/performance has never been good and I admit it sucks here too, but they do have the edge and premium for it.

 

[MarkPost]

well, as far as I know there is no a game using all cores/threads, so stock doesnt mean they are running at its base or all core clocks, but higher than that (due to turbo feature). For example, is very possible 7800X is running game benchs at 4.3-4.5, or 1600X at 3.7-3.8 etc. This is a very likely reason to not seeing a great difference with fixed OCs at 4.7 or 4.0

About productivity, you have 1700 or 1700X. They are cheaper than 7800X too (not to mention the cpu-mobo combo), with a bit better overall perf, stock or oced

 

[Insert_Nickname]

With the worse than expected gaming performance of Skylake-X, I think it would've made more sense to only offer 8 cores and up for X299 since coffee lake will make X299 obsolete for most people looking for a 6 core Intel solution, and I bet most people who buy Intel HEDT go for the 6 core models.

Not only that, but since CFL apparently uses identical cores (without AVX512, "old" cache structure) to KBL, we may even end up with CFL chips outperforming the 6 core SKL-Xs. That would be... embarrassing... for Intel.

 

[CHADBOGA]

I dont know if this is related to the turn of events in the CPU landscape with AMD's MCM vs Intels BIG chip but the head of INTC Engineering group left the company today. I have a feeling he must have had a good talking to about the current designs.

https://www.techpowerup.com/235385/francois-piednoel-quits-intel

He has been spending way too much time on twitter pushing SJW narratives and anti-Trump rhetoric, instead of concentrating on his job.

Little wonder Intel has failed to drive home what should be clear advantages.

His removal can only be a good thing for Intel, hopefully Krasnich follows him out the door.