Ryzen 1800X @ 3.6GHz With Turbo Disabled – Outperforms Intel’s 8-Cores In 6/8 Tests CPU Benchmarks

IdiotInCharge said:
In the desktop space, agreed. AMD's new products, just by AMD's sheer need to survive by grabbing marketshare to generate and sustain revenue, will force Intel to adjust somewhat.

On the other hand, I can't really complain; outside of gaming, on the desktop, there's been very little need for faster parts, and Intel hasn't let their pricing go wild either.
Click to expand...


Yeah I agree too, but its nice to see Intel get pushed :)

I'm excited now to see what they have in 3 years, coffee lake doesn't look any different than what they have now.
 
One thing a competitive AMD may do is force Intel to push the features down a rung; imagine getting S115x CPUs with six cores and HT as an i7 with the fastest clocks of the generation in the same US$300-350 bracket that the i7 7700K currently occupies?

I'd shake Lisa Su's hand for that one, even if AMD doesn't have a product that can compete on pure performance!
 
os2wiz said:
Snip of the fanboi crap.

I am extreemely happy that AMD has developed Ryzen and aggressively priced it so that 6 and 8 core chips with high performance are affordable for so many more people.
Click to expand...

The last sentence I agree with. I have no problem with AMD pushing the market. I hope they get back to their Athlon days and kick Intel in the ass. We all win if that happens.
 
IdiotInCharge said:
One thing a competitive AMD may do is force Intel to push the features down a rung; imagine getting S115x CPUs with six cores and HT as an i7 with the fastest clocks of the generation in the same US$300-350 bracket that the i7 7700K currently occupies?

I'd shake Lisa Su's hand for that one, even if AMD doesn't have a product that can compete on pure performance!
Click to expand...

That product was coming anyway, Ryzen or not. And its called Coffee Lake. But it needs to be done on 14nm++ to make it worthwhile clockwise. You wouldn't want to go backwards in ST performance.
 
razor1 said:
Looking at Kaby Lake, it is well optimized for lower end sku's where we can see the IPC change, but for their higher end chips, those IPC games are held back by some kind of bottleneck.
Click to expand...

Look at the relative die area for the CPU cores vs the GPU's on the newest chips. the GPU is more than half of the chip. Intel has been focusing these chips on lower power envelopes and improving graphics for laptops. I'm not a chip engineer, but I'm willing to guess that unless transistors get budgeted back toward the CPU cores we won't see big jumps. Optimization can only take you so far. The power reduction is also pretty useful in server environments (fewer watts = more cores) so has been no development pressure to do anything other than what they've been doing.
 
Activate: AMD said:
Look at the relative die area for the CPU cores vs the GPU's on the newest chips. the GPU is more than half of the chip. Intel has been focusing these chips on lower power envelopes and improving graphics for laptops. I'm not a chip engineer, but I'm willing to guess that unless transistors get budgeted back toward the CPU cores we won't see big jumps. Optimization can only take you so far. The power reduction is also pretty useful in server environments (fewer watts = more cores) so has been no development pressure to do anything other than what they've been doing.
Click to expand...

Throwing more transistors at the problem can only get you so far. The big jumps of the past was facilitated by the movement from transistors doing multiple jobs to transistors specialized to one job. Essentially, going from IPCs of less than 1 to 1. IPCs above 1 require branch prediction, I believe Intel is somewhere around 9 on average for the latest architectures. That is, Intel has engineered an architecture that is able to predict the next 9 results on average, which is quite frankly amazing. To see a jump from 0.3 IPC to 1 today would be like going from 9 to 27, or seeing 27 steps ahead. That is several magnitudes more difficult than just throwing transistors at the problem, which is what prompted multiple cores in the first place. The low hanging fruits of CPU design have already been picked, further progress in x86 silicon-based CPU performance would be very difficult to obtain.
 
rgMekanic said:
THen you lose the modern amenities like I said, DDR4, M.2, USB 3.1/Thunderbolt, etc. etc.
Click to expand...
2666v3 works on X99 motherboards just fine. Only issue would be NVM.e boot drive, last time i checked, and even that is only an issue with overclocking last time i checked.
 
lolfail9001 said:
Ebay 2 2666v3 qual samples, what is the issue :p? Intel gets nothing, you get 20 cores at decent frequency.
Click to expand...


You have to be careful with Intel ES on ebay. The only listing I see now that has working, (not dead, for parts), E5-2666 V3 cpus has some oddballs that aren't exactly the standard E5 2666 V3. It's 10 core, 10 thread, (no HT), and they are 2.9 Ghz with turbo to 3.1 Ghz. It's also $450. So clock speed is low, no HT, and it's more expensive than a 1700X. Oh, and it's Haswell-E, and Ryzen seems to be around Broadwell-e IPC, so this ES loses on pretty much all fronts.
eBay item number: 122361254452

The good E5 QS and ES now are expensive and go fast when available.

edit: I figured that listing above out. The guy has the listing wrong, he is selling E4-2663 V3, not 2666.
So that's another issue with ebay, people list things wrong, and sometimes it's easy with ES or QS cpus.
http://www.cpu-world.com/CPUs/Xeon/Intel-Xeon E5-2663 v3.html
 
Last edited:
pgaster said:
eBay item number: 122361254452
Click to expand...
Of course, it is first stepping ES.
pgaster said:
Oh, and it's Haswell-E, and Ryzen seems to be around Broadwell-e IPC, so this ES loses on pretty much all fronts.
Click to expand...
Well, outside of more cores. And of course there is always an exploit of 2683+Haswell clock bug, that suddenly turns it into respectable CPU.

That is getting way OT though.
 
You must log in or register to reply here.
Back
Top