What I found about 5775c EDRAM's impact on gaming performance.

[cucker tarlson]

I didn't know 5775c was so unpopular, look at the size of the list for devil's canyon ownership club (2 CPUs: 4790K and 4690K) and compare to the list of 5775c/5675c owners

http://www.overclock.net/t/1490324/the-intel-devils-canyon-owners-club
http://www.overclock.net/t/1583537/intel-broadwell-c-ownership-club

Crazy. Polish enthusiast forums are pretty much barren but still there are more 5675c/5775c owners there than here. Ppl look at 5775c in my sig sig and go like whut bro, but 7700K got moar megahurtzzz

Intel’s Skylake lineup is robbing us of the performance king we deserve. The one Skylake processor I want is the one that Intel isn't selling.

wow that 6785R sounds so dope yet so obscure I have never even heard of it. Wonder if it's possible to find an OEM/barebone desktop featuring this beast anywhere.


All that remotely resembles a desktop PC and comes with Skylake and 128MB edram is this NUC

http://allegro.pl/zestaw-nuc-boxnuc6i7kyk-i7-6770hq-8gb-ssd-1tb-m-2-i6933251324.html

It has a 6770HQ
https://ark.intel.com/pl/products/93341/Intel-Core-i7-6770HQ-Processor-6M-Cache-up-to-3_50-GHz

but it's still BGA

The NUC looks cracking and it does feature a usb type c/thunderbolt 3 but it's still gonna be limited to 5GB/s which will severely cripple any modern high-end or even mid-range card.
That NUC with alienware GPU dock would look so sick hooked up to my TV...

seems like up to 30% performance hit to an oc'd 980Ti, but I suppose a GTX 1050Ti or even a 1060 would be a good choice to go with the external GPU docking station.

 

[Fouquin]

I never understood why they didn't include this in later chips. It certainly made a large performance improvement. Wonder if it hurt profit margins.

The L4 cache was too expensive to manufacture and made the CPU too good

They did include it though. Skylake still has eDRAM but it runs off the system agent and is transparent to the system as DRAM. It no longer acts as an L4 off LLC but it does still sit in front of DRAM access if any device needs it.

 

[cucker tarlson]

They did include it though. Skylake still has eDRAM but it runs off the system agent and is transparent to the system as DRAM. It no longer acts as an L4 off LLC but it does still sit in front of DRAM access if any device needs it.

here's the 5775c one for comparison

Spoiler: pic

 

[Flaky]

eDRAM is still present on SKUs where it is most beneficial - laptops and compact platforms, when there is no overclocking headroom and means to use RAM with higher clocks.
On current socketed desktop platforms, this is not a problem. The faster main memory you get, the smaller benefit from eDRAM is, up to a certain point when main memory will be faster and have lower latency than eDRAM.

 

[cucker tarlson]

eDRAM is still present on SKUs where it is most beneficial - laptops and compact platforms, when there is no overclocking headroom and means to use RAM with higher clocks.
On current socketed desktop platforms, this is not a problem. The faster main memory you get, the smaller benefit from eDRAM is, up to a certain point when main memory will be faster and have lower latency than eDRAM.

well if you prefer to spend $100 more on memory than $30 on a CPU then yes.

Lower latency than edram ? I don't think DDR4 can do that. Niether 7700K nor 5930K running DDR4 3600 CL14 and some ridiculous (~4500MHz) northbridge speeds can go below 40ns like 5775c can do at stock.

Faster memory read/write ? Yes, tuned 3600 DDR4 will beat stock edram. 5775c @4.4GHz and EDRAM @2GHz with some decent 2133 CL9 DDR3 should still beat a 5GHz 7700K with DDR4 3600 by a whisker or at least be as fast at lower cost and power draw.

 

[Vayra86]

With the current DDR4 prices I might actually just try to find one of these and put some nice DDR3 next to it for my next build. I've always known the 5775c was special, but seeing it today, especially with regards to latency and min fps, quite an eye opener. THANKS for that man! Topping the charts over a 5 Ghz 7700k while pushing only 4.2 is just.. wow

Not a huge fan of the i7 on steroids that is the 7700k. Feels like a weaksauce AMD attempt like the 9590 to get big numbers.

I have a question too (actually seriously researching this now ): overclocking seems to cap out at the 4.0-4.4 Ghz range. Is there anyone who can provide input as to how lucky you'd have to be to hit 4.2 ~4.4 ? I saw OP at just around 1.375 or so, that's quite a lot... Are these reliable 4.2 Ghz chips, or is that optimistic

@cucker tarlson what are your temps for that 4.4 @ 1.38?

 

[cucker tarlson]

#1 I like about 5775c is how fun it is to play with overclocking it. I think it surpasses my old 2500K and I never expected I'd say that ever again, certainly not with what intel has been doing lately. Base clock is just 3300MHz, 4-core turbo is 3600MHz. That means overclocking it to 4400MHz/1.38v is an 22,2% increase. In comparison, it'd take 7700K (4.4GHz turbo on 4 cores) 5376MHz to reach 22,2% oc. You can overclock EDRAM on 5775c as well, and that will come at no power/temperature penalty and give you very nice returns in games. And it gets even sweeter when my tests clearly show the returns from using/overclocking EDRAM are higher the higher cpu frequency is as long as you're not GPU bottlenecked. All that at incredibly low temperatures and power consumption. During the testing period I ran cpu intensive games at low preset with additional voltage and clockspeed, CPU load was constantly +90% yet the CPU stayed in low 50s/high 40s and the fan on D15S was only spinning at 750rpm. Compared to 4790K I had that is just ridiculous for summer temperatures. Devil's Canyon at 4.6GHz 1.30v would hit 70s easily at higher fan speed.

As for oc, I've seen few of their owners claim to hit a wall at 4.0 GHz even with 1.37v. Most will do 4.2GHz at 1.35v-1.37v. Quite a few can do 4.3-4.4GHz at below 1.4v like mine does. Creme de la creme will do 4.5-4.6GHz but don't count on that, that's huge luck in sillicon lottery. Pumping up to 1.4v is pretty much nothing for those CPUs, like I said the temperatures are very,very moderate.
If you read the broadwell-c owner's club I posted a link to, you'll see people pumping 1.45v on water without delid while maintaining pretty reasonable temperatures, which is just impossible for devil's canyon or kaby lake.
Especially if you do what I did, exchanged 2400MHz CL11 1.65v sticks for slower but low latency 1600 CL8 1.35v and run them at 2133 CL9 1.512v. This way you're not contributing that much to temperature rise under load as 1.65v would do.

I did my tests at 1.3v 4200MHz cause I was 100% sure that was stable. 4.4GHz 1.38v has been running fine for some time but for the tests I did here I wanted to run something that I'd be 100% sure is stable. Temps is game rise about 5 degrees every 0.5v so expect low 40s at stock 1.2v, low 50s at 1.3v, high 50s/low 60s at 1.38v. All depends on ambient and fan setting tho. I can run games at 1.38v with low noise adapter (500/600rpm) but that gives me 67 degrees. Still very neat tho considering that like 30-40% of D15S fan speed.

 

[EarthDog]

How doesnt overclocking edram raise power or temps? Physics and all??

If ypu are going to test temps and power, you may want to move to a more consistent testing method like a stress test of some sort. Gaming, regardless if the cpu is at 90%, is way less stressful than stress testing.... temps clearly show that.

 

[cucker tarlson]

How doesnt overclocking edram raise power or temps? Physics and all??

If ypu are going to test temps and power, you may want to move to a more consistent testing method like a stress test of some sort. Gaming, regardless if the cpu is at 90%, is way less stressful than stress testing.... temps clearly show that.

The increase must be quite low so it doesn't produce any noticeable changes to temperature and draw readings. EDRAM is actually not that fast (well, it's actually quite slow compared to DDR4 or even DDR3) cause more it's low-latency oriented, and I'm not adding any voltage to it either. I'd be more surprised to see any considerable power draw differences than not see them.

I see you're quite being sour calling my reporting of gaming temps inconsistent. What's wrong with using game temps as reference unless you're just nitpicking. It's only inconsistent if I run too few test and/or have no idea of what I'm actually reporting. And by the way I mentioned IBT temps in my post as well so what is it exactly that your post was supposed to contribute the conversation ? Stress testing with IBT is producing power draw that I'd never,NEVER see in gaming, which is the topic of this thread,top of the page,capital letters.

 

[EarthDog]

Wow, people are just... wow... hackles go up at the slightest sign of questioning anything... What. The. Hell... peeps.

Anyway, cucker, all I was saying is that it doesn't make sense overclocking doesn't raise temps... because Physics. This may come down to the testing method or many other variables including not noticeable increases. I was saying to use something MORE CONSISTENT, like a stress test. That's all. Gaming temps, even in the same game can be all over the place. Variances are higher with variable loads.

Sorry I missed IBT... though it wasn't in post 57 which I responded to.

Geeeeeeeeeeeeeeeeezus Christ........

 

[cadaveca]

What's wrong with using game temps as reference unless you're just nitpicking.

Games do not provide consistent loads and typically do not use all of the resources available in a way as to be able to accurately draw conclusions about power usage. You can at best measure the peak over time, but even this peak measurement will be inconsistent.

 

[cucker tarlson]

Games do not provide consistent loads and typically do not use all of the resources available in a way as to be able to accurately draw conclusions about power usage. You can at best measure the peak over time, but even this peak measurement will be inconsistent.

Agreed. If you're testing different sequences/locations/games. Not if you're comparing same location in same game. That actually does reflect changes in power/temperatures pretty well in a standard use.

Wow, people are just... wow... hackles go up at the slightest sign of questioning anything... What. The. Hell... peeps.

Anyway, cucker, all I was saying is that it doesn't make sense overclocking doesn't raise temps... because Physics. This may come down to the testing method or many other variables including not noticeable increases. I was saying to use something MORE CONSISTENT, like a stress test. That's all. Gaming temps, even in the same game can be all over the place. Variances are higher with variable loads.

Sorry I missed IBT... though it wasn't in post 57 which I responded to.

Geeeeeeeeeeeeeeeeezus Christ........

I wish people while reading this thread would use 1/5th of the effort I tried to use while testing to ensure it produces fairly consistent results, not worthless crap. testing different settings in the exact same game locations in ALL I DO in this thread.


You believe stress test and built-in benchmarks, I prefer testing gaming loads and in-game scenarios. Fair enough, opinions vary, you're right in your way just as I'm right in mine. Trust me, I'm doing this for myself more than anyone, just sharing here to see other 5775c owenrs share their, but that's a bit of a downer so far.

 

[EarthDog]

Stress testing with IBT is producing power draw that I'd never,NEVER see in gaming, which is the topic of this thread,top of the page,capital letters.

Considering you brought up the point of power and temps...I find this post curious. I(we) are just responding to what you put out there and trying to help you confirm what you are saying by using a more consistent method. Physics tells us temsp will increase and power use. Just because you are not seeing it, doesn't mean it isn't there. Not a big deal, just saying.

Sorry if that didn't "contribute to the conversation" trying to get clarity on the point you are made...

EDIT: As I side note, I am a long time reviewer/editor (smaller website - freelance review for a huge one now), and Dave reviews motherboards here. It's not like we haven't done this before and know. That isn't to say we can't be wrong... certainly, it happens. I learn just like others do! Just sharing what we know from past testing and established norms. You could be spot on, but unfortunately, due to how games work, the variance is a lot larger than with stress testing and then the physics thing... so, I made a suggestion. Im not sour, not trying to be combative... nothing like that.

I'll leave it alone.

 

[cucker tarlson]

Well I guess I should report what I'm NOT seeing, that would be consistent according to you.

/just pulling your leg /

And how's IBT method more consistent when measuring EDRAM performance in games ? Does IBT even take advantage of EDRAM cache ? And how is it anywhere near to being related to gaming performance ?

 

[EarthDog]

Power and temps cucker... power and temps. I'm not going down this rabbit hole though cucker. Sorry I even said a word... wow. SMFH.

If you want to see if eDRAM is used in stress testing, enable and disable it like you do for games and see across a couple of tests (IBT, P95, Intel XTU, OCCT, etc). I am not certain either, but imagine it would make use of it.

 

[cucker tarlson]

LOL you imagine. Well, it doesn't. Not to any noticeable extent that would be worth proving. GFlops score in IBT stays the same. But I already knew that. It's just I never mentioned before as it would be irrelevant to the thread,which again, is entitled edram and performance in games, and as you can see I can be quite anal when people make it about whatever else THEY want it to be about.

I guess I'll make another one. "does 5775c get hotter in stress testing than in games ? Gee, I don't know but that's what we're about to find out here. What I can tell you for sure is that it's cucking a lot more power."

 

[EarthDog]

Yes, anal... no doubt. I just made a suggestion... jesus christ man..

Just because performance doesn't go up, doesn't mean it isn't being used either. This is were fly by night testing gets poeple in trouble... assumptions as a base. You may be right, but it is worth testing, to me. Obviously not to you, and not in this thread.

Good luck. Unsubscribed.

 

[Vayra86]

Power and temps cucker... power and temps. I'm not going down this rabbit hole though cucker. Sorry I even said a word... wow. SMFH.

If you want to see if eDRAM is used in stress testing, enable and disable it like you do for games and see across a couple of tests (IBT, P95, Intel XTU, OCCT, etc). I am not certain either, but imagine it would make use of it.

This would be interesting to see!

In other news, lets try to get along... There is mutual interest here and yet its turning into a combat zone... RELAX

 

[cucker tarlson]

Yes, anal... no doubt. I just made a suggestion... jesus christ man..

Just because performance doesn't go up, doesn't mean it isn't being used either. This is were fly by night testing gets poeple in trouble... assumptions as a base. You may be right, but it is worth testing, to me. Obviously not to you, and not in this thread.

Good luck. Unsubscribed.

Wow I'm glad you actually got it man!

 

[Kursah]

Let's keep the drama to yourselves folks, or take it to PM if any of you insist on dragging it on.

I am pretty sure as adults here that everyone can, or should be able to accept and even welcome criticism and feedback without getting defensive and aggressive about it. Let's all treat each other with respect, that is not a request that is part of the rules being a member here. Feel free to review the TPU guidelines, link is in my sig.

There have been some good points made all around, that's one of the nice things about TPU. We have experience at all levels and all sorts of interesting topics. Let's get along, or infractions will be earned and threads will be closed.

 

[cucker tarlson]

Yeah,we're cool, but not all criticism is constructive criticism.

The thread is pretty much dead anyway IMO,sad to admit that. I was expecting others to pitch in with their results rather than people nitpicking about irrelevant details in mine.